Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

02 Whole

Download as pdf or txt
Download as pdf or txt
You are on page 1of 260

Intrauterine Influences on

Obesity and Insulin Resistance in

Pre-pubertal Children

Oana Maftei

M.D., F.Rom.C.P.

Discipline of Public Health, School of Population Health and Clinical Practice

Discipline of Obstetrics and Gynaecology, School of Paediatrics and Reproductive Health

Faculty of Health Sciences, University of Adelaide

Australia

December 2011

A thesis submitted in partial fulfilment of the requirements for the degree of Doctor of Philosophy
“Truth lies all around us, but is only revealed to those who search for it.”

Nicolae Iorga (1871-1940)


This thesis is dedicated to my mother for her immense love, trust, support and encouragement not
only through the PhD candidature years, but throughout my entire life. I have learnt from her wisdom
that giving anything less than all I could give meant sacrificing a gift.
Table of contents

List of tables ............................................................................................................................ix


List of figures ........................................................................................................................... xii
Abstract .......................................................................................................................... xiv
Declaration ......................................................................................................................... xvii
Acknowledgements................................................................................................................. xviii
Abbreviations and acronyms .....................................................................................................xx
Conference presentations........................................................................................................ xxii

Chapter 1 Introduction

1.1 Developmental origins of health and disease .......................................................................1


1.2 Evidence of programming of childhood obesity and insulin resistance by maternal
body size and glucose intolerance during pregnancy .........................................................2
1.3 Research questions and hypotheses .....................................................................................4
1.4 Thesis objectives .......................................................................................................................5
1.5 Overview of the thesis...............................................................................................................6
1.6 Statement of contribution.........................................................................................................7

Chapter 2 Theoretical framework

2.1 Early life exposures and later development ..........................................................................8


2.2 Conceptualising pre-birth exposures .................................................................................. 11
2.2.1 Maternal pre-pregnancy obesity ........................................................................................ 12
2.2.1.1 Pathophysiology...............................................................................................................................12
2.2.1.2 Public health significance – Prevalence data .................................................................................14
2.2.1.3 Adverse health outcomes of maternal pre-pregnancy obesity ......................................................15
2.2.1.4 Clinical response to the epidemic of maternal obesity...................................................................17
2.2.2 Glucose tolerance status during pregnancy.................................................................... 19
2.2.2.1 Glucose regulation during pregnancy .............................................................................................19
2.2.2.2 Screening and diagnostic tests for glucose intolerance during pregnancy ..................................21

iv
2.2.2.3 Defining categories of gestational glucose intolerance based on oral glucose challenge test and
oral glucose tolerance test...............................................................................................................23
2.2.2.4 Risk factors for glucose intolerance during pregnancy..................................................................24
2.2.2.5 Public health significance – Prevalence data .................................................................................26
2.2.2.6 Adverse health outcomes of gestational glucose intolerance .......................................................29
2.2.2.7 Management of gestational glucose intolerance............................................................................32
2.2.3 Gestational weight gain....................................................................................................... 33
2.2.3.1 Pathophysiology...............................................................................................................................34
2.2.3.2 Public health significance - Prevalence data..................................................................................35
2.2.3.3 Adverse health outcomes of excessive gestational weight gain ...................................................36
2.2.3.4 Clinical response - Recommended gestational weight gain..........................................................37
2.2.4 Interrelations between maternal pre-pregnancy BMI, glucose tolerance status during
pregnancy and gestational weight gain ............................................................................ 37
2.3 Specific metabolic consequences in children.................................................................... 40
2.3.1 Obesity, body composition and fat pattern in children .................................................. 40
2.3.1.1 Public health significance – Prevalence data .................................................................................42
2.3.1.2 Adverse health outcomes of childhood obesity..............................................................................44
2.3.1.3 Measures of obesity.........................................................................................................................46
2.3.2 Insulin resistance in children ............................................................................................. 54
2.3.2.1 Public health significance – Prevalence data .................................................................................56
2.3.2.2 Adverse health outcomes of insulin resistance ..............................................................................57
2.3.2.3 Assessment of insulin resistance in children..................................................................................58
2.4 Evidence of early origins of childhood obesity and insulin resistance ......................... 61
2.4.1 Intrauterine programming of obesity and insulin resistance by maternal pre-
pregnancy obesity................................................................................................................ 61
2.4.2 Intrauterine programming of obesity and insulin resistance by maternal glucose
intolerance during pregnancy ............................................................................................ 70
2.4.3 Intrauterine programming of obesity and insulin resistance by maternal gestational
weight gain ............................................................................................................................ 88
2.5 Summary .................................................................................................................................. 97

Chapter 3 Methodology

3.1 The Generation 1 cohort ........................................................................................................ 98


3.1.1 Study design and overall aim ............................................................................................. 98
3.1.2 Baseline sampling process and selection criteria .......................................................... 99

v
3.1.3 Baseline data collection .................................................................................................... 100
3.1.4 Participation until this follow-up ...................................................................................... 101
3.2 Data collection for the follow-up at 9-10 years of age..................................................... 102
3.2.1 Approaching participant families..................................................................................... 103
3.2.2 Procedures undertaken with participant children ......................................................... 103
3.2.2.1 Interview..........................................................................................................................................103
3.2.2.2 Anthropometric measurements and assessment of body composition ......................................104
3.2.2.3 Fasting blood samples ...................................................................................................................105
3.2.3 Ethical considerations ....................................................................................................... 106
3.3 Data management ................................................................................................................. 107
3.4 Defining variables of interest .............................................................................................. 109
3.4.1 Intrauterine exposures....................................................................................................... 109
3.4.1.1 Maternal pre-pregnancy BMI.........................................................................................................109
3.4.1.2 Maternal glucose tolerance status during pregnancy ..................................................................110
3.4.1.3 Maternal gestational weight gain...................................................................................................111
3.4.2 Outcomes in children at the age of 9-10 years .............................................................. 112
3.4.2.1 Child BMI z-score...........................................................................................................................113
3.4.2.2 Child percentage body fat..............................................................................................................113
3.4.2.3 Child waist-to-height ratio ..............................................................................................................113
3.4.2.4 Child insulin resistance ..................................................................................................................113
3.4.3 Potential confounders ....................................................................................................... 114
3.5 Analysis plan ......................................................................................................................... 115
3.6 Summary ................................................................................................................................ 118

Chapter 4 Results

4.1 Descriptive statistics............................................................................................................ 119


4.1.1 Intrauterine exposures....................................................................................................... 119
4.1.1.1 Maternal pre-pregnancy BMI.........................................................................................................119
4.1.1.2 Maternal glucose tolerance status during pregnancy ..................................................................121
4.1.1.3 Maternal gestational weight gain...................................................................................................125
4.1.1.4 Interrelations between maternal pre-pregnancy BMI, glucose tolerance status during pregnancy
and gestational weight gain ...........................................................................................................126
4.1.2 Outcomes in children at the age of 9-10 years .............................................................. 129
4.1.2.1 Child BMI z-score...........................................................................................................................130
4.1.2.2 Child percentage body fat..............................................................................................................131

vi
4.1.2.3 Child waist-to-height ratio ..............................................................................................................132
4.1.2.4 Child insulin resistance ..................................................................................................................133
4.1.3 Potential confounders ....................................................................................................... 135
4.1.4 Non-participation assessment.......................................................................................... 137
4.2 Inferential statistics .............................................................................................................. 140
4.2.1 Maternal pre-pregnancy BMI and child outcomes......................................................... 141
4.2.1.1 Maternal pre-pregnancy BMI and child BMI z-score....................................................................141
4.2.1.2 Maternal pre-pregnancy BMI and child percentage body fat.......................................................142
4.2.1.3 Maternal pre-pregnancy BMI and child waist-to-height ratio .......................................................144
4.2.1.4 Maternal pre-pregnancy BMI and child insulin resistance...........................................................146
4.2.2 Maternal glucose tolerance status during pregnancy and child outcomes .............. 148
4.2.2.1 Maternal glucose tolerance status during pregnancy and child BMI z-score.............................148
4.2.2.2 Maternal glucose tolerance status during pregnancy and child percentage body fat................151
4.2.2.3 Maternal glucose tolerance status during pregnancy and child waist-to-height ratio ................154
4.2.2.4 Maternal glucose tolerance status during pregnancy and child insulin resistance ....................156
4.2.3 Maternal gestational weight gain and child outcomes ................................................. 159
4.2.3.1 Maternal gestational weight gain and child BMI z-score .............................................................159
4.2.3.2 Maternal gestational weight gain and child percentage body fat ................................................160
4.2.3.3 Maternal gestational weight gain and child waist-to-height ratio.................................................162
4.2.3.4 Maternal gestational weight gain and child insulin resistance.....................................................163
4.2.4 Two-way interactions between the three intrauterine exposures of interest in relation
to child outcomes ............................................................................................................... 164

Chapter 5 Discussion and conclusion

5.1 Overview of major findings ................................................................................................. 167


5.2 Study strengths and limitations.......................................................................................... 169
5.2.1 Study design ......................................................................................................................... 169
5.2.2 Measurements ...................................................................................................................... 172
5.2.3 Analyses................................................................................................................................ 176
5.3 Relationship to studies on early origins of child obesity and insulin resistance....... 178
5.3.1 Comparison of the Generation 1 sample with previous relevant studies........................... 178
5.3.2 Maternal pre-pregnancy BMI and child outcomes .............................................................. 180
5.3.3 Maternal glucose tolerance status during pregnancy and child outcomes ........................ 182
5.3.4 Maternal gestational weight gain and child outcomes ........................................................ 184
5.3.5 Interpreting mediation by current body size ........................................................................ 184

vii
5.3.6 Two-way interactions between main exposures in relation to child outcomes .................. 184
5.3.7 Potential underlying mechanisms ........................................................................................ 186
5.4 Implications and recommendations for public health..................................................... 190
5.5 Future directions for research ............................................................................................ 201
5.6 Conclusion............................................................................................................................. 205

References ........................................................................................................................ 207

viii
List of tables

Table 1. National prevalence of overweight and obesity in women of all ages from selected
developed countries ............................................................................................................... 14

Table 2. National prevalence of overweight and obesity among women of reproductive age in
Australia .................................................................................................................................. 15

Table 3. Cut-off values for plasma glucose levels recommended for the diagnosis of gestational
diabetes in oral glucose tolerance test .................................................................................. 23

Table 4. Prevalence and incidence rates of gestational diabetes reported in Australia .................... 27

Table 5. 2009 Institute of Medicine recommendations for gestational weight gain for singleton
pregnancies............................................................................................................................. 34

Table 6. Summary of longitudinal studies examining the influence of maternal pre-pregnancy


overweight / obesity on child obesity and insulin resistance ................................................ 64

Table 7. Summary of longitudinal studies examining the influence of maternal glucose tolerance
status during pregnancy on child obesity and insulin resistance ......................................... 75

Table 8. Summary of longitudinal studies examining the influence of maternal gestational weight
gain on child obesity............................................................................................................... 90

Table 9. Main strategies to prevent loss to follow-up and enhance response rates........................ 102

Table 10. Timing of last available measure of weight during pregnancy............................................ 112

Table 11. Maternal body size before pregnancy ................................................................................. 120

Table 12. Maternal pre-pregnancy BMI status..................................................................................... 121

Table 13. Summary of plasma glucose levels at OGCT and OGTT during pregnancy for Generation 1
women................................................................................................................................... 122

Table 14. Glucose tolerance profiles during pregnancy in Generation 1 women .............................. 123

Table 15. Total gestational weight gain across categories of pre-pregnancy BMI in Generation 1
women, relative to 2009 Institute of Medicine recommendations ...................................... 125

Table 16. Maternal pre-pregnancy BMI across the spectrum of glucose tolerance during pregnancy 126

ix
List of tables

Table 17. Maternal gestational weight gain across categories of pre-pregnancy BMI ...................... 128

Table 18. Maternal gestational weight gain and glucose tolerance status during pregnancy........... 129

Table 19. Anthropometric measurements in Generation 1 children at 9-10 years ............................ 130

Table 20. Summary measures for percentage body fat in Generation 1 children ............................. 131

Table 21. Summary measures of central adiposity in Generation 1 children..................................... 132

Table 22. Summary measures for fasting glucose, fasting insulin, and HOMA-IR in Generation 1
children.................................................................................................................................. 134

Table 23. Children’s anthropometric measurements and fasting measures of glucose homeostasis at
9-10 years ............................................................................................................................. 135

Table 24. Maternal and child characteristics in the Generation 1 participants at 9-10 year follow-up
relative to the original cohort................................................................................................ 139

Table 25. Child characteristics in the Generation 1 participants at 9-10 years who provided a fasting
blood sample collection relative to all participating children at 9-10 years ........................ 140

Table 26. Child BMI and BMI z-score across categories of maternal pre-pregnancy BMI................ 142

Table 27. Estimated change in child BMI z-score in relation to maternal pre-pregnancy BMI.......... 142

Table 28. Child percentage body fat across categories of maternal pre-pregnancy BMI.................. 143

Table 29. Estimated change in child percentage body fat in relation to maternal pre-pregnancy BMI 144

Table 30. Child waist-to-height ratio across categories of maternal pre-pregnancy BMI .................. 145

Table 31. Estimated change in child waist-to-height ratio in relation to maternal pre-pregnancy BMI. 145

Table 32. Child HOMA-IR across categories of maternal pre-pregnancy BMI .................................. 146

Table 33. Estimated change in child HOMA-IR in relation to maternal pre-pregnancy BMI.............. 147

Table 34. Child BMI and BMI z-score across the categories of maternal glucose tolerance during
pregnancy ............................................................................................................................. 148

Table 35. Estimated change in child BMI z-score in relation to maternal glucose tolerance status
during pregnancy.................................................................................................................. 150

Table 36. Child percentage body fat across the categories of maternal glucose tolerance during
pregnancy ............................................................................................................................. 151

x
List of tables

Table 37. Estimated change in child percentage body fat in relation to maternal glucose tolerance
status during pregnancy....................................................................................................... 153

Table 38. Child waist-to-height ratio across categories of maternal glucose tolerance during
pregnancy ............................................................................................................................. 154

Table 39. Estimated change in child waist-to-height ratio in relation to maternal glucose tolerance
status during pregnancy....................................................................................................... 155

Table 40. Child HOMA-IR across categories of maternal glucose tolerance during pregnancy ....... 156

Table 41. Estimated change in child HOMA-IR in relation to maternal glucose tolerance status during
pregnancy ............................................................................................................................. 158

Table 42. Estimated change in child BMI z-score in relation to maternal gestational weight gain.... 160

Table 43. Estimated change in child percentage body fat in relation to maternal gestational weight
gain........................................................................................................................................ 161

Table 44. Estimated change in child waist-to-height ratio in relation to maternal gestational weight
gain........................................................................................................................................ 162

Table 45. Estimated change in child HOMA-IR in relation to maternal gestational weight gain ....... 164

Table 46. Comparison of the Generation 1 sample with previous studies ......................................... 179

xi
List of figures

Figure 1. Flowchart of participation ................................................................................................... 101

Figure 2. Confounders in relation to exposures and outcomes ....................................................... 114

Figure 3. Correlation between maternal self-reported pre-pregnancy weight and measured weight
in early pregnancy .............................................................................................................. 120

Figure 4. Maternal pre-pregnancy BMI status .................................................................................. 121

Figure 5. Results of prenatal screening for gestational diabetes in Generation 1 women ............. 124

Figure 6. Maternal pre-pregnancy BMI across the spectrum of glucose tolerance during pregnancy. 127

Figure 7. Maternal gestational weight gain in relation to pre-pregnancy BMI ................................. 127

Figure 8. Maternal gestational weight gain across categories of pre-pregnancy BMI .................... 128

Figure 9. Maternal gestational weight gain and glucose tolerance status during pregnancy......... 129

Figure 10. BMI z-score in Generation 1 children ................................................................................ 130

Figure 11. Distribution of BMI categories in Generation 1 children ................................................... 131

Figure 12. Histograms by transformation for child percentage body fat............................................ 132

Figure 13. Histograms by transformation for child waist-to-height ratio ............................................ 133

Figure 14. Histograms by transformation for child HOMA-IR............................................................. 134

Figure 15. Child BMI z-score in relation to maternal pre-pregnancy BMI.......................................... 141

Figure 16. Child percentage body fat in relation to maternal pre-pregnancy BMI............................. 143

Figure 17. Child waist-to-height ratio in relation to maternal pre-pregnancy BMI............................. 144

Figure 18. Natural logarithm of child HOMA-IR in relation to maternal pre-pregnancy BMI............. 146

Figure 19. Child BMI z-score across categories of maternal glucose tolerance during pregnancy . 149

Figure 20. Child percentage body fat across the categories of maternal glucose tolerance during
pregnancy ........................................................................................................................... 151

Figure 21. Child waist-to-height ratio across categories of maternal glucose tolerance during
pregnancy ........................................................................................................................... 154

xii
List of figures

Figure 22. Child HOMA-IR across categories of maternal glucose tolerance during pregnancy..... 156

Figure 23. Child BMI z-score in relation to maternal gestational weight gain ................................... 159

Figure 24. Child percentage body fat in relation to maternal gestational weight gain ...................... 161

Figure 25. Child waist-to-height ratio in relation to maternal gestational weight gain....................... 162

Figure 26. Child HOMA-IR in relation to maternal gestational weight gain ....................................... 163

Figure 27. Interaction between maternal glucose tolerance status during pregnancy and pre-
pregnancy BMI in relation to child waist-to-height ratio.................................................... 165

Figure 28. Interaction between maternal pre-pregnancy BMI and gestational weight gain in relation
to child waist-to-height ratio ............................................................................................... 166

xiii
Abstract

Within the paradigm of developmental origins of health and disease, an intrauterine environment that
stimulates fetal overnutrition has been found to contribute to the risk of subsequent obesity in the
offspring. There is compelling epidemiological evidence for a positive association between maternal
obesity prior to pregnancy, gestational diabetes (GD) or excessive gestational weight gain, and the
development of childhood obesity (as measured by body mass index, BMI). However, the evidence
is limited and inconsistent with respect to more specific measures of adiposity (body composition or
fat pattern) and insulin resistance in children. Furthermore, the long-term effects of maternal
borderline gestational glucose intolerance (BGGI) on the offspring have not been considered.

Therefore, I sought to examine whether maternal obesity prior to pregnancy, gestational glucose
intolerance across the entire spectrum, and gestational weight gain have deleterious effects on the
development of obesity (both global and specific measures of adiposity) and insulin resistance in
pre-pubertal children. These associations are particularly important from a public health perspective
as, once identified, they may point towards potential windows for prevention of childhood obesity and
related metabolic disorders.

This project entailed a follow-up of an existing representative, prospective birth cohort study
(Generation 1 Study, n=557) in Adelaide, South Australia, recruited during 1998-2000. At the 9-10
year follow-up, rigorous anthropometric measurements were conducted in 443 children (80% of the
original cohort), of whom 163 consented to provide a fasting blood sample for the estimation of
insulin resistance based on homeostasis model assessment (HOMA-IR). Information on intrauterine
exposures and confounders was collected from the antenatal interviews and hospital records.
Maternal age, parity, smoking, pregnancy-induced hypertension, and education at the time of
pregnancy were considered as potential confounders for all the associations of interest, and child
current BMI z-score as a potential mediator on the pathway between the intrauterine exposures and
child insulin resistance. Data were analysed using multiple linear regression and generalized linear
models.

xiv
Abstract

Maternal pre-pregnancy BMI was positively associated with all three obesity-related measures
considered in the 9-10 year-old children (BMI z-score, percentage body fat estimated by bioelectrical
impedance analysis, and waist-to-height ratio); these relationships were robust to adjustment for
potential confounders (adjusted coefficients for each one kg/m2 increase in maternal pre-pregnancy
BMI were 0.08 (95% confidence interval 0.06, 0.10) for child BMI z-score, 0.44 (95% CI 0.31, 0.58)
for percentage body fat and 0.002 (95% CI 0.002, 0.003) for waist-to-height ratio). There was no
association between maternal pre-pregnancy BMI and HOMA-IR in children (with or without
adjustment); however, when child current BMI z-score was included as a mediating variable, the
relationship between maternal pre-pregnancy BMI and child HOMA-IR was inverse and significant
(adjusted change in child HOMA-IR for each one kg/m2 increase in maternal pre-pregnancy BMI was
-0.83% (95% CI -1.63, -0.02)).

Intrauterine exposure to glucose intolerance during pregnancy (either BGGI or GD) was not
associated with any of the three obesity-related measures in children at 9-10 years. Children of
mothers who developed GD during the index pregnancy had a higher HOMA-IR; this relationship
was robust to adjustment for potential confounders (adjusted change in child HOMA-IR if exposed to
maternal GD was 42.9% (95% CI 20.9, 68.9)) and partly mediated by child current BMI z-score. No
association was found between exposure to maternal BGGI and child HOMA-IR (with or without
confounder adjustment); however, when child current BMI z-score was added as a potential mediator,
exposure to BGGI was associated with a reduction in child HOMA-IR (adjusted change in child
HOMA-IR if exposed to maternal BGGI was -17.9% (95% CI -29.9, -3.96)).

There were no significant associations between maternal gestational weight gain and any of the
outcome measures of interest in unadjusted models. However, adjustment for pre-pregnancy BMI
led to a positive association between gestational weight gain and child BMI z-score (adjusted
changes in child BMI z-score for each one kg increase in maternal gestational weight gain was 0.032
(95% CI 0.007, 0.057)). Gestational weight gain was not associated with child insulin resistance,
and this did not change when child current BMI z-score was included as a potential mediator on the
pathway between gestational weight gain and child insulin resistance.

Potential two-way interactions between the main exposures were investigated in relation to all
outcomes of interest. Two significant interactions were identified: maternal pre-pregnancy BMI and
glucose tolerance status, and maternal pre-pregnancy BMI and gestational weight gain, with a
synergistic effect on child waist-to-height ratio.

xv
Abstract

These results suggest that childhood obesity and insulin resistance have origins, at least in part, in
intrauterine life, particularly in relation to maternal obesity at the time of pregnancy and GD. Further
research to differentiate between genetic, environmental and intrauterine programming is
recommended. That said, maternal pre-pregnancy BMI was the strongest predictor of child BMI z-
score, while GD appeared to have an independent effect on child insulin resistance, and both clinical
and public health actions to address these maternal factors are warranted for a range of reasons.

xvi
Declaration

I, Oana Maftei certify that this work contains no material which has been accepted for the award of
any other degree or diploma in any university or other tertiary institution and, to the best of my
knowledge and belief, contains no material previously published or written by another person, except
where due reference has been made in the text.

I give consent to this copy of my thesis, when deposited in the University Library, being made
available for loan and photocopying, subject to the provisions of the Copyright Act 1968.

I also give permission for the digital version of my thesis to be made available on the web, via the
University’s digital research repository, the Library catalogue and also through web search engines,
unless permission has been granted by the University to restrict access for a period of time.

Oana Maftei Date: 21 December 2011

xvii
Acknowledgements

It goes without saying that the research detailed in this thesis could never have occurred, had it not
been for the innumerable contributions of my supervisors, colleagues, study participants, friends and
family.

I owe an immense debt of gratitude to Associate Professor Vivienne Moore, primary supervisor, who
was willing to accept into her research group an overseas, head-strong Paediatrician with no
experience in epidemiological research and biostatistics. I thank her for her patience in guiding me
through important stages of the project, for supporting my professional development and for allowing
me a great deal of independence, which I have come to value with time. Not only has her vast
knowledge in so many and diverse areas of public health been an inspiration for me, but also her
personal kindness.

Dr Melissa Whitrow, one of my co-supervisors, deserves a big thank you for providing feedback on
all the drafts of this thesis. I have appreciated her structured way of thinking and her attention to detail.

Associate Professor Michael Davies, my other co-supervisor, has been of invaluable assistance,
particularly with the theoretical underpinning of this project, stimulating critical conversations and
providing feedback on the drafts of the thesis. I have admired his enthusiasm to discuss any aspects
of reproductive epidemiology.

Thanks are also due to Kendall Smith, study co-ordinator, for her dedication in facilitating access to
the cohort families, who were always welcoming. None of this project could have been conducted
without the commitment of the families and their willingness to contribute to research, despite their
busy and complicated lives. I am also thankful to the other fieldwork members (Stephanie
Champion, Trish Malbon, Deb Roffe and Amanda Hillgrove) for their contribution to data collection. I
thank Dr Miles DeBlasio for running the blood assays necessary for this research.

Special thanks go to Professor Phil Ryan, whose clarity in teaching biostatistics and ability to simplify
complex statistical issues enabled me to conduct the analyses presented in this thesis. I gratefully
acknowledge Dr Lynne Giles for backing up some of the statistical analyses contained in this thesis
and also Suzanne Edwards for data cleaning and imputation of some missing data. My colleagues in
the Life course and Intergenerational Health research group, all extraordinary people, Dr Emily Steele,

xviii
Acknowledgements

Dr Alice Rumbold, Dr Jennifer Marino, Dr Wendy March, and Malinda Steenkamp have contributed
by constant emotional support.

I am indebted to Maria Gardiner, for mentoring me in the last year of my candidature, hence
‘turbocharging’ the writing of this thesis. I will always remember her words: ‘writing is critical
thinking’. I thank other current PhD students for their friendship and for valuable feedback on several
drafts obtained from those involved in the Thesis Writing Group.

My gratitude also extends to Dr Alexia Pena, who was extremely generous in offering me a part time
job in a clinical research project, believed in me at challenging times when not many other people did
and encouraged me to follow my dreams despite my disillusions. Her ‘can-do’ attitude contributed
considerably to finalising this project within a reasonable timeframe. I am likewise grateful to my
best friend, Crina for her constant support.

I also acknowledge the University of Adelaide for granting me the Adelaide Scholarship International,
which allowed me to have a decent life throughout my candidature.

For what I am today, I owe enormous gratefulness to my parents, Elena and Ion, for embodying in
me the virtues of scholarship and for educating me to put my dreams into action in order to become
successful in life. My greatest debt of appreciation is owed to my husband Bob, whose love, care,
thoughtfulness and understanding empowered me to finalise this research. I cannot imagine how
impoverished my existence would be without him.

xix
Abbreviations and acronyms

%BF percentage body fat


ACHOIS Australasian Carbohydrate Intolerance Study in Pregnancy
ACOG American College of Obstetricians and Gynecologists
ADA American Diabetes Association
ADIPS Australasian Diabetes in Pregnancy Society
ALSPAC Avon Longitudinal Study of Parents and Children
BGGI borderline gestational glucose intolerance
BIA bioelectrical impedance analysis
BMI body mass index
BW birth weight
CDC Centers for Disease Control and Prevention
CI confidence interval
CT computed tomography
DXA dual energy X-ray absorptiometry
EPOCH Exploring Perinatal Outcomes among Children
FFM fat-free mass
FGIR fasting glucose-to-insulin ratio
FM fat mass
FSIVGTT frequently sampled intravenous glucose tolerance test
FTO fat mass and obesity associated gene
GD gestational diabetes
GLM generalised linear model
GWG gestational weight gain
HAPO Hyperglycemia and Adverse Pregnancy Outcomes
HbA1c haemoglobin A1c
HDL high-density lipoprotein
HOMA-IR homeostasis model assessment of insulin resistance
IGF insulin-like growth factor
IGT impaired glucose intolerance
IL interleukin

xx
Abbreviations and acronyms

IOM Institute of Medicine


IOTF International Obesity Task Force
IQR interquartile range
IR insulin resistance
LDL low-density lipoprotein
LMS lambda-mu-sigma
MRI magnetic resonance imaging
NDDG National Diabetes Data Group
NGT normal glucose tolerance
NZSSD New Zealand Society for the Study of Diabetes
OGCT oral glucose challenge test
OGTT oral glucose tolerance test
OR odds ratio
PAR population attributable risk
PCOS polycystic ovarian syndrome
QUICKI quantitative insulin sensitivity check index
r Pearson’s correlation coefficient
RCT randomised clinical trial
RR relative risk
SD standard deviation
SES socio-economic status
SFT skinfold thickness
TBW total body water
TNF tumour necrosis factor
TAFE Training and Further Education
WHO World Health Organization
WHtR waist-to-height ratio

xxi
Conference presentations

1. Maftei O, Whitrow MJ, Moore VM, Davies MJ. Intrauterine influences on offspring obesity in
prepubertal children - oral presentation, 7th World Congress on Developmental Origins of Health
and Disease, Portland, Oregon, USA, 18-21 September 2011

2. Maftei O, Whitrow MJ, Moore VM, Davies MJ. Intrauterine influences on insulin resistance in
prepubertal children - oral presentation (ranked in top 10%), 7th World Congress on
Developmental Origins of Health and Disease, Portland, Oregon, USA, 18-21 September 2011

3. Maftei O, Whitrow MJ, Moore VM, Davies MJ. Insulin resistance in prepubertal children in
relation to intrauterine exposure to overnutrition - oral e-poster presentation, 52nd Annual
Meeting of the European Society for Paediatric Research, Newcastle, UK, 14-17 October 2011

xxii
Chapter 1 Introduction

Childhood obesity and its related metabolic disorders, including insulin resistance, are an emerging
public health issue in developed countries. Given that weight reduction is notoriously difficult to
achieve and sustain, identifying effective preventative strategies is imperative. Relatively recent
epidemiological research has suggested that intrauterine life might provide a window for obesity
prevention. This suggestion has been based on the observation that maintaining healthy body
weight is more likely to be achieved by individuals with an optimal fetal environment, leaving aside
genetic considerations.

1.1 Developmental origins of health and disease

A growing body of evidence has shown that adverse conditions occurring during intrauterine life may
result in perturbations of fetal growth and ‘programming’ of physiological systems, including
metabolism, thereby predisposing the individual to chronic disease later in life. This constitutes the
‘fetal origins hypothesis’, also known as the ‘developmental origins of health and disease’
hypothesis, originally proposed by Barker in 1980s (Barker and Osmond 1986), and subsequently
supported by many other studies, both in humans and animal models, with various exposures and
outcomes of interest (Barker 1994; Barker 1995; Barker 2003; Gillman 2005; McMillen and Robinson
2005; Oken and Gillman 2003).

Several studies have indicated that changes in maternal metabolic milieu during pregnancy could
play a major role in fetal development, with subsequent effects on the size at birth. Both low and
high birth weight for a given gestational age have been associated with a greater risk of obesity and
related metabolic outcomes later in life (U-shaped curve) (Gillman 2005; Jovanovic 2000), but, as
low birth weight has become less common in developed countries, the research focus has shifted
towards fetal overnutrition.

1
Introduction

A number of intrauterine conditions have been identified as independent risk factors for greater size
at birth, including maternal characteristics brought to the pregnancy (such as body weight status)
and occurring during pregnancy (such as gestational glucose intolerance and weight gain). Data is
scarce regarding the long-term influence of these potentially modifiable risk factors on offspring
weight status, body composition, fat distribution and insulin sensitivity, independently, cumulatively,
or in complex interactions.

The current project sits within the paradigm of developmental origins of health and disease, aiming to
investigate the influence that maternal obesity prior to pregnancy, gestational glucose intolerance,
and/or gestational weight gain have on the development of obesity, increased percentage of body
fat, central adiposity and insulin resistance in the child.

1.2 Evidence of programming of childhood obesity and insulin resistance by


maternal body size and glucose intolerance during pregnancy

There is consistent epidemiological evidence of a positive association between maternal weight


status at the time of pregnancy and child global obesity, as measured by body mass index (BMI)
(Boerschmann et al. 2010; Lawlor et al. 2007b; Li et al. 2005; Pirkola et al. 2010; Reilly et al. 2005;
Whitaker 2004), although there are marked variations in the estimates reported on this association.
In contrast, the evidence is scarce with respect to the influence of maternal pre-pregnancy BMI on
more specific measures of adiposity in children, such as percentage body fat and fat pattern, and on
insulin sensitivity in childhood.

Gestational glucose intolerance may occur during the second half of pregnancy in those women
whose insulin secretion is insufficient to compensate for the increased insulin resistance that
characterises any pregnancy. Various degrees of glucose intolerance may develop, ranging from
severe (falling under the category of gestational diabetes) to milder (which does not meet the
diagnostic criteria for gestational diabetes and which in this project is defined as borderline
gestational glucose intolerance). There is compelling evidence regarding the detrimental effect of
gestational diabetes on fetal development and some recent studies have shown that even lower
levels of glucose intolerance may adversely affect intrauterine growth (HAPO Study Cooperative
Research Group 2008). However, the long-term effects of prenatal exposure to various degrees of

2
Introduction

maternal gestational glucose intolerance on health outcomes in the offspring subsequent to birth
have been far less documented.

There is relatively abundant evidence to suggest an association between maternal gestational


diabetes and the long-term risk of global obesity in the child, but it is inconsistent, with some studies
identifying an increased risk of obesity obesity (Boerschmann et al. 2010; Catalano et al. 2009a;
Egeland and Meltzer 2010; Krishnaveni et al. 2010; Silverman et al. 1991; Silverman et al. 1998),
while others not (Gillman et al. 2003; Jeffery et al. 2006; Lawlor et al. 2010; Pirkola et al. 2010;
Whitaker et al. 1998). Data regarding the association between maternal glucose intolerance during
pregnancy and body composition or fat distribution of the child, while more limited, are also
conflicting. After confounder adjustment, exposure to gestational diabetes was positively associated
with offspring measures of adiposity in some studies (Catalano et al. 2003b; Egeland and Meltzer
2010; Krishnaveni et al. 2010; Wright et al. 2009), but not in other studies (Catalano et al. 2009a;
Pirkola et al. 2010). Furthermore, there is little and inconclusive research on the association
between exposure to a diabetic intrauterine environment and insulin resistance during childhood
(Boerschmann et al. 2010; Catalano et al. 2009a; Egeland and Meltzer 2010; Jeffery et al. 2006;
Krishnaveni et al. 2010).

It has been recently suggested that persistent increases in maternal plasma glucose levels during
pregnancy that do not fulfil the diagnostic criteria for gestational diabetes may also result in the
programming of child obesity and insulin resistance. These associations have been addressed in
very few studies in non-high risk populations, with mixed findings (Chandler-Laney et al. 2011;
Egeland and Meltzer 2010; Hillier et al. 2007; Lawlor et al. 2010). There is no study investigating the
long-term influence of maternal glucose intolerance during pregnancy across the entire spectrum of
severity on child obesity, body composition, fat pattern and insulin resistance before puberty, with
thorough adjustment for confounders.

Recent studies have suggested that, in addition to maternal obesity and glucose intolerance during
pregnancy, gestational weight gain may also affect the intrauterine environment, with increases to
the risk of obesity later in life and potentially obesity-related metabolic disorders. With one exception
(Whitaker 2004), all published studies have found a positive association between maternal weight
gain during pregnancy and obesity in children. The evidence is very limited with regard to the
influence of maternal weight gain during pregnancy on more specific measures of adiposity in the
child, with inconsistent results (Fraser et al. 2010; Oken et al. 2007). No studies have been
published addressing the potential association between maternal excessive weight gain in

3
Introduction

pregnancy and child insulin resistance. Moreover, no studies have examined how maternal
gestational weight gain across the entire spectrum influences child body composition, fat distribution
and insulin resistance before puberty.

In summary, the existing literature on these relatively new questions about how predisposition to
childhood obesity and insulin resistance arises has a number of limitations including lack of data for
some associations, poor confounder adjustment, and non-representative samples. In addition, most
studies used categorical variables, thus limiting opportunities to determine the effect of small shifts in
a particular variable on the outcome of interest. A full summary and critique of this literature is
provided in Section 2.4.

1.3 Research questions and hypotheses

Numerous factors are known or suspected to affect the risk of developing obesity and related
metabolic disorders in children, including genetic, epigenetic and postnatal environmental factors,
but, as a whole, they are beyond the scope of this project. The current study aims to shed light on
the long-term influence of early life factors, in particular maternal obesity prior to pregnancy,
gestational glucose intolerance (both gestational diabetes and borderline gestational glucose
intolerance), and gestational weight gain on childhood obesity, body composition, fat distribution, and
insulin resistance, within a contemporary Australian cohort of 9-10 year old children.

Consistent with the developmental origins of health and disease paradigm, I hypothesise that:

1. Maternal pre-pregnancy obesity is positively associated with child global obesity (as measured
by BMI), percentage body fat, central adiposity and insulin resistance in pre-pubertal children;
the relationship with insulin resistance is independent of child current body size.

2. Maternal glucose intolerance during pregnancy across its entire spectrum of severity (from
mild glucose intolerance to gestational diabetes) is positively associated with child global
obesity (as measured by BMI), percentage body fat, central adiposity and insulin resistance in
pre-pubertal children; the relationship with insulin resistance is independent of child current
body size.

4
Introduction

3. Maternal weight gain during pregnancy is positively associated with offspring global obesity
(as measured by BMI), percentage body fat, central adiposity and insulin resistance in pre-
pubertal children; the relationship with insulin resistance is independent of child current body size.

4. Maternal pre-pregnancy obesity, glucose intolerance during pregnancy, and gestational weight
gain interact and synergistically increase offspring global obesity (as measured by BMI),
percentage body fat, central adiposity and insulin resistance in pre-pubertal children.

For each association considered, several potential confounders (maternal age, parity, smoking,
pregnancy-induced hypertension, and socioeconomic status) will be adjusted for.

1.4 Thesis objectives

This thesis aims to investigate the effects of intrauterine exposure to overnutrition due to maternal
obesity, gestational glucose intolerance and gestational weight gain, on global obesity, body
composition, fat pattern and insulin resistance among pre-pubertal children in a prospective birth
cohort study in Adelaide, South Australia. The specific objectives of this project are:

1. To access and assess historical measurements of


a. maternal weight status prior to pregnancy;
b. maternal glucose tolerance status during pregnancy;
c. maternal gestational weight gain
in a representative cohort of women in Adelaide, South Australia;

2. To assess
a. global obesity (measured by BMI);
b. body composition (percentage body fat);
c. fat pattern (waist-to-height ratio);
d. insulin resistance (homeostasis model assessment)
in pre-pubertal (9-10 year-old) children in a representative prospective birth cohort in
Adelaide, South Australia;

3. To investigate the associations between maternal pre-pregnancy BMI and child global obesity,
percentage body fat, waist-to-height ratio and insulin resistance at the age of 9-10 years;

5
Introduction

4. To investigate the associations between maternal glucose intolerance during pregnancy


across the spectrum and child global obesity, percentage body fat, waist-to-height ratio and
insulin resistance at the age of 9-10 years;

5. To investigate the associations between maternal weight gain during pregnancy and child
global obesity, percentage body fat, waist-to-height ratio and insulin resistance at the age of
9-10 years;

6. To assess interactions between maternal pregnancy characteristics in relation to offspring


global obesity, percentage body fat, waist-to-height ratio and insulin resistance at the age of
9-10 years.

1.5 Overview of the thesis

Following the introduction and statement of hypotheses and objectives in Chapter 1, Chapter 2
describes a theoretical framework to which this project seeks to contribute. It starts by outlining the
broader conceptual framework of early life origins of health and disease, and then discusses each
exposure and outcome of interest with an emphasis on their measurement and public health
significance. Finally, the chapter presents a summary and critique of the published literature
concerning the influences of maternal obesity, glucose intolerance and weight gain during pregnancy
upon childhood obesity and related metabolic disorders, delineating limitations of each study and
identifying the gaps in this area of epidemiological knowledge.

Chapter 3 describes the study methods, including the study design, relevant aspects of data
collection, the derivation of exposure and outcome variables examined in this thesis, and the
statistical methods applied. The Generation 1 cohort, upon which this project is based, is also briefly
outlined.

Chapter 4 presents the findings arising from the analyses, including descriptive statistics pertaining
to the study participants (both pregnancy-related data and data on anthropometry and glucose-
insulin homeostasis in children) and the results for the investigations concerning the intrauterine
influences of maternal obesity, gestational glucose intolerance and gestational weight gain upon
childhood obesity, body composition, fat pattern and insulin resistance.

6
Introduction

Chapter 5 discusses the results of each association of interest in context with previous studies,
outlines study strengths and limitations, describes potential underlying mechanisms, and presents
implications arising from this work and recommendations for future research in the field of early
origins of childhood obesity and insulin resistance.

1.6 Statement of contribution

My doctoral research was partially based on data collected from a cohort of pregnant women from
1998 through to 2000. The activities for which I was responsible under my supervisors’ guidance
included:

1. Conceptualisation of the specific research questions;

2. Participation in fieldwork at the 9-10 year follow-up, from June 2008 through to May 2010,
interviewing and performing anthropometric measurements in 144 children, representing one
third of participants;

3. Data management (except for imputation of some missing data) and data analysis;

4. Presentation and interpretation of results;

5. Discussion and implications of findings.

7
Chapter 2 Theoretical framework

This chapter describes the broad concept of life course which contextualises the current project and
its public health significance. Section 2.1 outlines the principles underlying early life origins of health
and disease, with an emphasis on the long-term influence of conditions operating in utero on the
development of obesity and related metabolic disorders in the offspring before puberty. The
modifiable intrauterine factors that this project focuses on, namely maternal obesity prior to
pregnancy, gestational glucose intolerance across the entire spectrum of severity, and excessive
weight gain during pregnancy are presented in Section 2.2. Each of them is described in terms of
pathophysiology, demographic trends, adverse outcomes for the mother and the child, and the
clinical response thus far. Section 2.3 describes the outcomes of interest for this project, specifically
global obesity, body composition, fat pattern and insulin resistance in children. Details on
demographic trends, adverse health outcomes and methods of assessment are considered for each
of these outcomes. Epidemiological evidence from previous international studies addressing
intrauterine programming of childhood obesity and insulin resistance is summarised and critiqued in
Section 2.4, flagging the need for an investigation based on a contemporary cohort of pre-pubertal
children.

2.1 Early life exposures and later development

In recent years, potential associations between factors acting at various stages of life and
development of certain chronic diseases have started to be explored within the conceptual
framework of the life course (Kuh and Ben-Shlomo 2004). The life course approach to chronic
disease emerged in late 1980s from the necessity to fill the gap in explaining the increasing
occurrence of cardiovascular disease, by challenging the aetiological model that emphasized the role
of adult lifestyle (smoking, diet, and lack of exercise) (Barker and Osmond 1986). This approach
refers to the study of long term effects of exposures occurring in different stages of life, from

8
Theoretical framework

conception through to adulthood, on later health status or development of disease (Kuh and Ben-
Shlomo 2004). Exposures may include biological and psychosocial events acting before conception,
during intrauterine life, infancy, childhood, adolescence, or in early adult life, which “independently,
cumulatively and interactively” influence disease risk later in life (Kuh and Ben-Shlomo 2004). The
life course approach has been extended to cover intergenerational effects, focussing on those
factors that are transmitted from one generation to another (Kuh and Ben-Shlomo 2004).

Within this broad framework sits the paradigm of developmental origins of health and disease, which
focuses on early stages of life characterised by increased levels of plasticity and vulnerability, such
as the prenatal and early postnatal periods, as potential determinants of the health trajectory later in
life (Gillman 2005). Developmental plasticity is the phenomenon by which a given genotype may
result in different structural and functional states in response to various environmental exposures
occurring at certain stages of development (Gluckman and Hanson 2007). These permanent
changes in function and/or structure of organs are thought to be associated with permanent
alterations in gene expression regulated by epigenetic factors (such as DNA methylation and histone
modification).

From a public health point of view, the life course approach to chronic disease and particularly the
developmental origins to health and disease paradigm are important for implementing health
promotion and disease prevention interventions in early life.

Three specific periods in early life have been hypothesised as critically important for the
development of obesity before puberty and may indicate windows for prevention: the prenatal period,
infancy and the adiposity rebound (Gillman 2004), which occurs around the age of 5-6 years (Reilly
et al. 2005). Given the positive association between body size and insulin resistance (Lawlor et al.
2005), it seems reasonable to consider the same periods as being critical for the development of
insulin resistance in pre-pubertal children. According to the critical period model, which may underlie
programming (Lucas 1998), exposures acting in certain periods of vulnerability during early life,
characterised by rapid cell replication, marked plasticity of the systems and high sensitivity to
environmental factors (Barker 1994), which have lifelong, irreversible effects on the structure and/or
function of organs and systems (Ben-Shlomo and Kuh 2002). In contrast to the critical period
framework for obesity development and related metabolic disorders, it is also plausible that risk
factors accumulate progressively over the life span, as advocated by the accumulation of risk model
(Ben-Shlomo and Kuh 2002). These risk factors may be independent from each other or clustered,
with synergistic or antagonistic effects (Ben-Shlomo and Kuh 2002). An example of a synergistic

9
Theoretical framework

effect is the relationship between insulin resistance and low birth weight, which is particularly strong
for individuals who become obese later in life (Lawlor et al. 2005).

As intrauterine life is characterised by the most rapid growth and great vulnerability to various insults,
it is the focus of the current project to examine potential influences of prenatal environment on global
obesity (BMI), percentage body fat, fat distribution, and insulin sensitivity in the child, thus fitting in
the original conceptual framework of early origins of health and disease (‘fetal origins hypothesis’)
proposed by Barker (1995). Numerous observational studies have investigated the development of
obesity and associated metabolic disorders in relation to size at birth as a measure of convenience
(Wells et al. 2007), widely available as part of routine medical records and with established reference
norms, but which in fact is only a crude marker of intrauterine growth for a given gestational age.
Most of these studies showed a positive association with attained BMI later in life (Parsons et al.
1999) and a negative one with central adiposity (Okosun et al. 2000) and insulin resistance (Law et
al. 1995; Phillips et al. 1994). However, it has been suggested that certain factors acting during
pregnancy may have an impact on the later risk of disease even in the absence of any effect on size
at birth (Pettitt et al. 1987). Therefore, directly examining these intrauterine factors (instead of birth
size) in relation to later development and metabolism could be of more relevance.

Birth weight has often been used as a proxy for the ‘quality’ of intrauterine growth. More recently,
some studies have investigated pregnancy-related conditions that affect fetal development as
potential early life determinants for later obesity, greater adiposity, central fat distribution and insulin
resistance. Fetal overnutrition and subsequent large size at birth (macrosomia or large-for-
gestational age) have been linked to maternal pre-pregnancy obesity, to alterations in glucose-insulin
metabolism during pregnancy (from a mild degree of glucose intolerance to gestational diabetes), to
increased gestational weight gain, as well as to an unbalanced diet with high intake of carbohydrates
(Clapp 2002). Macrosomia has been recognised as a potential link between intrauterine conditions
associated with fetal overnutrition and the later development of obesity (Vohr et al. 1999).

At the other extreme of birth weight distribution, restricted fetal growth has been associated with
low maternal pre-pregnancy BMI (Godfrey et al. 1997), low gestational weight gain (as a proxy for
insufficient nutrient supply to the fetus) (Rogers 2003), poor quality maternal diet in early pregnancy
(Ravelli et al. 1976), low maternal intake of proteins during pregnancy (Godfrey et al. 1997),
pregnancy-induced hypertension (Villar et al. 2006), preeclampsia (Sibai 2008), advanced maternal
age (Odibo et al. 2006), low parity (Wanzhen et al. 2006), multiplicity (Shinwell and Blickstein 2007;

10
Theoretical framework

Taylor et al. 1998), stress (Nkansah-Amankra et al. 2010) and smoking during pregnancy (Power
and Jefferis 2002; Rogers 2003).

Given the reshaping of the birth weight distribution observed in Western societies in recent years,
with intrauterine growth restriction occurring less frequently (Australian Institute of Health and
Welfare 2000; 2010), increasing attention has been directed towards larger size at birth and its
intrauterine determinants as potential risk factors for obesity and related outcomes. Therefore, this
project focuses on three modifiable factors known to alter intrauterine milieu, maternal pre-pregnancy
obesity, glucose intolerance during pregnancy across the entire spectrum of severity, and gestational
weight gain, and their relationship to the programming of obesity and insulin resistance in pre-
pubertal children.

2.2 Conceptualising pre-birth exposures

Intrauterine life is considered the most vulnerable period for future development, with programming
effects being observed as early as the pre-implantation embryo stage (Duranthon et al. 2008).
Plasticity in utero is much greater compared to any other stage of life, with the fertilised ovum
undergoing 42 cycles of cell division before birth, as opposed to only 5 after birth (Barker 1994).
This plasticity underpins the lifelong consequences that exposure to various factors during the fetal
period might have, including coronary heart disease, hypertension, cerebrovascular disease (stroke),
diabetes, obesity, altered body composition (including reduced bone mineral density), lung disease,
allergic disease, cancer, neuropsychiatric disorders, and infectious diseases (Kuh and Ben-Shlomo
2004).

Pregnancy is a physiologic state characterised by profound changes in multiple organ systems:


cardiovascular changes (expansion of plasma volume and red blood cells mass), renal changes
(greater glomerular filtration rate and increased tubular sodium reabsorption), endocrine processes
involving most hormonal axes (including the adipose-derived hormones that influence energy
metabolism, adiponectin (Catalano et al. 2006) and leptin (Kirwan et al. 2002)) and metabolic
changes (related to glucose, lipid and protein metabolism), which are all directed towards
maintaining the pregnancy and sustaining optimal fetal development (Hytten 1991). Among the
numerous factors operating during pregnancy that may affect this physiologic state, maternal
obesity, gestational glucose intolerance and excessive weight gain during pregnancy are particularly

11
Theoretical framework

important from a public health perspective, as they are increasingly common and, in principle,
modifiable.

This section presents separately the three intrauterine exposures of interest for the current project:
maternal pre-pregnancy obesity (Section 2.2.1), gestational glucose intolerance across the entire
spectrum (Section 2.2.2), and excessive gestational weight gain (Section 2.2.3). First, for each of
them, proposed underlying mechanisms for the metabolic perturbations they induce and their main
risk factors are briefly described. Next, the public health significance of each condition is highlighted,
with a focus on prevalence and incidence data worldwide and the detrimental consequences both on
the mother and her child. Last, clinical responses in relation to each of these maternal intrauterine
factors to limit the adverse health outcomes are described. Section 2.2.4 discusses the interrelations
identified by previous studies between maternal body size prior to pregnancy, gestational glucose
tolerance status and gestational weight gain.

2.2.1 Maternal pre-pregnancy obesity

Weight status is most commonly characterised based on body mass index (BMI), calculated as the
ratio of weight (in kilograms) to squared height (in metres). In adults BMI between 25 and 29.9
kg/m2 corresponds to overweight and BMI greater than 30 kg/m2 corresponds to obesity (World
Health Organization 1995).

2.2.1.1 Pathophysiology

It has been suggested that maternal obesity at the time of pregnancy, in combination with genetic
influences, profoundly affects intrauterine development both during embryonic (first 14 weeks of
gestation) and fetal periods, thus explaining the perturbations in both organ formation and fetal
growth often seen in pregnancies of obese women (King 2006). There is evidence that a fetus
responds to the nutrient transfer and oxygen supply from the mother through the placenta in two
ways: by immediate survival choices and by long-term integrated adjustments to maximise its
advantages in postnatal life (Gluckman et al. 2007).

The underlying mechanisms of the alterations in fetal growth induced by maternal obesity appear to
revolve primarily around increasing maternal insulin resistance from early to late pregnancy, which is
exacerbated in obese compared to lean women (Catalano et al. 1999). This increased insulin
resistance has three important metabolic consequences for the pregnant woman: a reduction in the

12
Theoretical framework

insulin-mediated glucose uptake into the cells, leading to hyperglycaemia; a decreased ability of
insulin to suppress whole-body lipolysis with secondary elevated levels of plasma free fatty acids;
and a decline in the suppression of amino acid turnover with an increase in plasma levels of amino
acids (King 2006). Therefore, fetuses of obese mothers are exposed to relatively high amounts of all
major fuel sources (glucose, lipids and amino acids), which in turn may independently contribute to
fetal overgrowth.

The distribution of adipose tissue stored during pregnancy was found to be more central (between
the costal and the upper thigh areas) in obese women compared to lean women (Ehrenberg et al.
2003), with implications for glucose-insulin metabolism, such as relative hyperinsulinemia and an
earlier (second trimester) maximal glucose response to an oral glucose load relative to women with
lower-body obesity (who display these changes only in the third trimester) (Landon et al. 1994).

Other pathologic processes linking maternal obesity with insulin resistance and extending to fetal
growth include adipocyte dysfunction, with subsequent lower circulating levels of adiponectin (a fat-
derived hormone with insulin-sensitising effects) (Retnakaran et al. 2004) and a subclinical chronic
inflammation with elevated levels of circulating cytokines, such as C-reactive protein (Retnakaran et
al. 2003).

In addition to the marked hyperinsulinemia that characterises the pregnancy of an obese woman,
gestational dyslipidemia is also exaggerated in obese women, who tend to display much higher
levels of triglycerides and free fatty acids, lower levels of high-density lipoprotein cholesterol, and a
significant increase in fat oxidation (Okereke et al. 2004). These changes in maternal lipid
metabolism have been shown to play a key role in the vascular complications of obesity during
pregnancy, such as preeclampsia (Ramsay et al. 2002), and to correlate with neonatal fat mass
(Schaefer-Graf et al. 2008).

Obesity effects on amino acid metabolism during pregnancy have been less documented. In non-
pregnant obese women, insulin-stimulated protein synthesis was reduced compared to average
weight women (Chevalier et al. 2005), due to the insulin resistance associated with obesity. Based
on these findings it was suggested that the anabolic response to pregnancy might be altered in
obese women, which could explain why some obese mothers have small-for-gestational age babies
(Nelson et al. 2010).

These obesity-related perturbations in the maternal nutritional environment have been associated
not only with immediate effects on fetal growth, but also with long-term outcomes in the developing

13
Theoretical framework

child, such as an increased risk of obesity. The long-tem metabolic consequences in pre-pubertal
children will be detailed in Section 2.4.1.

2.2.1.2 Public health significance – Prevalence data

Maternal obesity prior to conception is of great public health importance as it is now highly prevalent,
often associated with serious adverse health outcomes both for the mother and her child, and
involves significant costs to the health care system. International data on the prevalence of
overweight and obesity among women show an overall increasing trend both in developed and
developing countries (mainly in urban areas of developing countries) around the world (World Health
Organization 2009a). Table 1 exemplifies these rates for selected high income countries, relative to
WHO recommended BMI cut-off points, contrasting the figures from the late 1990s (when the cohort
on which this project is based was established) and the most recent data in the 2000s (World Health
Organization 2009a). Given the different sampling frames employed in different settings and age
groups considered, caution is needed for direct comparisons.

Table 1. National prevalence of overweight and obesity in women of all ages from selected developed
countries (World Health Organization 2009a)

Country Years of data Prevalence of overweight Prevalence of obesity


collection (BMI = 25-29.9 kg/m2) (BMI ≥30 kg/m2)
NOTE:
This table is included on page 14
of the print copy of the thesis held in
the University of Adelaide Library.

BMI - body mass index

Based on these data, approximately one third of Australian women can be classified as overweight,
and about a quarter as obese (World Health Organization 2009a). As risk factors for obesity
accumulate over the lifespan, the prevalence of obesity tends to increase with age; hence, rates
would be lower among women of childbearing age compared to those among older women.
Australian national prevalence of overweight and obesity (based on BMI calculated from measured
weight and height) among women of reproductive age over two time periods is presented in Table 2.

14
Theoretical framework

Overall, it appears that more contemporary women of childbearing age have a higher BMI compared
to the previous decade.

Table 2. National prevalence of overweight and obesity among women of reproductive age in
Australia (Australian Bureau of Statistics 2009b)

Age group Prevalence (%) of overweight Prevalence (%) of obesity


(years) (BMI = 25-29.9 kg/m2) (BMI ≥30 kg/m2)
1995 2007-2008 1995 2007-2008
18-24 17.7 20.7 8.6 14.2
25-34 23.1 26.4 15.0 18.0
35-44 29.4 32.4 16.1 22.7
BMI - body mass index

Obesity prevalence in pregnant women is less documented at a national level and is mainly
generated by observational studies in specific settings, but overall it seems to have risen in parallel
with the rates described in the general population of women of reproductive age. A large study
conducted in nine states in the USA (Pregnancy Risk Assessment Monitoring System) showed a
sharp increase in the rates of pre-pregnancy obesity (defined as BMI ≥ 29 kg/m2, calculated from
self-reported data) from 13% in 1993 to 22% ten years later (Kim et al. 2007). A more recent British
national report based on first trimester BMI data from over 600,000 pregnant women indicates that
the prevalence of maternal overweight increased slightly (from 22.3% in 1989 to 25.9% in 2007),
while the prevalence of obesity doubled (from 7.55% in 1989 to 15.6% in 2007) (Heslehurst et al.
2010). In Australia, no national data on overweight and obesity rates are available specifically in an
obstetric population. Current estimates of 20.2% for overweight and 13.5% for obesity were reported
based on pre-pregnancy BMI data from over 11,000 women with singleton pregnancies who
delivered in 1998-2002 at Mater Mother’s Hospital, Brisbane (covering 8% of all pregnancies in
Queensland, Australia), after excluding women with BMI < 20 kg/m2 (Callaway et al. 2006).

2.2.1.3 Adverse health outcomes of maternal pre-pregnancy obesity

Maternal pre-gravid obesity carries significant risks for the health of the woman and her child, which
extend beyond the pregnancy.

Adverse health outcomes for the mother

For the mother, obesity complications start prior to conception and can interfere with a woman’s
ability to conceive (Gesink Law et al. 2007). Following conception, overweight and obese women

15
Theoretical framework

have an increased risk of miscarriage (pregnancy loss before 20 weeks gestation), whether
conception was spontaneous or assisted (odds ratio OR=1.67, 95% CI 1.25, 2.25) (Metwally et al.
2008). One of the most commonly described pregnancy complications in overweight and obese
women is gestational diabetes (Guelinckx et al. 2008; Torloni et al. 2009), with OR ranging from 1.68
(95% CI 1.53, 1.84) (Sebire et al. 2001) to 2.71 (95% CI 1.32, 5.55) (Doherty et al. 2006) in
overweight women and from 2.95 (95% CI 2.05, 4.25) (Callaway et al. 2006) to 6.5 (95% CI 3.32,
12.74) (Doherty et al. 2006) in obese women, after confounder adjustments. Pregnancy of an obese
woman may also be complicated by hypertensive disorders (adjusted OR = 2.60 (95% CI 1.49, 4.55)
in overweight and 7.93 (95% CI 4.74, 13.27) in obese women (Doherty et al. 2006)), including
preeclampsia (Guelinckx et al. 2008). Maternal obesity is also known to increase the risk of venous
thromboembolic events both during pregnancy and postpartum (adjusted OR = 5.3 (95% CI 2.1,
13.5)) (Larsen et al. 2007). Other pregnancy complications of maternal obesity may include
increased risk of induced labour (adjusted OR = 2.44 (95% CI 1.72, 3.45) (Doherty et al. 2006)),
caesarean delivery (adjusted OR = 2.2 (95% CI 1.58, 3.12) (Doherty et al. 2006)) with anaesthetic
and operative complications (difficult intubation and placement of epidural, excessive bleeding and
postpartum infections) (Sebire et al. 2001). After delivery, obese women are more likely to
experience difficult initiation of breastfeeding (Oddy et al. 2006), haemorrhage, genital and urinary
tract infections, or wound infections (Sebire et al. 2001).

In the long term, obese mothers have an increased risk of developing hypertension, type 2 diabetes,
cancer (in particular endometrial cancer), osteoarthritis, non-alcoholic fatty liver disease, gallbladder
disease, pancreatitis, sleep apnoea, or depression (Pi-Sunyer 2009).

Adverse health outcomes for the child

For the fetus, maternal obesity increases the risk of congenital malformations (such as neural tube
defects, heart defects, cleft lip or palate, hydrocephaly, limb reduction), which are often missed
during the ultrasound examination and are major causes of stillbirth and infant mortality (Stothard et
al. 2009). Macrosomia and large-for-gestational age have been consistently reported as
consequences of maternal overweight (with adjusted OR ranging from 1.2 (95% CI 1.1, 1.4)
(Ehrenberg et al. 2004) to 1.57 (95% CI 1.50, 1.64) (Sebire et al. 2001)) or obesity (with adjusted OR
ranging from 1.6 (95% CI 1.4, 1.9) (Ehrenberg et al. 2004) to 2.36 (95% CI 2.23, 2.50) (Sebire et al.
2001)), independently of maternal gestational glucose tolerance (Jensen et al. 2003). Not only fetal
weight is affected by maternal obesity, but also body composition (greater percentage body fat) and
insulin sensitivity (higher insulin resistance) (Catalano et al. 2009b). Although findings from previous

16
Theoretical framework

studies are not always consistent with regard to the risk of preterm delivery associated with maternal
obesity (Callaway et al. 2006; Oddy et al. 2006), a systematic review and meta-analysis indicated an
increased risk of preterm birth in overweight and obese women (RR=1.24, 95% CI 1.13, 1.37)
(McDonald et al. 2010).

The far-reaching implications of maternal obesity on offspring consist of an increased risk to become
obese themselves (Blair et al. 2007; Boerschmann et al. 2010; Catalano et al. 2009a; Li et al. 2005;
Reilly et al. 2005; Whitaker 2004) or develop metabolic syndrome (Boney et al. 2005). A critical
appraisal of the published literature addressing the link between maternal pre-pregnancy overweight
or obesity and childhood obesity and insulin resistance is presented in Section 2.4.1.

All these health consequences of maternal obesity on pregnancy outcome also incur significant
economic burden to health care systems, due to more intensive hospital obstetric care required by
obese women and higher rates of admission to neonatal intensive care units among infants of obese
mothers (Chu et al. 2008; Galtier-Dereure et al. 2000).

2.2.1.4 Clinical response to the epidemic of maternal obesity

Extended discussion of possible responses to obesity in women of reproductive age will occur in
Chapter 5. For the moment it is worth noting that in order to prevent the above mentioned adverse
consequences and increase the chance of achieving healthy outcomes for mothers and their
children, it has been suggested that, ideally, obese women should receive counselling prior to
conception regarding the perinatal and long-tem risks associated with excessive weight, as well as
appropriate surveillance and support during pregnancy. In this view, the American Dietetic
Association and the American Society for Nutrition recommend counselling on the roles of diet and
physical activity in reproductive health in all overweight and obese women of childbearing age
(American Dietetic Association et al. 2009). It is believed that knowledge of these risks would
motivate women to adopt healthy eating and exercise regimen (Stotland and Caughey 2010), thus
preventing further excessive weight gain during pregnancy, which, in turn, has been associated with
adverse outcomes.

Information regarding effective intervention strategies during pregnancy in obese women to improve
maternal and child health outcomes is limited. It is well established though that pharmacological
treatment of obesity is unsafe during pregnancy and thus should not be recommended (Yogev and
Catalano 2009). Lifestyle intervention strategies for weight loss during pregnancy in obese women

17
Theoretical framework

are controversial, with some authors arguing for the benefits of a reduction in the risk of delivering
large-for-gestational age infants, while others emphasising the impact weight loss might have on
intrauterine growth restriction (Dodd and Robinson 2011). The current Institute of Medicine
recommendations for obese women are to avoid weight loss during pregnancy and to gain between
5 and 9.1 kg throughout pregnancy (Rasmussen and Yaktine 2009), much less than the
recommendations for women of average weight.

Physical exercise during pregnancy has been recognised as a beneficial factor that improves
maternal cardiovascular function (Clapp 2000) and insulin sensitivity (Clapp 2006b), with subsequent
reduction in plasma glucose levels, fetal exposure to lower levels of glucose and subsequent
decreased fetal insulin secretion, with potential influence on size at birth. The exercise regimen
plays an important role; while recreational exercise does not affect birth weight (Kramer and
McDonald 2006) vigorous endurance physical activity has been associated with a reduction in birth
weight (Clapp 2006a). However, putative risks to the fetus have been described during maternal
strenuous physical activity, such as decreased oxygen supply due to lower utero-placental blood flow
secondary to increased blood flow to maternal skeletal muscles (Kennelly et al. 2002) and decreased
glucose supply due to increased glucose use by maternal skeletal muscles (Bonen et al. 1992).

The effect of limiting the intake of carbohydrate and saturated fat during pregnancy in overweight or
obese women remains uncertain. Clapp (2002) found a reduction in gestational weight gain
associated with a low glycaemic index diet in obese pregnant women, but there is limited wider
evidence on the importance or otherwise of this property of maternal diet. In one other study, in
addition to the reduction in gestational weight gain, lower rates of prematurity and improved maternal
plasma lipids levels were found in overweight and obese pregnant women receiving a low-glycaemic
load diet during pregnancy, compared to those receiving a low-fat diet (Rhodes et al. 2010).

Several randomised controlled trials of lifestyle interventions (advice on physical activity and/or diet)
in overweight or obese pregnant women have been completed (Asbee et al. 2009; Thornton et al.
2009) and others are underway (e.g., LIMIT trial conducted in Adelaide, South Australia, “for limiting
weight gain in overweight and obese women during pregnancy to improve health outcomes” (Dodd
et al. 2011)). In the ongoing trials, in addition to the effect on limiting gestational weight gain, the
interventions will be examined with respect to the reduction in maternal insulin resistance and to the
rates of large-for-gestational age infants in obese women (Nelson et al. 2010).

18
Theoretical framework

2.2.2 Glucose tolerance status during pregnancy

Glucose metabolism undergoes major variations during pregnancy, as described in Section 2.2.2.1.
When these changes fail to be controlled by maternal compensatory mechanisms, gestational
glucose intolerance arises, with various degrees of severity, culminating in gestational diabetes.
Gestational diabetes (GD) has been defined as a clinical entity characterised by “carbohydrate
intolerance with variable severity and first recognition during pregnancy” (Metzger et al. 1998). An
amendment has been recently made to this definition, in that overt diabetes detected at the first
antenatal visit (before 24-28 weeks gestation) should not be considered gestational diabetes, but
rather pre-existing diabetes (International Association of Diabetes and Pregnancy Study Groups
Consensus Panel et al. 2010). Milder degrees of impaired glucose tolerance have also been
described during pregnancy, but their clinical significance is still controversial, as their adverse health
implications are widely unknown. Among them, borderline gestational glucose intolerance
(BGGI) has been defined as the mildest degree of glucose intolerance during pregnancy not meeting
the criteria for diagnosing gestational diabetes (Bonomo et al. 2005). Details on the screening for
GD are presented in Section 2.2.2.2. The following section presents the definitions of gestational
diabetes and borderline gestational glucose intolerance that are employed throughout this thesis.
Section 2.2.2.4 gives an overview of the factors considered to increase the risk of glucose
intolerance during pregnancy and constitutes the basis for identifying potential confounders that will
be included in the statistical analyses of this project. Section 2.2.2.5 describes the public health
significance of gestational glucose intolerance, Section 2.2.2.6 outlines the adverse health outcomes
of this pregnancy-related metabolic perturbation, while Section 2.2.2.7 presents, briefly, current
approaches in its management.

2.2.2.1 Glucose regulation during pregnancy

Glucose regulation in normal pregnancies

Any pregnancy is characterised by some degree of insulin resistance, commencing towards the end
of the second trimester and increasing progressively with advancing gestation to levels close to
those seen in patients with type 2 diabetes (Buchanan and Xiang 2005). This is, up to a certain
level, physiologic, representing an adaptive response to satisfy fetal nutritional demands for optimal
intrauterine growth. Insulin sensitivity is mainly impaired peripherally, in skeletal muscle and adipose
tissue (Catalano et al. 1993b). As a consequence of the impairment in insulin sensitivity, glucose
disposal in skeletal muscle and adipose tissue is reduced by about 40% in normal pregnancies,

19
Theoretical framework

which in turn stimulates maternal insulin secretion and suppresses hepatic glucose production in
order to maintain glucose levels within a relatively normal range (Buchanan and Xiang 2005;
Catalano et al. 1993b).

Glucose regulation in pregnancies complicated by glucose intolerance

Gestational diabetes occurs predominantly during the third trimester of the pregnancy, as a
consequence of increased insulin resistance and subsequent insufficient rise in insulin secretion to
overcome the insulin resistance (Catalano et al. 2003a).

In pregnancies complicated by glucose intolerance, in particular gestational diabetes, in addition to


the reduced insulin-stimulated glucose uptake by about 54% (Friedman et al. 1999), the first-phase
insulin response is insufficient to compensate for the peripheral insulin resistance and endogenous
glucose production is less suppressed in late gestation (Catalano et al. 1993b). In short, as a result
of the enhanced insulin resistance, maternal plasma glucose levels increase and glucose tolerance
fails to be maintained, with excessive transfer of maternal energy substrates, in particular glucose
and amino-acids through the placenta, which influence fetal development within the so-called ‘fuel
mediated teratogenesis’ (Freinkel 1980). Unlike glucose and other nutrients, maternal insulin does
not cross the placenta and, subsequently fetal plasma levels of insulin are dependent on secretion
from the fetal pancreas only, which starts from 8-10 weeks gestation (Beardsall et al. 2008). Given
that placental glucose transport depends on diffusion down a concentration gradient only, maternal
hyperglycaemia is followed by fetal hyperglycaemia. This, in turn, stimulates the fetal pancreas to
secrete more insulin compared to fetuses of normoglycaemic mothers. Insulin acts as a growth-
promoting factor (Beardsall et al. 2008). Thus, fetal hyperinsulinemia secondary to maternal
hyperglycaemia is followed by an accelerated rate of growth in utero and subsequently greater size
at birth (Vohr et al. 1995; Vohr and McGarvey 1997). The accelerated fetal growth, as a
consequence of fetal hyperinsulinism, mainly accounts for an increase in adipose tissue and
percentage body fat at birth, and not lean mass (Catalano et al. 1995), with potential metabolic
consequences later in life. Greater adiposity, likely due to fetal hyperinsulinism, has also been
observed in appropriate-for-gestational age infants of mothers with GD compared to those of women
with normal glucose tolerance during pregnancy (Catalano et al. 2003b).

Mechanisms of insulin resistance during pregnancy

A number of underlying mechanisms have been explored to explain the enhanced insulin resistance
during pregnancies complicated by GD, albeit that they are still largely unknown. It has been

20
Theoretical framework

suggested that insulin resistance developed during pregnancy would be directly associated with the
amount of maternal adiposity gained in early pregnancy, the insulin-desensitising effects of placental
derived hormones (such as human placental lactogen and human placental growth hormone)
(Barbour et al. 2007; Buchanan and Xiang 2005), and a series of adipokines (adiponectin, tumour
necrosis factor (TNF)-alpha) (Barbour et al. 2007). More specifically, it has been found that this
insulin resistance is linked to defects at post-receptor level (in the intracellular insulin signalling
pathway (Catalano 2010)) in insulin-sensitive tissues, such as adipose tissue, skeletal muscle, and
placenta (Colomiere et al. 2010).

It has been hypothesised that a certain degree of insulin resistance exists even prior to conception in
women who develop GD (Buchanan 2001). Supporting this hypothesis is the finding of higher levels
of C-peptide (a marker of glucose-triggered insulin release) in women with GD, compared to women
who maintain normal glucose tolerance during pregnancy (in the latter category of women, C-peptide
level increases between 20 and 32 weeks of gestation) (Weijers et al. 2002). In contrast,
pregnancies complicated by milder degrees of glucose intolerance display C-peptide levels similar to
normal pregnancies (Weijers et al. 2002), suggesting the absence of antedating insulin resistance.

2.2.2.2 Screening and diagnostic tests for glucose intolerance during pregnancy

As GD is generally asymptomatic (McIntyre et al. 2005), it can be detected only based on clinical
tests. Screening tests have been sought given the potential for adverse health outcomes of GD,
initially considering women’s risk of later developing type 2 diabetes, and more recently,
incorporating the harm to the offspring, in particular macrosomia (as detailed in Section 2.2.2.6).
The utility of universal screening for GD, which would maximise the diagnosis and give opportunity
for intervention to improve pregnancy outcomes, is still under debate (Metzger et al. 2007). The
National Institute for Health and Clinical Excellence in the UK and the American Diabetes
Association (ADA) recommend selective screening, on the basis of risk factor assessment, such as
pre-gravid obesity, advanced age, high-risk ethnicity and family history of diabetes (American
Diabetes Association 2002; 2006; National Institute for Health and Clinical Excellence 2008). In
contrast, the Australasian Diabetes in Pregnancy Society (ADIPS) has recommended universal
screening since 1998 (Government of South Australia 2007). Screening is usually recommended to
be performed at 24-28 weeks of pregnancy (American Diabetes Association 2006; Hoffman et al.
1998). However, pregnant women at high risk of GD are recommended to undergo the screening
test earlier in pregnancy (Hoffman et al. 1998).

21
Theoretical framework

Oral glucose challenge test

The most validated screening method for GD consists of a non-fasting oral glucose challenge test
(OGCT), with assessment of plasma glucose levels 1 hour after a 50 g oral glucose load (American
Diabetes Association 2002). The sensitivity of OGCT has been estimated as 79% and specificity as
86% (O'Sullivan and Mahan 1964). The optimum cut-offs for a positive challenge test are still
uncertain. Most commonly, women with a positive challenge test result (plasma glucose level
exceeding the threshold of 7.8 mmol/l) are either diagnosed with GD if the glucose level is higher
than 11.1 mmol/l, or tested further by an oral glucose tolerance test if the plasma glucose level is
between 7.8 and 11.1 mmol/l (World Health Organization 1985). Plasma glucose levels lower than
7.8 mmol/l are considered normal (World Health Organization 1985).

Oral glucose tolerance test

The oral glucose tolerance test (OGTT) is the diagnostic test for GD. It needs to be performed after
an 8-14 hour fasting period and minimum three days of unrestricted diet and unlimited physical
activity (American Diabetes Association 2002). These requirements are one of the reasons a
screening test is usually applied first, to rule out the majority of women who do not need to be
burdened in this manner. There are two protocols for the OGTT: one based on 100 g glucose load
and measurement of plasma glucose level every hour for 3 hours after oral loading (initially proposed
by O’Sullivan and Mahan (1964) and modified by Carpenter and Coustan (1982)), and the second,
based on 75 g glucose load and plasma glucose level measurements at baseline and after 2 hours
(World Health Organization 1985).

A number of sets of criteria for the diagnosis of GD have been proposed, with some variation in
glucose threshold values (Table 3). There is no universal consensus in the use of a particular set of
diagnostic criteria and hence the choice of particular glucose thresholds for the diagnosis of
gestational diabetes is according to the local institution’s policy. Moreover, it is important to note that
categories of glucose tolerance during pregnancy have been proposed to facilitate interpretation, but
in both OGCT and OGTT, plasma glucose level is a continuous variable. This aspect is particularly
important given the linear relationship found between hyperglycaemia during pregnancy, as a
continuum, rather than a threshold effect, and the risk of adverse outcomes (HAPO Study
Cooperative Research Group 2008). As an alternative to the sequential screening/diagnosis
strategy described above, performing the diagnostic OGTT directly without preceding it by an OGCT
has been recommended as a cost-effective approach in high risk populations (American Diabetes

22
Theoretical framework

Association 2002). However, in the general population, the two-step approach (OGCT followed by
OGTT) was more cost-effective than OGTT alone or no-screening at all (Nicholson et al. 2005).

Table 3. Cut-off values for plasma glucose levels (mmol/l) recommended for the diagnosis of
gestational diabetes in oral glucose tolerance test (OGTT)

NDDG ADA ACOG WHO ADIPS NZSSD


Carpenter and
Coustan
100 g glucose 100 g glucose 100 g glucose 75 g glucose 75 g glucose 75 g glucose
load load load load load load
Fasting glucose ≥ 5.8 ≥ 5.3 ≥ 5.8 ≥ 7.8 ≥ 5.5 ≥ 5.5
1-h post-load ≥10.5 ≥10.0 ≥10.6
2-h post-load ≥ 9.2 ≥ 8.6 ≥ 9.2 ≥11.1 ≥ 8.0 ≥ 9.0
3-h post-load ≥ 8.0 ≥ 7.8 ≥ 8.1
NDDG - National Diabetes Data Group; ADA - American Diabetes Association; ACOG - American College of Obstetricians and
Gynecologists; WHO - World Health Organization; ADIPS - Australasian Diabetes in Pregnancy Society; NZSSD - New Zealand
Society for the Study of Diabetes

Other screening methods for gestational diabetes

Other screening methods for GD have been proposed, but they are controversial and hence less
often used in practice. They include measurement of fasting plasma glucose level (HAPO Study
Cooperative Research Group 2008), measurement of random plasma glucose level (Government of
South Australia 2007), or measurement of glycosylated haemoglobin A1c (HbA1c). The latter method
provides an estimate of the mean plasma glucose levels in the three months preceding the test
(Gabbe 2003; Lapolla et al. 2007) and, if measured early in pregnancy, is particularly useful to detect
pre-gestational type 2 diabetes.

2.2.2.3 Defining categories of gestational glucose intolerance based on oral glucose


challenge test and oral glucose tolerance test

Stratification of pregnant women into glucose tolerance categories differs according to the diagnostic
criteria used. Based on the American Diabetes Association criteria, the diagnosis of GD requires at
least two out of four elevated plasma glucose levels above the thresholds for the OGTT over the 3
hours (American Diabetes Association 2006). If only one threshold value at OGTT is exceeded, the
woman is considered as having impaired glucose intolerance (IGT), while a positive OGCT followed
by a normal OGTT (no increased plasma glucose levels during OGTT) is regarded as borderline
gestational glucose intolerance (Bonomo et al. 2005).

23
Theoretical framework

In Australia, based on ADIPS criteria, women are diagnosed with GD if at least one of the OGTT
threshold values is exceeded (i.e., either fasting plasma glucose level is ≥ 5.5 mmol/l and/or 2-h
post-glucose load is ≥ 8.0 mmol/l) (Hoffman et al. 1998). In this context, borderline gestational
glucose intolerance (BGGI), can be defined based on a positive OGCT and a normal OGTT (Ju et al.
2008) (i.e., both fasting plasma glucose level is < 5.5 mmol/l and 2-h post-glucose load is < 8.0
mmol/l). For this project, ADIPS criteria are employed and this definition of BGGI is used throughout
the thesis.

2.2.2.4 Risk factors for glucose intolerance during pregnancy

Risk factors for gestational diabetes

Gestational diabetes is considered a multifactorial disease. The risk factors for GD, often
interrelated, are outlined below, grouped from a life course perspective.

Maternal and familial factors:

 Family history of type 2 diabetes in first-degree relatives (Kim et al. 2009; McLean et al. 2006;
Solomon et al. 1997);

 Being a member of an ethnic group with a high prevalence of diabetes, such as Australian
Aboriginal, Pacific Islander, Hispanic, Native American, African American, South and East
Asian (Hoffman et al. 1998; Solomon et al. 1997);

 Maternal own birth weight has an unclear relationship to the risk of developing GD. Some
studies indicate an inverse relationship between own birth weight and the risk of glucose
intolerance during pregnancy (Bo et al. 2003; Seghieri et al. 2002), while other studies support
a U-shaped or J-shaped relationship, with both high (> 4000 g) and low birth weight (< 2000 g)
increasing the subsequent risk of GD (Innes et al. 2002);

 Maternal short stature (Bo et al. 2001; Branchtein et al. 2000; Dode and dos Santos 2009;
Jang et al. 1998; Ogonowski and Miazgowski 2010);

 Polycystic ovary syndrome (Hanna and Peters 2002).

Risk factors related to previous pregnancy history:

 Gestational diabetes in a previous pregnancy (Bottalico 2007; Dudhbhai et al. 2006; Radesky
et al. 2008; Tieu et al. 2008), most likely due to the persistence of other risk factors;

24
Theoretical framework

 Previous history of macrosomia (Bottalico 2007; Corrado et al. 2006);

 Previous unexplained stillbirth (Robson et al. 2001);

 Previous history of complicated pregnancy: preeclampsia, hypertension, polyhydramnios


(Bottalico 2007; Corrado et al. 2006).

Risk factors related to current pregnancy:

 Advanced maternal age (Bottalico 2007; Corrado et al. 2006; Dudhbhai et al. 2006; Montan
2007; Solomon et al. 1997) as a continuum rather than a threshold effect;

 Parity (Dudhbhai et al. 2006; Seghieri et al. 2005);

 Pre-pregnancy obesity (Torloni et al. 2009); in addition to pre-pregnancy obesity, a greater


weight gain in the 5 years prior to pregnancy further increases the risk of GD (Hedderson et al.
2008);

 Gestational weight gain greater than 1.2 kg in the first trimester, or greater than 0.4 kg/week in
the second and third trimesters (Bonomo et al. 2005);

 Recurrent glycosuria (i.e., the presence of glucose in the urine) during pregnancy (American
Diabetes Association 2006; Bottalico 2007).

Other risk factors:

 Diet: high glycaemic index diets (Clapp 2002), increased intake of saturated fat (Bo et al.
2001); higher consumption of red and processed meat (Zhang et al. 2006a);

 Physical inactivity: regular exercise before pregnancy lowered the risk of GD (Zhang et al.
2006b), or even of mild gestational glucose intolerance, especially if they continued with light-
moderate physical activity in early pregnancy (Oken et al. 2006);

 Cigarette smoking (Wendland et al. 2008);

 Socioeconomic status (lower SES) (Anna et al. 2008; Australian Bureau of Statistics 2001; Bo
et al. 2002; Clausen et al. 2006).

According to ADIPS, the highest risk of developing GD is associated with maternal age over 30
years, obesity, family history of diabetes, GD in a previous pregnancy, certain ethnicity, and past
history of complicated pregnancy (Hoffman et al. 1998).

25
Theoretical framework

From a public health perspective, it is important to consider whether the risk factors for GD are
modifiable or not. Arguably, some of the risk factors may be modifiable only at population level and
not at individual level (e.g., women’s BMI at the time of pregnancy cannot be modified once they
present at the prenatal visits, although BMI is potentially modifiable at population level).

Risk factors for borderline gestational glucose intolerance

There are fewer studies investigating risk factors for milder degrees of glucose intolerance during
pregnancy, such as BGGI, but the risk factors seem to be similar to those for GD. These risk factors
include: pre-pregnancy obesity (Weijers et al. 2002), advanced maternal age (Ju et al. 2008), parity
(Dudhbhai et al. 2006), maternal low birth weight (Bo et al. 2003), increased intake of fat (Saldana et
al. 2004), and sedentary behaviours prior to and during pregnancy (Oken et al. 2006). Higher
gestational weight gain up to the time of the screening for GD did not increase the risk of GD or
BGGI (defined by a positive screening test and a negative 3 h OGTT), but it doubled the risk of
impaired glucose tolerance (defined as a positive screening test and one high plasma glucose level
at 3-h OGTT) (Herring et al. 2009). Smoking did not seem to be associated with the risk for mild
gestational glucose intolerance (Ju et al. 2008).

2.2.2.5 Public health significance – Prevalence data

Prevalence of gestational diabetes

Gestational diabetes is affecting increasing numbers of pregnant women worldwide (Ferrara 2007;
Hunt and Schuller 2007). Some reports express these rates in terms of prevalence (defined as the
total number of all individuals affected by the condition in a given population at a designated time),
while others as incidence (defined as the number of new individuals affected by the disease in a
specified population within a defined period of time). Given that GD is a transient condition, with a
relatively short duration (3-4 months), new cases do not add much to the pool of condition in the
population, as other cases resolve at the same time. Therefore, the incidence of GD does not differ
much from its prevalence in a population in a specified period of time.

The prevalence of GD depends mainly on the population racial and ethnic characteristics, the clinical
practice for detection of diabetes before and during pregnancy, and the diagnostic criteria used
(basically, using lower plasma glucose thresholds leads to a higher prevalence by including a larger
number of women into the diabetic group and vice versa). At the same time, the distribution of
various risk factors in the population, such as family history of diabetes, influences this prevalence.

26
Theoretical framework

Thus, the prevalence of GD varies in different studies according to different criteria of diagnosis and
country, ranging from 2.2% in Korea (Cheung and Byth 2003), to 4% in the USA (American Diabetes
Association 2006), and to 8.8% in Australia (Beischer et al. 1996). In Western countries, the most
likely reasons for this elevated prevalence are the higher rate of pre-pregnancy obesity, the
increasing maternal age at conception (Australian Bureau of Statistics 2006) and the more
widespread implementation of screening strategies for GD.

Studies conducted in Australia at different points in time, some hospital-based others population-
based, both retrospective and prospective, most of them using ADIPS criteria, showed different rates
of GD (Hunt and Schuller 2007). Their main findings are summarised in Table 4.

Table 4. Prevalence and incidence rates of gestational diabetes reported in Australia

Reference Study design, Setting Ethnicity (n) Prevalence (P) or


sample frame, incidence (I) of
time frame gestational diabetes

Beischer Hospital-based, Melbourne Not specified


et al. 1996 retrospective, (27,111 in 1971-1980) I: 2.9% in 1971-1980
1971-1994 (16,820 in 1991-1994) I: 8.8% in 1991-1994
Davey and Hospital-based, Melbourne Not specified I: 5.2%
Hamblin retrospective, (6,032)
2001 1996-1998
Ishak and Population-based, South Non-Aboriginal (225,000) P: 2.4% non-Aboriginal
Petocz retrospective, Australia Aboriginal (4,800) 4.3% Aboriginal
2003 1988-1999
Kim and Hospital-based, Cairns Aboriginal, Torres Islanders, I: 14.4% in 1992
Humphrey 1992-1996 Australian-born Caucasian, 13.4% in 1993
1999 others 11.1% in 1994
(7,576) 7.3% in 1995
5.3% in 1996
Moses et Population-based, New South Australasian, Aboriginal and I: 7.2% overall
al. 1994 retrospective, Wales, Pacific Islander, Asian, 7.1% Australasian
1993 Illawarra Northern and Southern 11.9% Asian
area European, others
(2,152)
Stone et Population-based, Victoria Non-Aboriginal (59,962) I: 3.6% overall
al. 2002 retrospective, data Aboriginal (438) 3.6% non-Aboriginal
linkage, 1996 4.3% Aboriginal
Yue et al. Hospital-based, Sydney Anglo-Celtic, Aboriginal, Asian P: 6.7% overall
1996 prospective, 1996 (5,250) 3.0% Anglo-Celtic
7.3% Arab
9% Vietnamese
10 % Aboriginal
15% Chinese
17% Indian

27
Theoretical framework

A retrospective population analysis of all deliveries in South Australia between 1988 and 1999
showed 2.5 fold higher rates of age-standardised GD in Aboriginal compared to non-Aboriginal
women (Ishak and Petocz 2003). However, the trends in GD prevalence over the 12 year period
studied were different, with a 72% increase among non-Aboriginal women and only 12% increase
among Aboriginal women, potentially attributable to a lower accessibility of health care services in
the latter group of women.

According to the most recent national data based on counts of diagnosed cases, the incidence of GD
among Australian women aged 15-49 years has increased by 20% in the last 5 years, reaching 4.6%
in 2005-2006 (Australian Institute of Health and Welfare 2008). The reported rate of GD in different
states of Australia ranged between 2.8% in Tasmania to 5.2% in Queensland, while in South
Australia it was 4.9% (Australian Institute of Health and Welfare 2009).

Advanced age places pregnant women at a higher risk of GD. Given the current trend in delayed
childbearing, this adds to the burden of this condition. The latest Australian report on the incidence
of GD indicated that 13% of women aged 44-49 years were affected, as opposed to only 1% among
pregnant women of age under 20 years (Australian Institute of Health and Welfare 2008). Based on
the same report, of all Australian women whose pregnancies were complicated by GD in 2005-2006,
the highest proportion was in the age group of 30-34 years (Australian Institute of Health and
Welfare 2008).

Ethnicity is another factor that influences the incidence of GD. Aboriginal and Torres Strait Islander
women have a 1.5 fold increased risk of developing GD compared to non-Indigenous Australian
women (Australian Institute of Health and Welfare 2008). Another interesting finding of this national
report refers to the double incidence rate of GD among women born overseas, especially in Southern
Asia, compared to Australian born women (Australian Institute of Health and Welfare 2008).

Prevalence of borderline gestational glucose intolerance

Although milder degrees of gestational glucose intolerance, defined by an abnormal OGCT and a
normal OGTT, are likely to be more prevalent than GD, fewer reports have addressed their public
health impact. Similar to GD, the use of different criteria to define mild impairment of glucose
tolerance during pregnancy provides different prevalence rates of this condition. In Australia, based
on ADIPS criteria, borderline gestational glucose intolerance can be defined as a positive OGCT
(plasma glucose level 1 hour post-50 g glucose load ≥ 7.8 mmol/l) but a negative OGTT (fasting
plasma glucose level < 5.5 mmol/l and 2-h plasma glucose level < 8.0 mmol/l post-75 g glucose

28
Theoretical framework

load). A recent analysis of OGCT and OGTT results available from 1,804 pregnant women who
participated in the Australian Collaborative Trial of Supplements (for which data collection took place
in 2002-2004) indicates a prevalence of BGGI of 8%, double the prevalence of GD (Ju et al. 2008).

2.2.2.6 Adverse health outcomes of gestational glucose intolerance

Adverse health outcomes of GD

A wide range of perinatal and long-term postpartum adverse health outcomes have been described
in pregnancies complicated by gestational diabetes, affecting both the mother and the child.

Adverse health outcomes for the mother

Women with GD have an increased risk of hypertensive disorders (RR=1.54, 95% CI 1.28, 2.11)
(Joffe et al. 1998), including preeclampsia (adjusted OR=1.6, 95% CI 1.4, 1.9) (Stone et al. 2002), a
higher need for induction of labour (adjusted OR=3.0, 95% CI 2.7, 3.4) (Stone et al. 2002) and
caesarean section (adjusted OR=1.7, 95% CI 1.6, 1.9) (Stone et al. 2002) compared to women with
normal glucose tolerance during pregnancy. These adverse outcomes for the mother also have flow
on effects for the infant.

Although GD is generally a transient condition and reverts to normal glucose tolerance after delivery,
in 30-70% of the women it is followed by type 2 diabetes in the 5 years following the pregnancy (Kim
et al. 2002; McLean et al. 2006), possibly due to a residual dysfunction of the pancreatic beta-cells
(Buchanan 2001). The risk of developing overt type 2 diabetes is exacerbated if the woman is obese
during pregnancy (Guelinckx et al. 2008). It has been suggested that GD represents early
unmasked type 2 diabetes (Yue et al. 1996). Population attributable risk (PAR) of gestational
diabetes for type 2 diabetes (i.e., the proportion of cases of type 2 diabetes in women that could be
related to the previous development of GD) varies in different populations and with different
diagnostic criteria. Based on four longitudinal studies conducted in Australia (Beischer et al. 1996;
Davey and Hamblin 2001; Martin et al. 1995; Moses et al. 1994), PAR was estimated to 0.21-0.31
(Cheung and Byth 2003). As a consequence, GD along with overweight and obesity have been
included in the key areas for monitoring diabetes in Australia, with a potential for primary prevention
of type 2 diabetes (Australian Institute of Health and Welfare 2006).

29
Theoretical framework

Adverse health outcomes for the child

As GD develops later during pregnancy, it does not affect the formation of fetal organs and does not
cause birth defects as seen in children of mothers with diabetes antedating pregnancy. However, it
occurs at a time when the fetus is growing intensively and therefore it can affect body size, body
composition and glucose-insulin metabolism. Perinatal risks for the newborn of a mother with GD
are well recognized and include macrosomia (Van Assche et al. 2001) (adjusted OR=2.0, 95% CI
1.8, 2.3) (Stone et al. 2002), secondary birth trauma (such as shoulder dystocia, bone fractures, and
nerve palsies), respiratory distress syndrome (adjusted OR=1.6, 95% CI 1.2, 2.2) (Stone et al. 2002),
hypoglycaemia (Langer et al. 2005) and neonatal jaundice (adjusted OR=1.4, 95% CI 1.2, 1.7)
(Stone et al. 2002). Not only size at birth, but body composition is influenced by the exposure to an
intrauterine diabetic environment. Increased body adiposity was described by skinfold measures
and total body electrical conductivity in newborns of mothers with GD, even if they had normal birth
weight (Catalano et al. 2003b).

The long-term detrimental consequences of exposure to GD for the child include an increased risk of
obesity (Boerschmann et al. 2010; Catalano et al. 2009a; Chandler-Laney et al. 2011; Egeland and
Meltzer 2010; Krishnaveni et al. 2010; Pettitt et al. 1983; Pettitt et al. 1985; Silverman et al. 1991),
insulin resistance (Boerschmann et al. 2010; Dabelea et al. 2000; Egeland and Meltzer 2010; Keely
et al. 2008; Krishnaveni et al. 2010), impaired glucose tolerance (Malcolm et al. 2006; Silverman et
al. 1995; Silverman et al. 1998), higher systolic blood pressure (Cho et al. 2000; Wright et al. 2009),
and potentially some degree of impaired intellectual development (Rizzo et al. 1997; Silverman et al.
1998). A critical appraisal of studies addressing the long-term metabolic implications of GD in the
offspring is presented in Section 2.4.2.

Adverse health outcomes of BGGI

Evidence that various degrees of glucose intolerance during pregnancy, less severe than gestational
diabetes, may be associated with adverse consequences is mounting. To date, the most compelling
evidence on the perinatal outcomes of elevated plasma glucose levels has been provided by the
multinational study on Hyperglycemia and Adverse Pregnancy Outcome (HAPO), conducted in
approximately 25,000 pregnant women from the USA, Canada, Europe, Asia and Australia, enrolled
during 2000-2006, all receiving a 75-g OGTT between 24-32 weeks gestation (women with overt GD,
i.e., fasting plasma glucose > 5.8 mmol/l or 2-hour plasma glucose level > 11.1 mmol/l, were
excluded) (HAPO Study Cooperative Research Group 2002). This large study found a positive,

30
Theoretical framework

continuous association between maternal plasma glucose (each of the fasting, 1-hour and 2-hour
levels during OGTT) in pregnancy across the spectrum and the risk of perinatal outcomes, such as
macrosomia, higher levels of umbilical cord C-peptide (fragment of insulin), neonatal hypoglycaemia
and caesarean section, with no evident cut-off value for plasma glucose levels at which these risks
increased (HAPO Study Cooperative Research Group 2008).

Adverse health outcomes for the mother

There is some evidence that mild glucose intolerance during pregnancy contributes to the risk of
preeclampsia (Dodd et al. 2007; Yogev et al. 2004), preterm birth (Yang et al. 2002) and need for
caesarean section (HAPO Study Cooperative Research Group 2008). In the longer term, it has been
shown that not only the risk of developing type 2 diabetes may be increased in women with BGGI,
but also the risk of cardiovascular disease, such as acute myocardial infarction or stroke; however,
this risk is lower than in women with pregnancies complicated by GD (Retnakaran and Shah 2009).
Moreover, it has been shown that among women with mild gestational glucose intolerance, those
with isolated hyperglycaemia at 1 hour during OGTT have the highest degree of postpartum
metabolic dysregulation, similar to those with GD (Retnakaran et al. 2008).

Adverse health outcomes for the child

Perinatal deleterious effects of intrauterine exposure to milder degrees of gestational glucose


intolerance include macrosomia (Dodd et al. 2007; HAPO Study Cooperative Research Group 2008;
Vambergue et al. 2000; Yang et al. 2002), birth injuries and neonatal hypoglycaemia (Dodd et al.
2007; HAPO Study Cooperative Research Group 2008). While acknowledging the potential role of
mild hyperglycaemia in pregnancy in excessive fetal growth, some reports suggest that the effect
might be attributed, at least partly, to maternal obesity at the time of pregnancy (Ricart et al. 2005).
Furthermore, it has been shown that maternal elevated level of triglycerides in the second trimester
of gestation independently predicted macrosomia for women with impaired glucose tolerance during
pregnancy (Bo et al. 2004).

Increased maternal plasma glucose levels in pregnancy, even if they are not high enough for a
diagnosis of GD, have been positively associated with offspring size during infancy (at 6-8, 24-36
and 96-120 weeks), and negatively associated with child’s growth trajectory by the age of 2 years,
known as a ‘catch-down’ or decelerated growth (Stenhouse et al. 2006). It has been suggested that
the adverse outcomes of hyperglycemia during pregnancy on the offspring extend beyond infancy as
a continuum across the range of maternal glucose tolerance (Beardsall et al. 2008), but the evidence

31
Theoretical framework

is limited. The few studies conducted in non-high risk populations have suggested that maternal
hyperglycaemia in pregnancy, across the spectrum, is related to the risk of obesity in the child
(Chandler-Laney et al. 2011; Hillier et al. 2007). A critical appraisal of studies addressing the long-
term metabolic implications of milder degrees of gestational glucose intolerance in the offspring is
presented in Section 2.4.2.

2.2.2.7 Management of gestational glucose intolerance

Management of GD

In order to assess the efficacy of preventing the adverse health implications posed by GD, several
randomised trials have focussed on the effects of tight glycaemic control compared to routine
antenatal care. The Australasian Carbohydrate Intolerance Study in Pregnancy (ACHOIS), a large-
scale randomized-treatment trial conducted in Adelaide, Australia, demonstrated that GD
management consisting of dietary advice, blood glucose monitoring, and insulin when required,
improved perinatal outcomes in both mother and child (Crowther et al. 2005). Briefly, infants in the
intervention group had reduced rates of serious perinatal outcomes (death, shoulder dystocia,
fracture and/or nerve palsy), lower mean birth weight, and lower rates of macrosomia, while their
mothers reported improved health-related quality of life, were more likely to need induction of labour,
with no differences in the rates of caesarean section, had lower weight gain during pregnancy
(between first booking and last antenatal visit) and lower rates of preeclampsia (Crowther et al.
2005). Based on ACHOIS findings, the current management of GD in Australia consists of dietary
advice (Tieu et al. 2008), blood glucose monitoring (with an aim for fasting plasma glucose of 3.5-5.5
mmol/l and 2 hour postprandial plasma glucose of 4.0-7.0 mmol/l), and insulin when required (i.e., if
fasting plasma glucose is ≥ 5.5 mmol/l at least once a week or postprandial plasma glucose is ≥ 7.5
mmol/l at least twice a week) (Government of South Australia 2007).

A more recent randomised clinical trial of similar scale conducted in the US, in which women with
mild GD (defined as two or three plasma glucose levels above the thresholds, but fasting glucose <
5.3 mmol/l at OGTT) were randomly assigned to treatment or routine care, showed no difference in
the rates of a composite outcome including stillbirth, perinatal death, neonatal hyperbilirubinemia,
hyperinsulinemia, and birth trauma, but found clear benefits of treatment in reducing mean birth
weight, neonatal fat mass, rates of large-for-gestational-age or macrosomia, shoulder dystocia,
caesarean delivery and preeclampsia (Landon et al. 2009).

32
Theoretical framework

In recent years, oral glucose lowering drugs (such as Glibenclamide or Metformin) have been
proposed as an alternative to insulin during pregnancies complicated by GD. Several randomised
controlled trials and observational studies are currently underway to evaluate the perinatal effects of
such treatments on maternal and neonatal outcomes (Alwan et al. 2009). It is, at this stage,
uncertain what beneficial or adverse effects they may have in the long-term for both mothers and
children.

Further research is being undertaken to examine whether management of GD has long-term benefits
for the mother and the child. Based on a US data-linkage study, Hillier et al. (2007) showed that
among 5-7 year old children exposed to GD, the risk of overweight (OR=1.89, 95% CI 1.30, 2.76)
and obesity (OR=1.82, 95% CI 1.15, 2.88) was attenuated if the mother received dietary advice and
insulin if required (ORweight>85th =1.29, 95% CI 0.85, 1.97; ORweight>95th=1.38, 95% CI 0.84, 2.27). In
contrast, data from the 4-5 year follow-up of the children whose mothers took part in the ACHOIS
trial did not support a benefit of treatment for GD on weight status in children (Gillman et al. 2010).

Management of BGGI

Intervention strategies designed to decrease the occurrence of adverse perinatal complications in


women with BGGI have also started to be evaluated, but so far evidence has been insufficient to
justify routine treatment. A randomised clinical trial in Italy assessed the effect of treating BGGI by
offering dietary advice and blood glucose monitoring (Bonomo et al. 2005). Bonomo et al. (2005)
reported a lower incidence of large-for-gestational age infants and lower neonatal ponderal index
(weight divided by length3 x 100) in the treatment group compared to non-treatment group or
controls. Long-term benefits of treatment for BGGI are completely unknown.

2.2.3 Gestational weight gain

Gestational weight gain (GWG) has been regarded as an indicator of maternal and fetal wellbeing
(Institute of Medicine 1990). It has been suggested that excessive GWG, similar to maternal obesity
prior to pregnancy, may induce metabolic processes that influence the intrauterine environment, with
subsequent effects on fetal development.

Initial Institute of Medicine (IOM) recommendations for GWG (Institute of Medicine 1990) were
directed to correct insufficient GWG and thus reduce prevalence of low birth weight infants.
Subsequently, given the increasing trend of obesity prevalence, these recommendations have been

33
Theoretical framework

revised (Rasmussen and Yaktine 2009) and the focus has been extended to also include excessive
GWG. Current IOM recommendations for GWG are stratified by maternal weight status at the time
of pregnancy (Rasmussen and Yaktine 2009), as outlined in Table 5.

Table 5. 2009 Institute of Medicine (IOM) recommendations for gestational weight gain (GWG) for
singleton pregnancies (Rasmussen and Yaktine 2009)

IOM recommendations 2009 for GWG


Pre-pregnancy BMI category

NOTE:
This table is included on page 34
of the print copy of the thesis held in
the University of Adelaide Library.

2.2.3.1 Pathophysiology

Weight gain represents a physiological process during pregnancy, aimed to facilitate transfer of
nutrients across the placenta and thus maintain normal fetal development (King 2006). It comprises
products of conception (including the fetus, placenta and the amniotic fluid), changes in maternal
tissues (uterus, breasts, blood, and, to a smaller extent, liver and intestinal mucosa), and changes in
body composition (with a significant increase in total body water, subcutaneous fat and protein
reserves) (Hytten 1991).

Compared to the water or protein components, the amount of fat stored during pregnancy is more
strongly correlated with total GWG and is thought to contribute the most to the development of
offspring obesity later in life (Committee on the Impact of Pregnancy Weight on Maternal and Child
Health 2007). The majority of fat deposited during pregnancy is subcutaneous; this fat store is built
progressively up to about 30 weeks of gestation, mainly in the abdominal region, back and upper
thighs, representing about one third of GWG (Hytten 1991).

The rate of weight gain is not constant throughout pregnancy and therefore the global pattern over
time has a slight sigmoid shape, rather than linear (Hytten 1991). The weekly rates are lower before
16-18 weeks of gestation (0.36 kg/wk), higher between 16-32 weeks (0.45 kg/wk), followed by slower
rates towards the term (0.36 kg/wk) (Hytten 1991). The maximum weekly weight gain was found at

34
Theoretical framework

20-24 weeks (Dawes and Grudzinskas 1991). GWG is influenced by maternal pre-pregnancy weight
status, age, parity (Hytten 1991), and less so by smoking, blood pressure (Dawes and Grudzinskas
1991; Widschut 2006) or diet (Widschut 2006).

There are wide variations among studies with respect to the predictors of GWG, but pre-pregnancy
weight status seems to play a major role, being inversely associated with GWG. Four categories of
determinants of GWG have been identified, with complex interactions among them, which are still
incompletely understood and require further research. They include biological factors (such as
maternal pre-pregnancy BMI, age, parity, stature (Hytten 1991)), metabolic factors (variations in
insulin (Scholl and Chen 2002) and leptin (Stein et al. 1998) levels), social factors (e.g., maternal
education, socioeconomic status, smoking, substance abuse, unintended pregnancy, and health
provider advice regarding physical activity and diet) (Committee on the Impact of Pregnancy Weight
on Maternal and Child Health 2007) and, more recently, genetic factors (Dishy et al. 2003; Tok et al.
2006).

2.2.3.2 Public health significance - Prevalence data

In parallel with the increasing prevalence of obesity amongst women of childbearing age, there has
been an increase in the proportion of women with GWG exceeding IOM recommendations,
particularly in overweight and obese women, across all population groups (Committee on the Impact
of Pregnancy Weight on Maternal and Child Health 2007; Rasmussen and Yaktine 2009). US
national reports based on self-reported GWG recorded in birth certificates indicate a marked
increase in the proportion of women with singleton term pregnancy gaining more than 18.2 kg during
pregnancy from 16% in 1990 (National Center for Health Statistics 2007) to 21% in 2006 (National
Center for Health Statistics 2009), reaching a plateau in 2008 (National Center for Health Statistics
2010). At the same time, a small decline over time in the proportion of women gaining below the
recommended weight was also noted (National Center for Health Statistics 2010). More recent
figures based on data obtained from the Pediatric and Pregnancy Nutrition Surveillance System and
the current (2009) IOM recommendations indicate that less than one third of the white, non-Hispanic
US women gain within the recommended ranges, while 18.4% gain below and 52.6% above these
recommendations (Centers for Disease Control and Prevention 2010). There is no national
Australian data regarding weight gain in pregnancy.

35
Theoretical framework

2.2.3.3 Adverse health outcomes of excessive gestational weight gain

Independently of pre-pregnancy BMI, or in combination with it, excessive GWG has been associated
with a range of adverse health outcomes for the mother and the child, both short- and long-term.

Adverse health outcomes for the mother

Perinatal complications of excessive GWG for the mother include an increased risk of preeclampsia
(adjusted OR=2.31, 95% CI 2.15, 2.49 in normal weight women; adjusted OR=1.88, 95% CI 1.72,
2.06 in overweight women) (Cedergren 2006), caesarean section (adjusted OR=1.24, 95% CI 1.19,
1.29 in normal weight women; adjusted OR=1.23, 95% CI 1.17, 1.30 in overweight women)
(Cedergren 2006), and postpartum weight retention (adjusted OR=1.68, 95% CI 1.40, 2.01 in normal
weight women; adjusted OR=1.97, 95% CI 1.55, 2.50 in overweight women) (Oken et al. 2009;
Olson et al. 2003). In the longer-term, women with excessive weight gain during pregnancy have a
greater risk of developing obesity, in particular if associated with failure to lose weight postpartum
(Rooney and Schauberger 2002). Current evidence regarding interactions between GWG and pre-
pregnancy BMI in predicting these adverse maternal outcomes is inconsistent (Committee on the
Impact of Pregnancy Weight on Maternal and Child Health 2007).

Adverse health outcomes for the child

Children whose mothers gain weight excessively during pregnancy are more likely to be large for
gestational age at birth (adjusted OR=2.73, 95% CI 2.60, 2.83 in normal weight women; adjusted
OR=2.14, 95% CI 2.01, 2.28 in overweight women) (Cedergren 2006; Oken et al. 2007; Oken et al.
2009) and have a greater amount of body fat at birth (Catalano et al. 2003b). However, maternal
pre-pregnancy BMI seems to be a stronger predictor for these outcomes than GWG itself
(Committee on the Impact of Pregnancy Weight on Maternal and Child Health 2007). Birth weight
appeared to be best predicted by early pregnancy weight gain, while weight gain in the third trimester
was not associated with birth weight (Sermer et al. 1995).

The long-term adverse outcomes of excessive GWG described in the offspring include a greater risk
of obesity from childhood (Moreira et al. 2007; Oken et al. 2007; Schack-Nielsen et al. 2010;
Wrotniak et al. 2008) through to adulthood (Schack-Nielsen et al. 2010; Stuebe et al. 2009) and a
higher systolic blood pressure in childhood (Oken et al. 2007) and early adulthood (Mamun et al.
2009), the latter effect being potentially mediated by body size. Further details on the long-term
influence of maternal GWG on child obesity are presented in Section 2.4.3.

36
Theoretical framework

2.2.3.4 Clinical response - Recommended gestational weight gain

As mentioned above, given the accumulated data on adverse outcomes of excessive weight gain
during pregnancy, the IOM has recently updated their recommendations to provide appropriate
guidance both during pregnancy and before conception, with the ultimate goal of achieving healthy
outcomes for the women and their offspring.

A small randomised controlled trial of dietary advice to limit GWG to 6-7 kg in obese women showed
favourable effects on maternal insulin and leptin levels, reducing the risk of GD without lowering
maternal glucose levels, which are important to fetal growth (Wolff et al. 2008). Accordingly, it was
suggested that this kind of dietary counselling could in turn reduce the risk of obesity in the next
generation; however, more research is needed in this area.

2.2.4 Interrelations between maternal pre-pregnancy BMI, glucose tolerance status during
pregnancy and gestational weight gain

The intrauterine milieu undergoes major metabolic changes during pregnancy, which are largely
influenced by maternal body size, glucose tolerance status and gestational weight gain. These three
factors are not independent of each other, but rather interrelated, centred on the amount of energy
substrate passed on to the growing fetus, mainly glucose, and compensatory fetal insulin secretion.

Maternal pre-pregnancy BMI and glucose tolerance status during pregnancy

Obesity is a strong determinant for the development of glucose intolerance, not only in non-pregnant
states, but also during gestation (Catalano 2010). Maternal obesity at conception is often
accompanied by a (subclinical) reduction in pre-gravid insulin sensitivity (Catalano et al. 1999), which
is further enhanced with advancing gestation; as a result, a higher degree of insulin resistance is
observed in obese pregnant women compared to those with normal weight status. It has been
shown that obese women who develop GD have a greater insulin response, higher insulin resistance
and lower suppression of endogenous glucose production compared to obese women who maintain
normal glucose tolerance during pregnancy (Catalano et al. 1999).

The increased insulin resistance encountered in obese women leads to a greater availability of
nutrients (glucose, free fatty acids, and amino acids) to the fetus and some degree of placental
dysfunction (Yogev and Catalano 2009). Subsequently, hyperglycaemia stimulates fetal insulin
production, with secondary hyperinsulinemia, which contributes to fetal overgrowth.
37
Theoretical framework

Women who are overweight or obese when they enter pregnancy, even if they maintain a normal
glucose tolerance, are more likely to have infants with a greater amount of adipose tissue compared
to infants of women who are lean or average weight before conception (Sewell et al. 2006). This
excessive intrauterine fat deposition in the fetus has been attributed to the increased availability of
free fatty acids to the fetus as a consequence of maternal lipolysis secondary to the lower sensitivity
to insulin that characterises obese women (Catalano et al. 1995; Catalano et al. 2009b).

Moreover, infants of obese mothers, although not always heavier at birth, are more insulin resistant
(mainly peripherally, at skeletal muscle and adipose tissue level) compared to offspring of normal
weight mothers (Catalano et al. 2009b). This correlation between maternal pre-pregnancy BMI and
infant insulin resistance at birth was robust to adjustments for fat mass and percentage body fat,
suggesting potential genetic or epigenetic underlying mechanisms for fetal insulin resistance
(Catalano et al. 2009b).

It has been posited that in obese pregnant women, the subclinical chronic inflammatory state may
interfere with placental growth and function (Challier et al. 2008), with potential negative effects on
maternal insulin sensitivity and limited fat deposition, but excessive placental and fetal growth.
Accordingly, it was hypothesised that fetal obesity could be attributed to the transfer of
proinflammatory cytokines across the placenta (Catalano et al. 2009a). However, it has been shown
that most cytokines (TNF-α, IL-1β, and IL-6) do not cross the placenta (Aaltonen et al. 2005) and
thus maternal inflammation cannot be linked directly to obesity-related fetal inflammation (Catalano
et al. 2009a).

Maternal pre-pregnancy BMI and gestational weight gain

Among the factors affecting GWG, maternal weight status at entry to pregnancy appears to be
important. Obese women have an overall tendency towards a lower weight gain throughout
pregnancy (Chu et al. 2009; Lederman et al. 1997). The amount of fat stored during pregnancy has
been shown to be negatively associated with maternal pre-pregnancy BMI (Lederman et al. 1997) or
percentage body fat (Ehrenberg et al. 2003). Not only the amount, but also the distribution of fat
accrued during pregnancy differs according to pre-gravid body composition; thus women with less
than 25% body fat tend to accumulate fat more peripherally, while women with more than 25% body
fat tend to accumulate fat more centrally, irrespective of their gestational glucose tolerance status
(Ehrenberg et al. 2003).

38
Theoretical framework

Maternal glucose-insulin homeostasis during pregnancy and gestational weight gain

Studies on the association between indicators of maternal glucose homeostasis in early pregnancy
and GWG emerged after the finding of a significantly lower weight gain in insulin resistant, non-
pregnant young adult Pima Indians followed-up prospectively over 3 years (Swinburn et al. 1991).
However, the studies conducted in pregnant women on the association between insulin sensitivity
and weight gain have yielded discrepant findings. Scholl and Chen (2002) found that women with
higher fasting insulin levels in the first trimester of pregnancy (women regarded as having a ‘thrifty’
metabolism) had greater rates of GWG, with a suggestion of a stronger relationship in women with
normal weight. In contrast, Stuebe et al. (2010) did not detect an overall association between fasting
insulin levels or insulin resistance (as quantified by homeostasis model assessment) in early
pregnancy and total GWG; however, when analysed separately according to maternal weight status
in early pregnancy, hyperinsulinemia was associated with lower GWG in obese women and with
greater GWG in lean women (BMI<20 kg/m2). In a small cohort of lean women (with <25% body fat)
followed up at various times throughout pregnancy, it has been shown that women with reduced
insulin sensitivity prior to pregnancy gained significantly less fat mass in the first 12-14 weeks of
gestation compared to women without pre-gravid alterations of glucose-insulin metabolism, but the
amount of maternal fat accrued between early and late pregnancy was not associated with insulin
sensitivity (probably because of the greater contribution of fetal and placental tissues rather than
maternal fat to weight gain after the first trimester of pregnancy) (Catalano et al. 1998).

Gestational weight gain has also been specifically evaluated in relation to maternal glucose tolerance
status during pregnancy, with marked inconsistencies in findings. A systematic review concluded
there was weak evidence for an association between GWG and impaired glucose tolerance during
pregnancy (Viswanathan et al. 2008). Some of the discrepancies may be due to the different
methodologies employed by various researchers, mainly different criteria used for screening and
diagnosis of GD or milder degrees of glucose intolerance. Other difficulties in interpreting this
relationship arise from the fact that treatment recommended to women with GD, primarily dietary
advice, may also influence weight gain after the diagnosis. In order to overcome this limitation, some
studies focussed on weight gain prior to diagnosis of GD instead of total GWG. One of these studies
(Saldana et al. 2006) showed that weight gain ratio (observed : expected) was significantly
associated with an increased risk of developing impaired glucose tolerance among overweight
women, but only marginally associated with an increased risk of developing GD. Similarly, weight
gain in the first two trimesters of pregnancy (but not total GWG) was positively associated with the

39
Theoretical framework

likelihood of developing GD (Kieffer et al. 2001). In contrast, no difference in weight gain in the first
28 weeks of gestation was found between women with abnormal versus normal OGCT (Hackmon et
al. 2007).

Other authors reported lower total GWG in women with GD compared to women who maintained
normal glucose tolerance (Catalano et al. 1993a; Kieffer et al. 2006), but after adjustment for pre-
pregnancy weight for height this association remained significant only for underweight women
(Catalano et al. 1993a). On the other hand, other authors found no difference in total GWG or
percentage body fat between women with GD and controls (Ehrenberg et al. 2003).

2.3 Specific metabolic consequences in children

A range of metabolic perturbations have been identified in children exposed to intrauterine conditions
such as maternal pre-pregnancy obesity, gestational diabetes or borderline gestational glucose
intolerance, and excessive gestational weight gain, as described in Sections 2.2.1.3, 2.2.2.6 and
2.2.3.3, respectively. Although the reports regarding these changes are not entirely consistent, they
include child obesity, insulin resistance, impaired glucose tolerance, and higher systolic blood
pressure. The focus of this project is child obesity, body composition, fat pattern (Section 2.3.1), and
insulin resistance (Section 2.3.2). These outcomes are described in terms of public health
significance, namely prevalence, adverse health outcomes and methods of assessment. Both
sections conclude with a listing of the outcome measures in children to be used in the current study.

2.3.1 Obesity, body composition and fat pattern in children

“Obesity is neither a behavior nor a disease, but rather a general phenotype of numerous pathologic
biochemical processes impinging on a complex negative feedback pathway for the control of energy
balance” (Lustig 2003). However, the concept that the obesity epidemic is a simple consequence of
disturbances in energy balance has started to be challenged by recent research showing that the
determinants of obesity are a complex mix of genetic, environmental, psycho-social, and cultural
factors.

Obesity is a chronic, multifactorial condition resulting from the interaction between genetic
inheritance and environmental factors, which may act both during intrauterine life (e.g., GD) and after
birth (e.g., high calorie diet, reduced physical activity, behavioural, social and cultural processes).
40
Theoretical framework

From a genetic perspective, only a few monogenic obesity syndromes have been described, caused
by mutations in genes for hormones involved in the neuroendocrine regulation of energy balance, or
their receptors (such as melanocortin 4 receptor, proopiomelanocortin, leptin, leptin receptor)
(Chagnon et al. 2003; Farooqi 2006). These mutations, by influencing eating behaviour and/or energy
expenditure, lead to a positive energy balance starting in infancy (Mietus-Snyder and Lustig 2008).

Body weight status of each individual is influenced by a multitude of factors starting prenatally and
spanning the life course, including genetic, biological, environmental, behavioural, and socio-
economic factors, as well as the complex interactions between them (World Health Organization
2000). Risk factors for overweight and obesity in children suggested in the literature include parental
obesity (Beyerlein et al. 2010; Parsons et al. 1999), birth weight (partly affected by gestational age),
maternal smoking in pregnancy, rapid growth in the first two years of life, earlier adiposity rebound,
dietary factors (including exclusive formula feeding in infancy), sedentary lifestyle, and social factors
(such as low level of parental education) (Beyerlein et al. 2010; Parsons et al. 1999; Reilly et al. 2005).

Two obesity phenotypes were described over 50 years ago (Vague 1956) and they are still used in
contemporary studies (Landsberg 2008): upper-body obesity (central, abdominal, android, or ‘apple
shape’) with important metabolic complications, and lower-body obesity (peripheral, gluteal-femoral,
gynoid, or ‘pear shape’).

Total body adipose tissue can be broadly divided into two main components: subcutaneous
(superficial or deep in relation to the fascial plane) and internal, which in turn, has two sub-divisions:
visceral (intrathoracic, intra-abdominal and intrapelvic) and non-visceral (intramuscular,
intermuscular and paraoseal) (Shen et al. 2003). These compartments have different roles. Of them,
intra-abdominal adipose tissue is more strongly linked to adverse metabolic complications, including
insulin resistance, type 2 diabetes, dyslipidemia, and cardiovascular disease (Goran and Gower 1999).

Intra-abdominal adipose tissue is involved in modulating cardiovascular disease risk not only in
obese, but in lean individuals as well (Gower et al. 1999). The underlying mechanisms involve
secretion by the visceral fat of metabolically active factors (e.g., IL-6, TNF-α, adiponectin or resistin),
and release of free fatty acids into the hepatic portal vein, which increase hepatic glucose production
and interfere with hepatic insulin extraction (Frayn 2000).

Weight management in obese children and adolescents is multidisciplinary. According to the


severity of obesity, management may include behavioural lifestyle changes regarding dietary and
physical activity patterns, pharmacological approaches (e.g., Orlistat, a medication which reduces

41
Theoretical framework

the absorption of fat from food; Sibutramine, a medication which acts centrally to reduce food intake;
Metformin, an insulin sensitizer, recommended particularly in adolescents with insulin resistance)
and surgical interventions (e.g., laparoscopic banding) (Hearnshaw and Matyka 2010).
Pharmacotherapy and surgery are reserved for adolescents with severe obesity (Whitlock et al.
2010). Although medium- to high-intensity lifestyle interventions, sometimes combined with
pharmacotherapy, have short-term (i.e.,12 months) benefits in obese children and adolescents
(Whitlock et al. 2010), sustained benefits of obesity treatment are difficult to achieve (Oude et al.
2009), with individuals having a tendency to return to the previously stabilised ‘set point’ for body
weight status (Keesey and Hirvonen 1997; Oude et al. 2009). This has resulted in a major focus on
identifying preventative strategies, particularly upstream of current behaviours, which may be
responsible for the lack of response to current obesity management. In this context, investigating
whether and how intrauterine exposures may contribute to the development of obesity during
childhood is of great relevance.

2.3.1.1 Public health significance - Prevalence data

Childhood obesity has become a major public health issue in both developed and developing
countries throughout the world, due to the increasing prevalence, associated complications, and cost
to the health care system.

Obesity prevalence differs across countries and comparisons are limited by the use of different
reference populations, as well as by the inconsistent definitions of obesity (some based on weight for
height, or weight-for-height z-scores, others on BMI centiles or BMI z-scores) (Lobstein et al. 2004).
The use of BMI centile and BMI z-score is described in Section 2.3.1.3. A relatively recent review on
the prevalence of childhood overweight and obesity in over 60 countries around the world reported
an increase in all the countries (except for Russia and Poland in the 1990s) (Wang and Lobstein
2006). The average annual increase in childhood obesity prevalence was particularly high in
economically developed countries, and in societies that underwent rapid socio-economic transitions
(Wang and Lobstein 2006).

In the US, obesity (defined as BMI>95th centile of 1999-2000 reference population) prevalence in
children aged 6-11 years grew from 4% in 1971-1974 to 11.3% in 1988-1994 and further to 15.3% in
1999-2000 (Ogden et al. 2002). In the UK, the prevalence of obesity, as defined by the International
Obesity Task Force (IOTF) trebled between 1984 and 1998 (from 1.7% to 5.4% in boys and from
2.6% to 7.8% in girls), while the prevalence of overweight and obesity combined doubled over the

42
Theoretical framework

same period of time (from 9% to 20.7% in boys and from 13.5% to 27.4% in girls) (Lobstein et al.
2003; Wang and Lobstein 2006).

In Australia, based on national data (Australian Health and Fitness Survey, 1985; National Nutrition
Survey, 1995), the prevalence of childhood obesity (as defined by IOTF) tripled between the mid-
1980s and mid-1990s (from 1.4% to 4.7% in boys and from 1.2% to 5.5% in girls) and the prevalence
of overweight and obesity combined doubled over the same time period (from 10.7% to 20% in boys
and from 11.8% to 21.5% in girls) (Magarey et al. 2001). This equates to a rise in the prevalence of
overweight and obesity in school-age children by almost 1% per year (Wang and Lobstein 2006).
According to the latest report in 2007-2008, a quarter of Australian school-age children were
overweight or obese (Australian Bureau of Statistics 2009a). Based on data from the National
Health Survey, the prevalence of childhood obesity was still increasing in 2007-2008, but at a slower
rate, reaching 8% in school-age boys and 6% in girls, while the prevalence of overweight plateaued
at around 17% for both sexes at this age (Australian Bureau of Statistics 2009a). The prevalence of
overweight and obesity showed a socio-economic gradient, with higher rates in children living in areas
of greatest relative disadvantage (20% and 12%, respectively) (Australian Bureau of Statistics 2009a).

A flattening in the prevalence of childhood overweight and obesity over the last decade has been
described in several countries, such as Australia (Olds et al. 2010), USA (Ogden et al. 2008) and
France (Peneau et al. 2009). Olds et al (2010), while acknowledging paediatric overweight as a
public health issue, have indicated a stabilisation in the rates of overweight and obesity (as
measured by BMI) throughout childhood and adolescence in Australia after 1996, based on a meta-
analysis of 41 studies conducted between 1985-2008, with raw data available for over 70,000
children. As the authors suggest, this plateau may be a result of the multiple initiatives regarding
optimal physical activity and nutrition for children; a saturation of the obesogenic environment (so
that predisposed children have already become overweight, while the others are resilient to high-
caloric diets or sedentary lifestyle); or a potential underestimation of the issue by using a surrogate
measure of overweight such as BMI, as opposed to more sensitive measures of adiposity, such as
skinfold thickness (Olds et al. 2010). The latter explanation is supported by a meta-analysis
including 154 studies on skinfold thickness measurements in children and adolescents from
developed countries around the world (1951-2004), with raw data available for over 50,000 subjects
(Olds 2009). The average estimated increase in percentage body fat based on Slaughter equations
(Slaughter et al. 1988) was of 0.9% per decade, with the highest rates being identified in peri-

43
Theoretical framework

pubertal children (Olds 2009). Similarly, a secular increase in waist circumference in children has
been reported in Australia (Dollman and Olds 2006) and the UK (McCarthy et al. 2003).

2.3.1.2 Adverse health outcomes of childhood obesity

Childhood obesity has been associated with a wide range of adverse health outcomes, of variable
severity, both immediate (as described below) and in the long-term. Overweight and obesity
themselves, as well as related morbidities, tend to persist from childhood into adult years, probably
due to factors promoting obesity that also track over the lifespan, such as lifestyle habits (Freedman
et al. 2005a; Parsons et al. 1999; Whitaker et al. 1997). Reviewed data from multiple studies show a
greater likelihood of overweight tracking from childhood to adulthood with increasing excessive
weight and increasing age (Singh et al. 2008). Overweight children have at least double the risk of
becoming an overweight adult compared with healthy-weight children (Singh et al. 2008). Measures
of adiposity also track from childhood into adulthood. For instance, relatively strong correlations
were described between the age of 13 years and young adulthood for BMI (r=0.67, p<0.001), waist
circumference (r=0.56, p<0.001), triceps (r=0.51, p=0.003) and subscapular (r=0.52, p=0.002)
skinfold thickness (Steinberger et al. 2001). Moreover, increased cardiovascular mortality has been
documented in adults who were obese during childhood (Gunnell et al. 1998).

Obesity-related complications accumulate with time and are more often described in adults, but they
have started to be identified in children and adolescents as well. Like in adults, these complications
may affect almost every organ system, particularly in morbidly obese children and adolescents
(Daniels 2009; Han et al. 2010), as described below:

 cardiovascular: hypertension (Sorof and Daniels 2002), left ventricular hypertrophy


(Yoshinaga et al. 1995), early development of atherosclerotic lesions (Berenson et al. 1992),
endothelial dysfunction (Aggoun et al. 2008; Peña et al. 2006), and impaired arterial
distensibility (Banach et al. 2010), chronic inflammation (Sacheck 2008), unfavourable levels
of haemostatic factors (Alessi et al. 2000; Ferguson et al. 1998);

 metabolic and endocrine: dyslipidemia (i.e., high triglyceride levels and low high-density
lipoprotein cholesterol levels) (Daniels 2009), insulin resistance (Goran et al. 2003;
Steinberger et al. 2001), impaired glucose tolerance (Sinha et al. 2002), early onset of type 2
diabetes (American Diabetes Association 2000; Goran et al. 2003; Pinhas-Hamiel et al. 1996),
metabolic syndrome (Weiss et al. 2004), unmasking symptoms of polycystic ovary syndrome

44
Theoretical framework

(Franks 2008), accelerated sexual maturation in girls (Rosenfield et al. 2009) and delayed
transition to puberty in boys (Biro et al. 2006);

 gastrointestinal: non-alcoholic fatty liver disease (Nanda 2004), cholelithiasis (Kaechele et al.
2006), iron deficiency (due to reduced absorption from the gut) (Nead et al. 2004), vitamin D
deficiency (due to increased storage in the adipose tissue) (Alemzadeh et al. 2008), gastro-
oesophageal reflux (Koebnick et al. 2010)

 pulmonary: obstructive sleep apnoea (Mallory et al. 1989), asthma (Rodriguez et al. 2002),
limited exercise tolerance (Norman et al. 2005);

 renal: glomerular hyperperfusion and hyperfiltration (Srivastava 2006);

 orthopaedic: lower limb malalignment (e.g., tibia vara, slipped capital femoral epiphysis)
(Daniels 2009; Dietz 1998), osteoarthritis (Han et al. 2010);

 central nervous system: pseudotumour cerebri (disorder characterised by increased


intracranial pressure, presenting with headaches and possibly visual impairment in severe
cases) (Dietz 1998);

 psychosocial: depression (Britz et al. 2000), stigmatisation, social isolation, negative self-
esteem (Griffiths et al. 2010), and overall impaired quality of life (Schwimmer et al. 2003).

The consequences of childhood obesity on the healthcare system have also started to be
documented, but data is still scarce and largely limited to the USA (Trasande 2011). There is
ambiguity in reports on obesity-related expenditures, as they employ different methodologies and
consider different cost components (direct versus indirect). Based on American data from the 2001-
2003 Medical Expenditure Panel Survey, it is currently estimated that the annual total healthcare
expenditures (including drug prescription, outpatient visits and emergency room costs) for
overweight and obese children (aged 6-17 years) are US$180 and US$220 per person, respectively,
higher than for normal weight children (Finkelstein and Trogdon 2008). In contrast, the only
European quantification of the economic impact of childhood obesity based on the cross-sectional
German Interview and Examination Survey for Children and Adolescents, did not show a significant
increase in the annual total costs associated with childhood overweight or obesity (Wenig 2010).
However, the German data indicated higher utilisation of healthcare services and a significant
increment in annual physician costs of €27 and €62 per person among overweight and obese
children, respectively, relative to their normal weight peers (Wenig 2010). Additionally, costs

45
Theoretical framework

associated with adult complications of obesity need consideration, due to the high rate of persistence
of childhood obesity into adult life.

2.3.1.3 Measures of obesity

There is no universally accepted classification of overweight and obesity in children. A number of


ways to measure obesity have been proposed and used, from broad indicators of body size to
measures of adipose tissue at specific sites. This section describes and critiques some of these
measurement tools.

BMI and BMI z-score

The most pragmatic definition of overweight, and thus most extensively used for screening and in
epidemiological studies, is centred on BMI, which is calculated as the ratio of weight (in kilograms)
divided by the square of height (in metres). In the international classification of adult overweight and
obesity, BMI cut-off points are 25 kg/m2 and 30 kg/m2, respectively, in both females and males
(World Health Organization 1995). However, these preset cut-offs cannot be used for children or
adolescents, as during these stages of life, growth patterns vary markedly with age and sex, with
BMI being greater in boys at all ages (Berkey et al. 2000). BMI increases steeply during infancy
peaking between 6 and 12 months, then falls during toddlerhood and the preschool years
(Silverwood et al. 2009), and starts to increase again around the age of 5-6 years, which is known as
‘adiposity rebound’ (Reilly et al. 2005). The maximum annual increase in BMI occurs in early puberty
(0.7 kg/m2/year in 11 year-old girls and 0.8 kg/m2/year in 12 year-old boys (Berkey et al. 2000)). BMI
variation during childhood could be attributed to changes in fat-free mass among lean children
(Freedman et al. 2005b), particularly in boys (Neovius et al. 2004), while among obese children this
BMI variation is mainly due to changes in fat mass (Freedman et al. 2005b).

To account for this inconsistent change in weight relative to height across childhood, age- and sex-
specific centile curves have been created using large samples from different populations, which
allow comparisons between children over time, as well as between children of different ages and
sex. These centiles can be internal (which are less valid if derived from a small sample) or external.
Two of the external centiles widely used are the ones adopted by the Centers for Disease Control
and Prevention (CDC) and by the International Obesity Task Force (IOTF). In the USA, BMI values
are compared to BMI age- and sex-specific centiles curves derived from five National Health and
Nutrition Examination Surveys (1963-1994) and five supplementary data sources; the 85th and 95th

46
Theoretical framework

percentiles are considered the lower cut-offs for ‘at risk for overweight’ and ‘overweight’, respectively,
and have been adopted by CDC (Kuczmarski et al. 2000). In most other settings, the current
definition of childhood overweight and obesity is the one adopted by IOTF (Cole et al. 2000), which is
gaining increasing worldwide acceptance. The method used to construct these less arbitrary age-
and sex-specific centile curves for BMI was developed based on pooled international data (large
heterogeneous sample) collected from surveys conducted between 1963 and 1993, in six countries
(Brazil, Hong Kong, Singapore, the Netherlands, UK and USA) (Cole et al. 2000). According to
IOTF/Cole classification, the lower thresholds for ‘overweight’ and ‘obesity’ in children are defined by
the age- and sex-specific centile curves corresponding to a BMI at 18 years equal to the adult cut-off
points of 25 kg/m2 and 30 kg/m2, respectively (Cole et al. 2000). In addition to these external
centiles, national systems of classifying childhood obesity based on local (internal) BMI reference
curves are available in some countries (Guillaume 1999), useful mainly for national purposes and not
for international comparisons.

Another BMI-related measure commonly used for comparisons of obesity rates among children over
time or from different populations is the age- and sex-specific z-score. Essentially, this is a measure
of how far a child’s BMI departs from the mean. In more detail, BMI z-score is computed as the
deviation of one individual’s BMI from the mean BMI for the reference population, divided by the
standard deviation for the reference population. This calculation can be based on the widely used
LMS method (lambda-mu-sigma method (Cole 1988)), which summarises the changing distribution
by three curves representing the median (M), coefficient of variation (S) and skewness (L), the latter
expressed as a Box-Cox power (Cole and Green 1992). It was suggested that among the metrics of
relative weight for height in children, BMI z-scores would explain the variation in percentage body fat
better than BMI centiles (Field et al. 2003).

Although simple, cheap, reproducible and extremely useful for screening of overweight status in both
clinical settings and epidemiological studies, BMI and related measures have some limitations as a
proxy indicator for adiposity because they do not depict the relative contributions of lean and fat
mass to total body mass, nor the fat distribution. It has been shown that BMI has high specificity
(98%), but a low to moderate sensitivity (36% in boys, 60% in girls) for identifying excess adiposity in
children (Reilly et al. 1999).

A significant variation in body composition can be found for a given BMI at any age. Individuals with
increased muscle mass may have a BMI as high as those with excessive deposition of fat, hence
measuring only an index of relative weight to height such as BMI may be a source of

47
Theoretical framework

misclassification of overweight or obese. To minimise this misclassification and better identify


individuals at risk for adverse health outcomes, various methods of body composition assessment
have been developed to gauge the amount of fat more directly.

Percentage body fat

As opposed to BMI which is a marker of relative weight for height, percentage body fat (%BF) is a
more specific measure of adiposity, indicating which overweight individuals are truly overfat rather
than having increased muscle mass. Variations in %BF reflect changes both in fat mass (FM) and
fat-free mass (FFM).

Body composition varies considerably with age, sex, race, and is influenced by physical activity level,
diet, underlying medical conditions (e.g., FM is reduced in malnutrition), and hydration status (Forbes
1978). Physiologically, after birth, the amount of body fat declines slowly until the end of infancy,
when it starts to progressively accumulate throughout childhood, at similar rates in both girls and
boys (Forbes 1978). With puberty, a sexual dimorphism of body fat deposition becomes apparent,
with females gaining proportionately more fat, while males gain more muscle (Forbes 1978).

Due to these variations in body composition, age- and sex-specific body fat reference curves have
been developed, similar to the BMI centile curves. For example, %BF estimated by dual energy X-
ray absorptiometry (method described below) were linked to the IOTF cut-offs for overweight and
obesity in a study involving 661 white children aged 3-18 years, from New Zealand (Taylor et al.
2002). In this study, %BF ranged from 18-23% in overweight boys, 20-34% in overweight girls, 24-
36% obese boys, and 26-46% in obese girls (Taylor et al. 2002), which supported the need for age-
and sex-specific definitions to classify children as overfat or obese based on %BF, instead of using
one single %BF value as a cut-off. McCarthy et al. (2006) did this by constructing age- and sex-
specific %BF reference charts and classified children with a %BF less than the 2nd centile as
underfat, while the 85th and 95th centiles defined the lower limits for overfat and obese, respectively
(corresponding to the overweight and obese categories of the IOTF reference curves of BMI). Their
reference population was UK Caucasian children aged 5-18 years who had bioelectrical impedance
analysis (method described below) in 1985.

A multitude of methods for estimating body fat are now in use, with different levels of accuracy,
complexity, feasibility, and cost. They range from equations based on simple anthropometric
measurements (e.g., skinfold thickness) or bioelectrical impedance analysis, to more complex
48
Theoretical framework

imaging techniques, such as dual energy X-ray absorptiometry, computed tomography, or magnetic
resonance imaging, as well as volume displacement methods (underwater weighing, air
displacement plethysmography), or dilution techniques (Lee and Gallagher 2008). All these in vivo
methods of body composition assessment rely on numerous assumptions which may not always be
met, and therefore cannot yield estimates as accurate as the direct chemical analysis (Goran 1998),
but the latter method is not feasible in human studies. Having different levels of accuracy, these
techniques do not provide identical estimates for body composition in the same individual, which
makes interpretation and comparisons challenging. The most accurate and feasible technique,
which is thus considered as the reference, is the 4-compartment model (Radley et al. 2009). Within
this method, fat or fat-free mass are estimated based on the measurement of the other three
components: total body water, body density and body mineral content (Radley et al. 2009). This
technique is difficult to perform, often uncomfortable (e.g., when body density measurement involves
underwater weighing) and time consuming, which are major issues especially in child studies.

Skinfold thickness (SFT) measurements in different body sites (e.g., biceps, triceps, subscapular,
abdominal, suprailiac, medial calf) offer a good indicator of subcutaneous fat amount and distribution
(upper versus lower body) (Mei et al. 2002). Unlike BMI, SFT measurement can identify non-
overweight children who have excess subcutaneous adiposity (Freedman et al. 2007b). However,
SFT measurements give no indication on the amount of visceral fat, which is the metabolically active
component. Body fat estimates from SFT are prone to examiner’s error and intensive training is
imperative to maximise reliability of measurements.

Numerous equations have been proposed for estimating %BF from SFT in children (Deurenberg et
al. 1990; Slaughter et al. 1988), with no gold standard decided upon so far. Among them, the
Slaughter equations (sex-specific equations for pre-pubertal, pubertal and post-pubertal stage)
(Slaughter et al. 1988) provide %BF estimates most highly correlated with dual energy X-ray
absorptiometry estimates, and without fat-dependent bias (Rodriguez et al. 2005). Simpler equations
such as the sum of triceps and subscapular SFT, which correlates better with DXA estimates than
BMI, have proved useful for comparisons between subjects, but they are meaningless as standalone
figures (Freedman et al. 2007b).

Computed tomography (CT) and magnetic resonance imaging (MRI) have the major advantage
of providing measures of regional fat distribution (subcutaneous, visceral, intermuscular) (Lee and
Gallagher 2008). However, due to their technical complexity, high cost, and, with CT, radiation

49
Theoretical framework

exposure, their use is limited mainly to clinical practice and few epidemiological studies for validation
of ‘field’ methods (e.g., bioelectrical impedance analysis) (Mattsson and Thomas 2006).

Dual energy X-ray absorptiometry (DXA) is based on a 3-compartment model which estimates the
differential absorption of X-ray energy by fat, bone and other lean tissue. Although precise with
respect to body composition assessment in paediatric populations and strongly correlated with the
four-compartment model (Sopher et al. 2004), DXA is not very accurate, particularly in obese
children and adolescents, with a tendency to overestimate FM and underestimate FFM in
comparison to the four-compartment model (Wells et al. 2010). Other limitations of this method
include the high cost, the lengthy time required, and that it measures all fatty elements in the body,
not only the adipose tissue (a limitation of all two- and three-compartment models).

Bioelectrical impedance analysis (BIA) determines the resistance of the body tissues to a small
alternating electrical current (Kyle et al. 2004a). The adipose tissue impedes the electric current
more than muscle or bone. The method then calculates the resistance index (as the ratio of height
squared and resistance) to predict total body water (TBW) and, implicitly, FFM (as adipose tissue is
almost anhydrous) (Wright et al. 2007). Fat mass is subsequently calculated by subtraction of FFM
from the total body weight. Fat and lean mass vary with height and thus cannot be interpreted as
standalone figures; instead, FM can be expressed as a percentage of total weight (%BF), which is an
indicator of lean and fat mass adjusted for body size.

BIA can be performed using the arm-to-leg method (with electrodes on the wrist and ankle, such as
Bodystat), the foot-to-foot method (using a stand-on machine, such as Tanita), and tetrapolar, with
comparable results (Wright et al. 2007).

The accuracy and precision of this method in children have been extensively debated, with some
authors describing the method as acceptable (Schaefer et al. 1994), while others as biased and
imprecise (Reilly et al. 1996). It has been suggested that the lack of correlation with estimates from
more accurate methods “reflects limitations in the regression equations used which, by necessity,
have been developed with small numbers of children over wide age ranges” (Wright et al. 2007).
Data from a cross-sectional study of 30 overweight or obese Australian children revealed that among
the embedded equations for BIA, the Schaefer equation offered the %BF estimate closest to that
estimated by DXA (Cleary et al. 2008).

When comparing BIA estimates to anthropometric measurements, some studies support BIA as a
more precise tool for estimating adiposity than BMI (Rush et al. 2003; Tyrrell et al. 2001) or SFT

50
Theoretical framework

(Kettaneh et al. 2005; Pecoraro et al. 2003). In contrast, data from a cross-sectional study in non-
obese pre-menarcheal girls found BIA no better than triceps SFT in predicting %BF (Bandini and Vu
1997). Similarly, weaker correlations between BIA and anthropometric measurements have been
reported in pubertal boys (Kettaneh et al. 2005).

Relative to DXA, which is often regarded as a reference for body composition assessment
particularly in children, it has been noted that BIA tends to underestimate %BF in overweight
subjects and overestimate it in lean subjects (Eisenmann et al. 2004). However, in a small (n=17)
cross-sectional study in overweight and obese preadolescent children, despite the significant
differences between BIA and DXA estimates, BIA measures of body composition were strongly
correlated to the ones provided by DXA (correlation coefficients for %BF, FM, and FFM were 0.85,
0.97, and 0.94, respectively) (Goldfield et al. 2006).

Relative to the 4-compartment model, BIA (using Tanita in a sample of overweight and obese
children) provided accurate estimates of body composition at a group level, but not at individual
level, with a tendency to overestimate FFM (Radley et al. 2009). Foot-to-foot BIA overestimates
FFM and underestimates FM in both obese and non-obese children, compared with air-displacement
plethysmography (Azcona et al. 2006). The magnitude of this problem is hard to gauge from the
available data, but over- and underestimation appeared to be in the order of 5%.

Despite the debatable accuracy for estimating %BF, BIA is highly suitable for epidemiological studies
in the field, particularly in children, as it is portable, easy to perform, non-invasive, quick and
increasingly inexpensive. Other advantages of this method are the minimal intra- and inter-observer
variability (Diaz et al. 1989) and the good reproducibility, with less than 1% error on repeated
measurements in the same conditions (Segal et al. 1991), which permits ranking of individuals within
a study.

Given these advantages, BIA was the method of choice for this research project. However, although
BIA is a reasonable method for estimating body composition, it does not allow for the description of
fat distribution, or the separate quantification of intra-abdominal (visceral) fat, which is the most
metabolically active.

51
Theoretical framework

Fat pattern

Fat distribution within specific body regions, or even within non-adipose tissue, is of great interest in
terms of metabolism. Two major phenotypes of relative distribution of the excess fat, with unique
characteristics, have been described several decades ago in relation to cardiovascular and
metabolic disease risk: central and peripheral adiposity (Ashwell et al. 1985; Vague 1956).

The centralised distribution of fat has been linked to a cluster of metabolic disturbances including
fasting hyperinsulinemia, insulin resistance, blood lipid disorders, as well as high blood pressure
(Browning et al. 2010), both in adults (Despres and Lemieux 2006; Harris et al. 2000; Li and
McDermott 2010) and children (Botton et al. 2007; Garnett et al. 2008; Hara et al. 2002; Kahn et al.
2005; Savva et al. 2000). Abdominal adiposity may be regarded as an endocrine organ secreting
various hormones (e.g., leptin, adiponectin) and cytokines (e.g., TNF-α, IL-6), which lead to hepatic
lipogenesis and insulin resistance, release of free fatty acids from adipocytes, macrophage infiltration
into the adipose tissue, or ectopic lipid storage (Kershaw and Flier 2004).

Similar to %BF, fat pattern can be estimated based on simple anthropometric measures, such as
waist circumference (paediatric centile curves are available (McCarthy et al. 2001)), waist-to-hip
ratio, waist-to-height ratio, sum of SFT (e.g., triceps, biceps, subscapular, suprailiac, abdominal), or
ratio of central to peripheral SFT (e.g., triceps to subscapular skinfold thicknesses). More accurately,
the central pattern of adiposity may be determined by the same imaging techniques (CT, MRI or
DXA) used for body composition assessment, which, due to the high cost, are usually not feasible in
epidemiological studies.

Waist circumference is a simple measure of central adiposity, with a relatively good sensitivity in
predicting cardiovascular risk. It measures both subcutaneous and visceral adipose tissue at
abdominal level. Waist circumference is more informative when used as a ratio to other body
measures. The ratio of waist-to-hip has been used as an indication of ‘apple’ or ‘pear’ shape. More
recent research has suggested the ratio of waist-to-height is a better predictor of cardiovascular risk
(Browning et al. 2010).

Waist-to-height ratio (WHtR) was first described in the Framingham Study in 1988 in an attempt to
account for some of the limitations of BMI in predicting cardiovascular disease risk in children and
adolescents (Higgins et al. 1988) and was later adopted by other investigators as an indicator of fat
distribution. This ratio has been validated as a good anthropometric measure of abdominal obesity
both in adult (Ashwell et al. 1996; Cox et al. 1997; Hsieh and Muto 2005) and paediatric populations

52
Theoretical framework

(McCarthy and Ashwell 2006; Nambiar et al. 2009). However, it cannot differentiate between intra-
abdominal (visceral), the metabolically active depot (Carr et al. 2004), and subcutaneous abdominal
adipose tissue, a differentiation which requires more expensive imaging techniques.

Several studies indicated an improved ability of WHtR to predict adverse cardiovascular risk factors,
such as high blood pressure and lipid disorders in children and adolescents, compared to BMI (Kahn
et al. 2005; Savva et al. 2000), whereas other studies did not support WHtR superiority, probably
due to the high correlation with BMI (Freedman et al. 2007a).

In contrast to other anthropometric measurements in children, WHtR does not significantly vary with
age and therefore expression as z-scores is thought to be not warranted (Freedman et al. 2007a).
Similar to adults, a cut-off of 0.5 for WHtR has been proposed to define an exacerbated cardiovascular
risk in children (McCarthy and Ashwell 2006). It has been argued that using a single cut-off for all
ages provides a simplified public health message with no loss in accuracy: “Keep your waist
circumference to less than half your height” (Ashwell and Hsieh 2005; McCarthy and Ashwell 2006).
However, despite the practicality of such a simple message, relationships involving WHtR are likely
to be continuous.

Data from a large community-based cross-sectional study, the Bogalusa Heart Study, support the
utility of 0.5 as a cut-off for WHtR in detecting children at cardio-metabolic risk, even among those
with normal weight (Mokha et al. 2010). Normal weight children with central adiposity (defined by
WHtR>0.5), also known as ‘metabolically obese normal weight’ (Stefan et al. 2008), have more
unfavourable cardio-metabolic risk profiles (higher systolic and diastolic blood pressure, abnormal
lipid profile, increased levels of fasting glucose and insulin, higher insulin resistance) than normal
weight children without central adiposity (Mokha et al. 2010). In contrast, overweight and obese
children without central adiposity have lower levels of these cardio-metabolic variables
(‘metabolically benign obesity’ (Stefan et al. 2008)) compared to those with centralised distribution of
the fat (Mokha et al. 2010).

In summary, the following obesity-related outcomes in children are considered in the current study as
continuous variables: (1) BMI z-score using the IOTF age- and sex- specific centiles curves, as a
measure of global obesity; (2) %BF estimated by BIA; and (3) fat pattern defined by WHtR.

53
Theoretical framework

2.3.2 Insulin resistance in children

Insulin resistance (IR) is a state characterised by an insufficient biological response of the tissues to
normal or higher levels of insulin (Matthaei et al. 2000). The diminished insulin action involves a
reduction in the glucose uptake by skeletal muscles and fat tissue, as well as a decreased
suppression of hepatic glucose production; consequently, hyperglycaemia arises, followed by
compensatory hyperinsulinemia (Matthaei et al. 2000; Savage et al. 2005). The critical role of IR in
human disease was recognised over 20 years ago (Reaven 1988).

There is evidence that the development of IR is influenced by several genes (Matthaei et al. 2000),
as well as obesity (Lee 2006), in particular visceral adiposity (Cruz et al. 2002), and also by
sedentary lifestyle, diet (e.g., rich in saturated and n-6 polyunsaturated fatty acids, with a high
carbohydrate to fat ratio, or low-fiber diet (Cañete et al. 2007)), puberty, sex, race/ethnicity, and
programming by some perinatal factors (such as small size at birth, prematurity) (Lee 2006). In
addition, more recent reports have considered intrauterine exposure to maternal obesity or GD as a
potential determinant of IR in the offspring, independent of genetic effect (see Sections 2.4.1 and 2.4.2).

Adipose tissue plays a central role in the development of IR, possibly through the production of free
fatty acids (‘portal theory’) and several adipocytokines (‘endocrine’ paradigm), such as adiponectin
(an insulin-sensitizing hormone), leptin, TNF–α, IL-6, and resistin. All these factors have been
implicated in the alteration of insulin signalling at various levels (Chiarelli and Marcovecchio 2008).
The underlying mechanisms responsible for IR are not entirely understood, but they include a
number of defects in the insulin signalling cascade, which affect phosphorylation of the insulin
receptor and inhibitory proteins (Matthaei et al. 2000; Savage et al. 2005).

Certain sites of fat deposition have been associated with increased risk of IR. First, visceral fat, with
its greater lipolytic activity relative to subcutaneous fat, generates larger amounts of free fatty acids
and glycerol to be carried to the liver, with subsequent effects on hepatic insulin sensitivity (Matthaei
et al. 2000). Second, lipid accumulation in muscular cells may alter insulin signalling at peripheral
levels (Weiss et al. 2005). Finally, it has been recently suggested that intermuscular thigh adipose
tissue, as opposed to the subcutaneous fat (Boettcher et al. 2009), as well as fat deposited around
blood vessels (Yudkin et al. 2005) may also contribute to the development of IR.

Adiposity is the most important risk factor for IR, which explains almost one third of the variance in
insulin sensitivity (Lee et al. 2006). Insulin resistance is strongly associated with obesity (Dwyer et
al. 2002; Lee et al. 2006), but it can occur in non-obese individuals as well, particularly if they are

54
Theoretical framework

hyperlipidemic (Al-Mahmood et al. 2007). Moreover, it has been suggested that IR is both a result
and a cause of obesity (Lustig 2003). Insulin resistance leads to hyperinsulinemia, which may
interfere with leptin signal transduction in the hypothalamus promoting leptin resistance, which
decreases resting energy expenditure and increases appetite, with subsequent weight gain (Lustig
2003).

A number of cross-sectional studies in children support the role of adiposity in the development of IR,
based on positive correlations between FM (measured by DXA) and IR in pre-pubertal children
(Gower et al. 1999), between central adiposity (estimated by SFT) and fasting insulin levels
(Freedman et al. 1987), or between visceral adiposity (measured by MRI) and fasting insulin and IR
(determined by the frequently sampled intravenous tolerance test) in obese Hispanic children (Cruz
et al. 2002).

In addition to the increased risk of IR associated with a family history of type 2 diabetes or a certain
ethnic background, a number of maternal factors have been recently identified as possibly
contributing to IR in children. They include GD (Boerschmann et al. 2010; Catalano et al. 2009b;
Egeland and Meltzer 2010), maternal diet during pregnancy (Shiell et al. 2000; Yajnik et al. 2008),
older maternal age (Loos et al. 2002), pregnancy-induced hypertension (Himmelmann et al. 1997),
low maternal BMI (Ravelli et al. 1998; Shiell et al. 2000), and prenatal psychosocial stress (Entringer
et al. 2008). The great majority of observational studies focussing on the early life origins of IR have
described a link with intrauterine growth restriction (Newsome et al. 2003) and also supported a role
for rapid catch-up growth after birth (Newsome et al. 2003; Ong and Dunger 2004; Veening et al.
2002). The explanations proposed for the negative association between IR and birth weight include
in utero ‘programming’ by poor intrauterine nutrition and direct effect of genes influencing IR, which
reduce insulin-related growth (‘fetal insulin hypothesis’) (Frayling and Hattersley 2001; Hattersley
and Tooke 1999).

Once clinically manifested (commonly associated with acanthosis nigricans, which appears as
hyperpigmented, velvety plaques at the base of the neck and in the flexures) IR deteriorates over
time if not treated (Quinn et al. 2009). Metformin, usually in association with lifestyle interventions,
has shown benefits in improving insulin sensitivity (along with a decrease in BMI) in children above
the age of 10 years, but due to the side effects (nausea, diarrhoea and abdominal discomfort) it is
not commonly recommended (Quinn et al. 2009).

55
Theoretical framework

Insulin resistance is recognised as a key feature of metabolic syndrome, which constitutes a cluster
of metabolic disorders including central obesity, impaired glucose tolerance, hyperglycaemia,
dyslipidemia (high level of triglycerides, low level of HDL-cholesterol) and high blood pressure
(Zimmet et al. 2005). Several definitions for metabolic syndrome are in use for adults, the most
commonly used being those proposed by the National Cholesterol Education Program (Expert Panel
on Detection Evaluation and Treatment of High Blood Cholesterol in Adults 2001), World Health
Organization (Expert Panel on Detection Evaluation and Treatment of High Blood Cholesterol in
Adults 2001), and International Diabetes Federation (Alberti et al. 2005). Only recently has a
consensus definition of metabolic syndrome in children and adolescents been established and
endorsed by the International Diabetes Federation (Zimmet et al. 2007), which thus replaces the use
of disparate definitions based on modified criteria for adults (Cook et al. 2003; Cruz et al. 2004; Ford
et al. 2005; Weiss et al. 2004). The use of various definitions for metabolic syndrome has
contributed to reporting of different prevalence rates in the population.

2.3.2.1 Public health significance – Prevalence data

Insulin resistance in children and adolescents, largely confined to those who are overweight or
obese, has become a topic of interest following the rising prevalence of obesity in the paediatric
populations and the detection of type 2 diabetes at younger ages (Kaufman 2002).

Most published reports refer to the prevalence of metabolic syndrome, and not specifically to the
prevalence of IR in children and adolescents. For example, US estimates based on the National
Health and Nutrition Examination Survey 1999-2000 data indicate that 6.4% of adolescents (12-19
years) have a metabolic syndrome phenotype (32.1% of obese, 7.1% of overweight and under 1% of
normal weight adolescents), an increase from the 4.2% reported from the 1988-1992 data collection
(Duncan et al. 2004). Lower rates of metabolic syndrome (about 5%) were reported in younger (7-9
year-old) US children (DuBose et al. 2006) compared to adolescents.

A recent review of 36 studies published between August 2007 and January 2009, including general
population samples and samples of overweight or obese children and adolescents (only one of these
studies being conducted in Australia, in Aboriginal children (Sellers et al. 2008)) explicitly describing
the definition employed for the diagnosis of metabolic syndrome (with clear cut-off values for all the
risk factors), concluded that the mean prevalence of metabolic syndrome in children and adolescents
worldwide is 10%, ranging between 2% in normal weight individuals and 32% among obese (Tailor
et al. 2010). Based on these data, it appears that although the prevalence of metabolic syndrome is

56
Theoretical framework

not particularly high in children, it is 15 times higher amongst obese children, supporting screening
for metabolic syndrome in this group (Tailor et al. 2010). This review also identified a higher
prevalence of metabolic syndrome in males than females, probably due to higher rates of central
obesity (Tailor et al. 2010).

As mentioned, very few studies have reported prevalence of IR alone in children. Data from a study
of American adolescents showed that more than half of the obese adolescents had increased IR, as
determined by homeostasis model assessment (HOMA-IR >4.39) (Lee et al. 2006). Despite using
less strict criteria for defining IR, two European studies reported somewhat lower rates in 10 year-old
obese children from Crete (31%, based on HOMA-IR>2.1) (Manios et al. 2008) and Southern Italy
(40.8%, using a cut-off for HOMA-IR of 2.5) (Valerio et al. 2006). The difference between the US
and latter two studies is likely due to the increase in IR during puberty.

Of concern, type 2 diabetes has also emerged as a health issue in obese children and adolescents,
especially in individuals with family history, or in high risk ethnic groups, such as African-American,
Hispanic-American, Pima Indians, Americans with an Asian descent (American Diabetes Association
2000) or the Indigenous Australian population (Maple-Brown et al. 2010). An increase in the
prevalence of type 2 diabetes in the past decade has been documented in the US and the UK. For
example, while in 1998 six per 1000 British children were prescribed an oral antidiabetic drug, in
2005 significantly more (50 per 1000) children received this type of medication (Hsia et al. 2009).

2.3.2.2 Adverse health outcomes of insulin resistance

Insulin resistance is regarded as a pathophysiologic link between adiposity and associated metabolic
disorders and chronic cardiovascular disease (Lee 2006; Weiss and Kaufman 2008). Complications
associated with IR are well documented in adults and recent studies have shown that they may also
occur in children and adolescents, as described below:

 impaired glucose tolerance and type 2 diabetes (Alberti et al. 2004), when pancreatic β-cell
function fails to compensate for IR;

 high blood pressure (Maffeis et al. 2010), presumably due to an insulin-mediated effect on the
sympathetic nervous system (Landsberg 1999) and on renal sodium reabsorption (Hall 1997);

 abnormal lipid profile (high triglyceride and low-density lipoprotein cholesterol levels, and low
high-density lipoprotein cholesterol level), premature development of atherosclerosis (Le et al.
2010);

57
Theoretical framework

 enhanced blood coagulability due to higher levels of plasminogen activator inhibitor-1 and
fibrinogen (Matsuzawa 2005);

 hepatic steatosis characterised by the accumulation of triglycerides in hepatic cells as a result


of the increased flux of free fatty acids to the liver and the enhanced hepatic lipogenesis
(Browning and Horton 2004);

 systemic inflammation, potentially mediating the relationship between IR and asthma (Al-
Shawwa et al. 2007).

It is known that, similar to overweight and obesity, IR and associated morbidities track from
childhood into adult life, independently of (Sinaiko et al. 2006) and synergistically with BMI (Morrison
et al. 2009). This suggests that metabolic syndrome in adults could be, at least partly, prevented by
early detection of IR during childhood followed by appropriate management. Given all these adverse
health outcomes of IR, research directed towards the identification of early life factors which
contribute to alterations in insulin sensitivity in childhood deserves priority.

2.3.2.3 Assessment of insulin resistance in children

A range of tools have been developed for assessing insulin sensitivity in humans and most of them
have been validated in children (Lee 2006). They vary in complexity and the conditions in which they
are performed (steady- or dynamic-state), which dictate their application depending on the purpose
of the study. Some of these methods provide a direct measure of insulin sensitivity by quantifying
insulin-mediated glucose uptake under steady-state conditions (such as hyperinsulinemic-
euglycaemic glucose clamp or the insulin suppression test), while others (such as the minimal model
analysis of frequently sampled intravenous glucose tolerance test and the oral glucose tolerance
test) use dynamic data to generate an indirect estimate of insulin sensitivity (Muniyappa et al. 2008).
However, all these methods are impractical in studies with large numbers of participants, especially
children. As an alternative, a number of equations based on fasting insulin and glucose have been
proposed and validated for clinical research and epidemiological studies of children (Gungor et al.
2004). They include fasting insulin level, fasting glucose to fasting insulin ratio (FGIR), the
homeostasis model assessment, the quantitative insulin sensitivity check index (QUICKI). The
Belfiore glycaemia and McAuley index (McAuley et al. 2007) are used less commonly. The most
widely used methods for estimating IR are briefly described and critiqued below.

58
Theoretical framework

The gold standard method for assessing IR in humans is the hyperinsulinemic-euglycaemic


clamp, valid both in adult (Morey 2003) and child populations (Arslanian 2005). It is the only reliable
method to identify IR in individuals with frank hyperglycaemia or diabetes (Morey 2003). This
technique involves intravenous administration of insulin to raise plasma insulin to a certain level,
followed by an infusion of 20% dextrose at variable rates, thereby ‘clamping’ the plasma glucose
concentration to a basal level. Insulin resistance is inversely proportional to the amount of glucose
required to maintain basal glucose levels. The technique is extremely complex and invasive,
requiring double vascular access, one for insulin and glucose infusions, and one for frequent blood
sampling to measure plasma glucose level (every 2-5 minutes) and insulin level (every 10-15
minutes) over 3 hours (Arslanian 2005).

Another method of insulin sensitivity assessment is the insulin suppression test. This technique
relies on the intravenous administration of somatostatin, which suppresses endogenous insulin
production, followed by a simultaneous infusion of insulin and glucose (to ensure steady-state
conditions) for 3 hours, during which repeated blood samples are collected (Muniyappa et al. 2008).
Insulin sensitivity is inversely proportional to the plasma glucose concentration measured in the
steady-state period (Muniyappa et al. 2008).

Among the indirect methods of estimating IR is the frequently sampled intravenous glucose
tolerance test (FSIVGTT), which involves intravenous infusion of a certain amount of glucose,
followed by a bolus of insulin, and multiple blood sampling at exact timing usually over a period of 3
hours; plasma glucose and insulin levels at specific time points are then used to derive an index of
insulin sensitivity (Muniyappa et al. 2008).

Alternative methods have been proposed based on a 2-hour oral glucose tolerance test (OGTT),
which requires multiple blood sampling (at baseline, 30, 60, and 120 minutes following the oral
administration of 1.75 g/kg glucose). Plasma glucose and insulin concentrations measured during
this test are used in specific equations to derive whole body insulin sensitivity indices (Matsuda and
DeFronzo 1999; Soonthornpun et al. 2003). OGTT, although less cumbersome compared to the
hyperinsulinemic-euglycaemic clamp, is still impractical in paediatric epidemiological studies as it
involves multiple blood sampling and requires co-operation over a prolonged period of time
(Muniyappa et al. 2008).

Although the four methods outlined above provide valuable information for physiological studies and
in clinical settings, they are expensive, time-consuming and laborious (they require intravenous

59
Theoretical framework

catheters and multiple blood sampling), hence not suitable for application in large epidemiological
studies, especially of children (Morey 2003). In contrast to these invasive methods, simple proxy
indices for insulin sensitivity have been derived based on fasting plasma insulin and glucose
concentrations. These methods, whilst still requiring a fasting blood sample, are less invasive and
less time consuming, hence more acceptable to use in epidemiological studies of children.

Among the indices of insulin sensitivity derived from fasting measurements of insulin and glucose,
homeostasis model assessment of insulin resistance (HOMA-IR) is now the most commonly
used in epidemiological and clinical research. HOMA-IR has a physiological basis and reflects the
feedback loop between tissues involved in glucose regulation, namely the liver (hepatic glucose
production) and the β-cells (insulin secretion), in homeostatic (fasting) conditions (Matthews et al.
1985). The original model was derived from mathematical interpretations of the interaction between
IR and β-cell function; these estimations are then used to predict fasting steady-state insulin and
glucose concentrations (Matthews et al. 1985). The original model approximates IR according to the
following formula (simplified from iterative equations):

HOMA-IR = Fasting plasma insulin (µU/ml) x Fasting plasma glucose (mmol/l) / 22.5

(Matthews et al. 1985)

The denominator (22.5) is an adjustment to normal population levels; it is the product of the fasting
plasma insulin (5 μU/ml) and fasting plasma glucose (4.5 mmol/l) levels considered normal in a
healthy individual (Matthews et al. 1985).

Estimates of IR derived from HOMA-IR have been validated against more complex methods within
individuals with normal or impaired glucose tolerance, and mild forms of diabetes, provided the blood
collection takes place in steady-state conditions (Wallace et al. 2004). High correlations have been
identified in adults between insulin sensitivity estimated by HOMA-IR and measured by the
hyperinsulinemic-euglycaemic clamp (r=0.88, p<0.0001 (Matthews et al. 1985); r=0.73, p<0.0001
(Katsuki et al. 2001)), and between HOMA-IR and the index derived from the minimal model
FSIVGTT (r=0.7, p<0.001 (García-Estévez et al. 2003). The validity of HOMA-IR has been assessed
in various populations of children and adolescents, most studies reporting good correlations against
the clamp-derived index (r=0.91, p<0.1 (Gungor et al. 2004)) or the minimal model FSIGTT in obese
children (r=0.89, p<0.01 (Conwell et al. 2004)). In contrast, other studies have indicated a weaker
correlation between HOMA-IR and the index obtained from the minimal model FSIGTT (r=-0.4,
p<0.001) in twins and children born small-for-gestational age (Cutfield et al. 2003). Compared to

60
Theoretical framework

other indices of IR derived from fasting plasma insulin and glucose (such as QUICKI or FGIR),
HOMA-IR in obese children and adolescents is more accurate (has higher sensitivity and specificity),
and therefore is more appropriate when other measures of insulin sensitivity are not possible (Keskin
et al. 2005).

Normal IR is indicated by a value of 1 for HOMA-IR (Wallace et al. 2004). For diabetic adults,
HOMA-IR ranges from 2.61 to 2.89 (Matthews et al. 1985). Among children and adolescents, the
cut-off value for detecting metabolic syndrome with high sensitivity and specificity has not been
conclusively established. Different mean values for HOMA-IR have been reported, ranging from 2.3
in normal weight adolescents (Lee et al. 2006), to 2.5 in pre-pubertal overweight children (Madeira et
al. 2008), and 3.16 (Keskin et al. 2005) or 4.93 (Lee et al. 2006) in obese adolescents.

HOMA-IR has been deemed appropriate to quantify IR as a one-off measure and to assess
longitudinal changes in IR (e.g., in order to examine the natural history of diabetes) (Wallace et al.
2004). HOMA-IR is chosen in the current study over other indices based on fasting plasma insulin
and glucose, as it has been proven to be accurate in children and adolescents (Keskin et al. 2005).

2.4 Evidence of early origins of childhood obesity and insulin resistance

The focus of interest for this project was on the influence of three modifiable factors known to alter
the intrauterine milieu (maternal pre-pregnancy obesity, glucose intolerance during pregnancy across
the entire spectrum, and gestational weight gain), separately and concomitantly, on the programming
of obesity and insulin resistance in pre-pubertal children. The following three sections include a
summary and critique of published epidemiological studies addressing the influence that each of
these intrauterine factors have on obesity and insulin resistance in childhood, identifying limitations
and gaps in the existing literature. A greater emphasis is given to longitudinal studies, known to
provide the strongest evidence for such associations.

2.4.1 Intrauterine programming of obesity and insulin resistance by maternal pre-


pregnancy obesity

Epidemiological studies increasingly suggest a link between maternal pre-pregnancy obesity and
metabolic perturbations in the offspring. Maternal obesity at the time of pregnancy has been
suggested as an independent risk factor for obesity in the next generation (Nelson et al. 2010). A
61
Theoretical framework

summary of the longitudinal studies examining the relationship between maternal obesity status at
the time of pregnancy and child obesity, adiposity, fat pattern and insulin resistance beyond infancy
in low risk populations is presented in Table 6 and discussed below.

Seven longitudinal studies comprising contemporary samples of children (born between 1982 and
2000) of various ages, from around the world provide consistent evidence for a positive association
between maternal overweight status at the time of pregnancy and the development of childhood
obesity, as measured by BMI (Boerschmann et al. 2010; Lawlor et al. 2007b; Li et al. 2005; Pirkola et
al. 2010; Reilly et al. 2005; Whitaker 2004). Of relevance to the age group considered in the present
study, OR for maternal weight and child obesity ranged from 4 (95% CI 1.23, 13.2) in 6-11 year-old
children (Catalano et al. 2009a) to 7 (95% CI 1.8, 27.7) in 11 year-olds (Boerschmann et al. 2010).
Based on data from studies that included at least two age groups, it appears that the association
might increase in peri-pubertal years compared with early childhood (Boerschmann et al. 2010; Li et
al. 2005), but it is difficult to draw conclusions from only two studies. Variations may also be due to
differences in exposure and outcome definitions (critiqued below) or to differences in populations’
characteristics (e.g., in the study by Boerschmann et al. (2010) children were born to mothers with
GD). A further increased risk of obesity was observed in children of obese mothers who also
developed GD during the index pregnancy compared to those whose mothers were obese but
maintained normal glucose tolerance during pregnancy (Pirkola et al. 2010).

The few prospective studies looking beyond weight relative to height of child in relation to maternal
overweight status prior to pregnancy similarly reported a positive association for child percentage
body fat in pre-pubertal years (Blair et al. 2007; Catalano et al. 2009a) and for waist-to-height ratio in
adolescents (Pirkola et al. 2010). The latter association was found to be amplified by GD (Pirkola et
al. 2010).

Two sets of evidence are in favour of a transmitted effect of maternal obesity through the intrauterine
milieu in addition to an inherited (genetic) one on the increased risk of overweight status in the
offspring. First, maternal pre-pregnancy BMI appeared as a better predictor of child body size than
paternal BMI at the time of conception (Lawlor et al. 2007b; Reilly et al. 2005). Furthermore, based
on data from the Avon Longitudinal Study of Parents and Children (ALSPAC), the risk of childhood
obesity at 7 years increased exponentially when both parents were obese at conception (OR=10.4
(95% CI 5.11-21.3) when both parents were obese; OR=4.25 (95% CI 2.86-6.32) when mother was
obese; OR=2.54 (95% CI 1.72-3.75) when father was obese) (Reilly et al. 2005). Second, evidence
for the role of the intrauterine milieu in programming obesity risk in the child was revealed by a study

62
Theoretical framework

of siblings discordant for in utero exposure to maternal obesity, which showed a reduction in the
prevalence of obesity by 52% in offspring born after bariatric surgery, reaching levels similar to the
general population (Kral et al. 2006); in the latter study, it is possible that women also changed their
diet and lifestyle, with potential positive effects on child body size.

The evidence is scarce and inconsistent with respect to the effect of maternal obesity on insulin
resistance in the offspring. One family-based study conducted in young adults found a significantly
higher degree of insulin resistance in offspring of mothers who were obese at the time of pregnancy
compared to those of normal weight mothers (Mingrone et al. 2008). In a recent cohort study,
Catalano et al. (2009b) documented for the first time that insulin resistance in the offspring exposed
to maternal pre-pregnancy obesity is apparent at birth and reported a positive relationship (r=0.27,
p=0.002) between maternal pre-pregnancy BMI and neonatal insulin resistance as measured by
HOMA-IR, based on umbilical cord blood samples (mothers had fasted to prepare for caesarean
section, thus ensuring steady-state conditions). These two studies are not included in Table 6 as
they were not studies of children beyond infancy. Of greater relevance to the present study, a
longitudinal study in 8-11 year old children of women with GD did not show an association between
maternal pre-pregnancy BMI and child insulin resistance (Boerschmann et al. 2010).

63
Table 6. Summary of longitudinal studies examining the influence of maternal pre-pregnancy overweight / obesity on child obesity and insulin resistance

Reference Setting Population Study Sample Age at Measurement of Outcome Covariates, Effect of maternal pre-pregnancy weight status
design size in outcome maternal pre- measurement confounders
analyses (years) pregnancy BMI Unadjusted Adjusted

Blair et al. 2007 New Zealand Caucasian Prospective 591 7 self-reported pre- %BF (BIA) maternal age, Overweight mothers: Overweight mothers:
pregnancy weight smoking, +3% (p<0.01) +3.7% (p<0.01)
and height (after hypertension in Obese mothers: Obese mothers:
Auckland delivery) pregnancy, SES, BW,
Birthweight +4% (p<0.01) +4% (p<0.01)
breastfeeding, rapid
Collaborative weight gain in
Study Nwt/Owt/Ob infancy, sedentary
time
Boerschmann Germany Caucasian Prospective 89 8 early pregnancy BMI ≥ 90th centile birth size, smoking, Overweight mothers: Not reported for
et al. 2010 (>98%) 74 11 weight and height (German therapy of GD 8 y: OR=7.4 (0.8, 68.2) overweight mothers
(measured by reference) 11 y: OR=3.1 (0.5, 19.5)
- all women physicians)
with GD Obese mothers: Obese mothers:
8 y: OR=16 (1.8, 140.9) 8 y: OR=5.1 (1.4, 18.6)
Nwt/Owt/Ob 11 y: OR=9.3 (1.8, 48.7) 11 y: OR=7.0 (1.8, 27.7)
HOMA-IR
No association pre- No association pre-
pregnancy BMI and pregnancy BMI and
HOMA-IR in offspring of HOMA-IR in offspring of
GD mothers GD mothers
Catalano et al. USA Caucasian, Prospective 89 6-11 self-reported pre- Weight centile GD, GWG, family Obese mothers: Obese mothers:
2009a African pregnancy weight (2000 CDC) history of diabetes, OR=3.75 (1.39, 10.1) OR=4.03 (1.23, 13.2)
American, (at delivery), (highest tertile) maternal age, parity,
Hispanic, measured height 1-h glucose
Asian screening, smoking,
paternal weight and
- elective Nwt/Owt vs. Ob height
caesarean
section 63 % BF (DXA) + %BF at birth Obese mothers: Obese mothers:
(highest tertile) OR=5.45 (1.62, 18.4) OR=7.75 (1.51, 37.4)
64
Reference Setting Population Study Sample Age at Measurement of Outcome Covariates, Effect of maternal pre-pregnancy weight status
design size in outcome maternal pre- measurement confounders
analyses (years) pregnancy BMI Unadjusted Adjusted

Lawlor et al. Australia Not Prospective 3,795 14 self-reported pre- Internal age- and Paternal BMI at the Not reported +1SD in maternal pre-
2007b specified pregnancy weight sex- specific SD time of pregnancy, pregnancy BMI ->
and height (at first for BMI income, parental +0.36SD in child BMI
Mater antenatal clinic) education, maternal (0.32, 0.40)
University age, smoking
Study of
+ size at birth +1SD in maternal pre-
Pregnancy internal age-
pregnancy BMI ->
and its specific SD for
Outcomes BMI +0.362SD in child BMI
(0.323, 0.402)
Li et al. 2005 USA White, Retrospective 2,636 2-16 self-reported pre- BMI ≥ 95th centile GWG, gestational Not reported Owt mothers:
Black, pregnancy weight (CDC) age, birth order, 2-6 y: OR=2.4 (1.4, 4.5)
Hispanic and height (after race/ethnicity, parity, 7-10 y: OR=2.2 (1.2, 4.1)
National delivery) age, smoking, alcohol 11-16y: OR=3.3 (1.7, 6.2)
Longitudinal use, education, family
Survey of Ob mothers:
income, BW,
Youth Uwt/Nwt/Owt/Ob breastfeeding 2-6 y: OR=2.8 (1.3, 6)
7-10y: OR=5.7 (2.9, 11.7)
11-16y: OR=4.3 (1.7, 10.8)
Pirkola et al. Finland Caucasian Prospective 4,186 16 Nwt vs. Owt/Ob BMI ≥ 85th centile GD (selective Relative to Nwt women Relative to Nwt women
2010 Northern (IOTF) screening - high risk) without risk factors for GD: without risk factors for GD:
Finland Birth Pre-pregnancy Owt/Ob women with Owt/Ob women with
Cohort of smoking, paternal normal OGTT normal OGTT
1986 BMI, size at birth OR=2.92 (2.03, 4.22) OR=2.56 (1.69, 3.88)
Owt/Ob women with GD Owt/Ob women with GD
OR=5.03 (2.54, 9.97) OR=4.05 (1.90, 9.16)
Waist-to-height Relative to Nwt women Relative to Nwt women
ratio > 0.5 without risk factors for GD: without risk factors for GD:
Owt/Ob women with Owt/Ob women with
normal OGTT normal OGTT
OR=2.97 (1.95, 4.51) OR=2.60 (1.62, 4.17)
Owt/Ob women with GD Owt/Ob women with GD
OR=4.25 (1.97, 9.16) OR=3.82 (1.66, 8.82)
65
Reference Setting Population Study Sample Age at Measurement of Outcome Covariates, Effect of maternal pre-pregnancy weight status
design size in outcome maternal pre- measurement confounders
analyses (years) pregnancy BMI Unadjusted Adjusted

Reilly et al. UK White, Prospective 7,758 7 self-reported BMI ≥ 95th centile SES, education, BW, OR=4.66 (3.28, 6.64) OR=4.25 (2.86, 6.32)
2005 Avon non-white weight and height (1990 British) gender, parity,
Longitudinal (in early smoking, gestational
Study of pregnancy) age, breastfeeding,
Parents and ethnicity, maternal
Children Nwt/Owt vs. Ob age

Whitaker 2004 USA White, Retrospective 8,494 2-5 weight and height BMI ≥ 95th centile BW, gestational age, Prevalence of child obesity: Owt mothers:
black, + data measured in the (CDC) gender, race / Nwt mothers 2 y: OR=1.42 (1.13, 1.79)
Hispanic, linkage first trimester of ethnicity, parity, age, 2 y: 6.4% 4 y: OR=1.75 (1.4, 2.18)
Special other pregnancy smoking, education, 4 y: 9%
Supplemental net GWG (self- Owt mothers Ob mothers:
Nutrition Low-
income report), marital status 2 y: 9% 2 y: OR=2.28 (1.84, 2.83)
Program for Uwt/Nwt/Owt/Ob/ 4 y: 14% 4 y: OR=3.07 (2.48, 3.79)
Women, families morbidly Ob Ob mothers
Infants, and 2 y: 13.9% No significant interactions
Children 4 y: 22.8% with covariates
%BF – percentage body fat, BIA – bioelectrical impedance analysis, BW – birth weight, CDC – Centers for Disease Control and Prevention, DXA – dual energy X-ray absorptiometry, GD – gestational diabetes, GWG – gestational weight gain, HOMA-IR – homeostasis model
assessment of insulin resistance, IOTF – International Obesity Task Force, Nwt – normal weight, Ob – obese, OGTT – oral glucose tolerance test, Owt – overweight, SES – socioeconomic status, SD – standard deviation, Uwt - underweight
66
Theoretical framework

Critique of longitudinal studies investigating the influence of maternal pre-pregnancy


overweight status on child obesity and insulin resistance

The summarised studies concerning the association between maternal pre-pregnancy BMI and child
obesity and insulin resistance have a number of limitations, including their exposure and outcome
assessment, sampling frames and representativeness. A critique of these studies and identification
of their limitations follow.

Exposure definition

All but one study defined maternal overweight status as a categorical variable, using BMI cut-offs
recommended by WHO (World Health Organization 1995). Although data was available on BMI as a
continuous variable in each of these studies, it was not used in analytical models. Hence the effect
size of increasing pre-pregnancy BMI by one unit on child obesity risk could not be reported. The
exception was Lawlor et al. (2007b) who reported a third of an SD increase in child BMI z-score
(internally derived) for each SD increase in maternal BMI.

No studies directly measured maternal weight prior to pregnancy. Most studies based their maternal
pre-pregnancy BMI calculation on pre-pregnancy weight self-reported during pregnancy (Lawlor et
al. 2007b; Reilly et al. 2005) or at/after the time of delivery (Blair et al. 2007; Catalano et al. 2009a; Li
et al. 2005). Three of these studies also used self-reported height (Blair et al. 2007; Li et al. 2005;
Reilly et al. 2005). This has potential implications on the interpretation of the results due to
information bias. Given women’s tendency to under-report their weight, particularly in overweight
and obese women, with subsequent potential underestimation of pre-pregnancy obesity, it is
possible that the associations reported underestimate the true association. Boerschmann et al.
(2010) and Whitaker (2004) used direct measurements of weight and height in the first trimester of
gestation as proxy measures for maternal pre-pregnancy weight and height; there may be some
variation from weight at conception, but for most women this is likely to be small.

Outcome definition

Different definitions of childhood overweight and obesity were employed by these studies. All but
one (Lawlor et al. 2007b) used externally derived BMI centiles for a specific population, two based on
American reference data (Li et al. 2005; Whitaker 2004), one on British (Pirkola et al. 2010; Reilly et
al. 2005), one on German (Boerschmann et al. 2010) and only one on IOTF centiles (Pirkola et al.
2010). Thus, the reported adjusted ORs for childhood overweight and obesity are not directly

67
Theoretical framework

comparable between studies, ranging from 2.28 to 7 in children aged 2-16 years. The ORs appear
to increase with age; for instance, Li et al. (2005) reported OR of obesity in children of obese
mothers ranging from 2.8 in 2-6 year olds, to 5.7 in 7-10 year olds and 4.3 in 11-16 year olds.

Only one study reported how the spectrum of maternal pre-pregnancy BMI influences child BMI
(internally derived age- and sex-specific BMI z-score) (Lawlor et al. 2007b). This is important as it
removes the influence of arbitrary cut-offs between studies and allows the investigation of smaller
unit shifts in BMI.

Studies examining only child BMI as an outcome are limited to drawing conclusions on weight
relative to height, and cannot comment on the amount of adiposity or fat distribution. Only three
studies looked beyond BMI in examining intrauterine programming of childhood obesity by maternal
pre-pregnancy overweight status. Of them, two studies focused on child %BF, estimated by BIA
(Blair et al. 2007) or DXA (Catalano et al. 2009a). While both studies showed a positive association
between maternal pre-pregnancy overweight status and child %BF, neither provided information on
how pre-pregnancy BMI across the spectrum can influence child %BF.

The only longitudinal study assessing offspring fat pattern in relation to maternal overweight during
pregnancy was conducted in the Finnish cohort of adolescents and defined abdominal obesity as
waist-to-height ratio > 0.5 (Pirkola et al. 2010). By using a dichotomous variable, small shifts in
waist-to-height ratio that might occur as a result of intrauterine exposure to maternal overweight, but
without shifting the ratio above 0.5, could be missed.

Confounders

Each study adjusted for a different set of covariates, including various indicators of SES, maternal
age, parity, smoking, hypertension, paternal weight, birth weight or breastfeeding (the latter two
factors are more likely to be on the causal pathway than confounding factors). Glucose tolerance
during pregnancy, a factor that is thought to be associated both with maternal pre-pregnancy obesity
and child obesity, was taken into account only by Catalano et al. (2009a) and Pirkola et al. (2010).
Similarly, adjustment for weight gain during pregnancy was done in only three studies (Catalano et
al. 2009a; Li et al. 2005; Whitaker 2004). None of the studies included all the potential confounders I
have identified, namely glucose tolerance during pregnancy, gestational weight gain, maternal age,
parity, smoking, pregnancy-induced hypertension, and education at the time of pregnancy (see
Section 3.4.3 for confounder justification). Poor confounding adjustment could result in inflated ORs
if an underlying factor is driving the association.

68
Theoretical framework

Sample representativeness and external validity

Although some of the samples on which these analyses were based comprised over 2,000
participants from various ethnic backgrounds (Li et al. 2005; Pirkola et al. 2010; Reilly et al. 2005;
Whitaker 2004), others were relatively small, with under 100 participants (Boerschmann et al. 2010;
Catalano et al. 2009a). However, the effect size in pre-pubertal children did not seem to be
influenced by the sample size (e.g., OR=4.25 in ALSPAC (Reilly et al. 2005), OR=4 in the small-
scale study of Catalano et al.(2009a)).

Studies were not always representative of the general population. For instance, the cohort of
German mother-child pairs reported by Boerschmann et al. (2010) included only women who
developed GD during the index pregnancy. Thus those findings are unlikely to be generalisable to
mothers who maintain normal glucose tolerance during pregnancy, which might explain the higher
OR for child obesity identified in this study compared to other studies with children of similar age.
Another example of non-random sample is given by the Special Supplementation Nutrition Program
for Women, Infants and Children, which included only low income families (Whitaker 2004). With
regard to socioeconomic status, some of the studies did not report whether their sample comprised
the full SES range (Boerschmann et al. 2010; Catalano et al. 2009a).

Another important factor that might affect representativeness in longitudinal studies with long-term
follow-up like the ones described in this section is participation rate, particularly if women or children
with certain characteristics (e.g., lower SES) are less likely to participate at the study outset or in
follow-up. Participation at the time of outcome measurement (number followed-up divided by
number of participants at baseline) ranged between 38.4% at 8 years and 31.9% at 11 years in the
German study with offspring of GD mothers (Boerschmann et al. 2010), to 53% at 14 years in Mater
University Study of Pregnancy and Its Outcomes (Lawlor et al. 2007b), to 55.5% at 7 years in
ALSPAC (Reilly et al. 2005), 67.9% in Auckland Birthweight Collaborative Study (Blair et al. 2007)
and 80% in the Northern Finland Birth Cohort (Pirkola et al. 2010). High non-participation can lead
to response bias, which may affect the inferences drawn from the study in question. Of note,
comparison of baseline characteristics between participants and non-participants were reported only
in two of the studies reviewed in this section. In the study by Lawlor et al. (2007b) participants at
follow-up had lower birth weight, were less likely to be from low-income families, their parents were
more educated, their mothers were older, but parental BMI at the time of pregnancy was similar to
the original cohort. In contrast, in the study by Pirkola et al. (2010) parental BMI at pregnancy was

69
Theoretical framework

lower in the subgroup taking part in the follow-up compared to the initial cohort. Without full
disclosure of attrition, it is challenging to evaluate the validity of the findings.

Conclusion

Whilst a number of longitudinal studies have found positive associations between maternal pre-
pregnancy overweight status and child obesity, firm conclusions on these associations cannot be
drawn due to the limitations mentioned above. Thus, this critical appraisal demonstrates a need for
a systematic replication with extension of exposures, outcomes and control for potential confounders
in a study comprising a representative, contemporary Caucasian population of children. With the
exception of one study that examined how maternal pre-pregnancy BMI across the entire spectrum
influences child BMI (using internally derived SD scores), progressive effects of discrete changes in
maternal body size prior to pregnancy on child body size, body composition, fat pattern and insulin
resistance have not been investigated. These are the gaps identified in the literature that this thesis
seeks to fill. Removing arbitrary cut-offs (using a continuous exposure and outcome measures) is
important because it can demonstrate that even incremental increases in maternal body size,
regardless of meeting criteria for overweight, might increase risk in the child. The influence of
maternal body size on child BMI is certainly important, but the potential effects on more specific
measures of adiposity (percentage body fat or fat distribution) and insulin resistance are of even
greater relevance for the future metabolic health of the child, as they are better predictors of
cardiovascular risk than BMI (Goran and Gower 1999; Steinberger et al. 2001). Therefore, these
potential associations, sparingly assessed to date, warrant examination, as they might indicate
targets for obesity and insulin resistance prevention. Moreover, the incomplete adjustment for
confounders in previous studies may have resulted in overestimation of the true associations.

2.4.2 Intrauterine programming of obesity and insulin resistance by maternal glucose


intolerance during pregnancy

A growing body of epidemiological literature has been devoted to the description of long-term
metabolic alterations associated with prenatal exposure to maternal glucose intolerance during
pregnancy. A summary of longitudinal studies investigating intrauterine programming of obesity,
body composition, fat pattern and insulin resistance in children (beyond neonatal period) by maternal
glucose intolerance during pregnancy in non-high risk populations worldwide is presented in Table 7
and discussed below.

70
Theoretical framework

Intrauterine programming of obesity and insulin resistance by gestational diabetes

The evidence of an association between maternal GD and long-term risk of obesity in the child,
although relatively abundant, is inconsistent. Early reports came from longitudinal prospective
studies in Pima Indians, the population with the highest known incidence and prevalence of type 2
diabetes in the world, which showed that in utero exposure to diabetes results in increased weight for
height during childhood, independent of size at birth (Pettitt et al. 1983; 1987). It could be argued
that the increased risk of obesity in this population is primarily due to genetic factors and not to the
diabetic intrauterine environment. In support of the latter are the findings of a study conducted in
siblings of Pima Indians (with two or more siblings per family), whose mothers developed type 2
diabetes between two pregnancies (meaning that children born before the diagnosis were not
exposed to diabetes in utero, while subsequent siblings were) (Dabelea et al. 2000). Participants
aged 9 years and older, who were exposed to diabetes in utero had higher BMI (+2.6 kg/m2,
p=0.003) compared to their unexposed siblings (who presumably had a similar risk of inheriting the
same genetic predisposition), but there was no difference in BMI between siblings born before and
after their father’s diagnosis of type 2 diabetes. Moreover, data from the same study showed that
intrauterine exposure to type 2 diabetes trebled siblings’ risk of developing type 2 diabetes
themselves (OR=3.7, 95% CI 1.3, 11.3) compared to those not exposed (Dabelea et al. 2000).

The findings from Pima Indian studies stimulated initiation of studies in populations with lower
prevalence of GD, which have generated conflicting results. Studies report increased weight relative
to height (Catalano et al. 2009a; Crume et al. 2011a; Egeland and Meltzer 2010; Krishnaveni et al.
2010; Silverman et al. 1991; Silverman et al. 1998), higher BMI growth velocity (Crume et al. 2011b),
or greater prevalence of obesity (Boerschmann et al. 2010) in pre-pubertal and pubertal children
exposed to maternal GD compared to those born to mothers who maintained normal glucose
tolerance during pregnancy. It was also observed that BMI in adolescents of diabetic mothers (either
pre-gestational diabetes or GD) correlated with fetal insulin secretion, as measured by the amniotic
fluid insulin (Silverman et al. 1998). In contrast, other studies of populations without a high risk for
diabetes did not find a robust association between GD and global obesity in children aged 5-16
years, after adjustment for potential confounders, including maternal body size (Gillman et al. 2003;
Jeffery et al. 2006; Pirkola et al. 2010; Whitaker et al. 1998). One of these studies concluded that
the GD influence on child obesity risk in peri-pubertal years was partially explained by the effect on
birth weight (Gillman et al. 2003). The findings from Whitaker et al. (1998) and Pirkola et al. (2010)
may have underestimated the true association between GD and child obesity as women diagnosed

71
Theoretical framework

with GD were given treatment (including dietary advice, glucose monitoring and insulin if needed);
this treatment may have lessened the severity of fetal exposure to maternal glucose intolerance.
The findings of Jeffery et al. (2006) need to be interpreted with caution given that maternal third
trimester fasting glucose levels were documented only in a small subgroup of women (n=26) with
random glucose level >6.5 mmol/l.

Similarly, there is little and inconsistent evidence on the relationship between glucose intolerance
during pregnancy and child body composition. After confounder adjustment, exposure to GD was
positively associated with offspring fat mass at birth (Catalano et al. 2003b), subscapular and triceps
skinfold thickness at 3 years (Wright et al. 2009) and at 9.5 years (Krishnaveni et al. 2010), but not
with percentage body fat estimated by DXA at 6-11 years (Catalano et al. 2009a).

Childhood central adiposity in relation to intrauterine exposure to maternal glucose intolerance was
explored in four longitudinal studies (Catalano et al. 2009a; Crume et al. 2011a; Egeland and Meltzer
2010; Pirkola et al. 2010), again with contradictory results. The Exploring Perinatal Outcomes
among Children (EPOCH) study showed increased waist circumference (by 3.2 cm, p=0.05 in fully
adjusted model) and marginally greater subcutaneous adipose tissue in the abdominal region
quantified by MRI (by 26 cm3, p=0.08 in fully adjusted model) in 6-13 year-old offspring exposed to
diabetes in utero compared to those not exposed; however, the association initially observed
between maternal diabetes and the ratio of subscapular to triceps skinfold thickness in the offspring
was attenuated to null after adjustment for maternal pre-pregnancy BMI (Crume et al. 2011a). In
another study, female offspring exposed to GD or impaired glucose tolerance while in utero had
significantly greater waist circumference at age 15 years compared to those of normal glucose
tolerant mothers (mean difference 5.9 cm after covariate adjustment) (Egeland and Meltzer 2010).
The findings of this study are limited by the relatively small sample size (90 mothers with GD or
impaired glucose tolerance, and 99 mothers with normal glucose tolerance during pregnancy), low
participation rates in both groups (51.6% and 24.4%, respectively) (Egeland and Meltzer 2010) and
exclusion of male children. The other two studies did not identify an association between maternal
GD and offspring waist circumference at 6-11 years (Catalano et al. 2009a) or waist-to-height ratio in
16 year-old offspring of normal weight mothers (Pirkola et al. 2010). Variations may be due to
differences in the age at which offspring were assessed, as well as differences in exposure and
outcome definitions, to be explored later in this section.

Exposure to a diabetic intrauterine environment has also been linked to childhood glucose-insulin
homeostasis, but the evidence has been inconclusive. A greater prevalence of impaired glucose

72
Theoretical framework

tolerance was identified in adolescents exposed to maternal diabetes in utero (with no differential
effect between GD, type 1 or type 2 diabetes during pregnancy) compared to those not exposed and
it correlated with the higher level of amniotic fluid insulin, as a marker of excessive fetal insulin
secretion (Silverman et al. 1995). Inconsistencies also exist in relation to the risk of insulin
resistance in children whose mothers developed glucose intolerance during pregnancy. Three
recent studies identified increased HOMA-IR in peri-pubertal children (Boerschmann et al. 2010;
Egeland and Meltzer 2010; Krishnaveni et al. 2010), while two others found no increase in younger
children (Catalano et al. 2009a; Jeffery et al. 2006) of mothers with GD. Of note, adjustments for
potential confounders were not performed in any of these studies with HOMA-IR as primary
outcome. Moreover, only the study by Krishnaveni et al. (2010) considered the potential mediating
effect of current BMI, which attenuated to null the association initially identified. It is possible that
puberty was the driving factor for the increased HOMA-IR reported by the former three studies.

A longitudinal study linking maternal GD and offspring size at birth with later development of
metabolic syndrome showed that children born large-for-gestational-age whose mothers had GD
were at increased risk of developing insulin resistance by 11 years as estimated by the fasting
glucose to insulin ratio <7, but not by HOMA-IR (Boney et al. 2005). Moreover, these children were
reported to have an increased risk of metabolic syndrome based on a definition that, importantly, did
not consider waist circumference, despite this conventionally being a core component of the
metabolic syndrome (Zimmet et al. 2007).

Intrauterine programming of obesity and insulin resistance by milder degrees of glucose


intolerance during pregnancy

There is some epidemiological evidence suggesting that milder elevations of maternal glycaemia
during pregnancy (not diagnostic for GD) may also be related to an increased risk of obesity in the
offspring. Initial studies conducted in American Indians (mainly Pima Indians) showed positive
associations between maternal 2-hour glucose level during OGTT and offspring age-, sex- and
height-specific weight at 5-14 years (Pettitt et al. 1991), and between maternal third trimester
glucose level and offspring BMI and waist circumference in subjects aged 10-14 years (but not other
age groups), whose mothers had an abnormal OGCT, but a normal OGTT (Franks et al. 2006).

Few studies have addressed intrauterine programming of child obesity by maternal chronic
hyperglycaemia during pregnancy in non-high risk populations. One data linkage study reported an

73
Theoretical framework

increased risk of overweight and obesity in children who were not macrosomic at birth and whose
mothers were in the highest quartile of plasma glucose levels at OGCT (Hillier et al. 2007). An
additional argument for a potential link between various degrees of maternal glucose intolerance
during pregnancy and offspring obesity emerging from this study was the finding of lower obesity
rates in children whose mothers received treatment for GD compared to those whose did not (Hillier
et al. 2007). Importantly, no adjustment for maternal pre-pregnancy body size was performed in
these analyses, which might confound the results.

Data from the ALSPAC study showed some evidence of an association between higher maternal
glucose levels during pregnancy and several obesity-related measures in 9-11 year-old children
(Lawlor et al. 2010). Maternal glycosuria (considered as a measure of a lower degree of
hyperglycaemia than that characterising GD) was associated with 53% increased risk of child global
overweight or obesity, and 39% increased risk of child central adiposity; these relationships were
attenuated (to 35% and 31%, respectively) by adjustment for maternal pre-pregnancy BMI, but
remained statistically significant (Lawlor et al. 2010).

Similar to some extent, in a small, retrospective study, Chandler-Laney (2011) showed that maternal
plasma glucose level at OGCT as a continuous variable correlated positively with child BMI, fat
mass, fat free mass and waist circumference at the age of 5-10 years.

Only one study reported on the association between various degrees of glucose intolerance during
pregnancy and offspring insulin resistance during puberty (Egeland and Meltzer 2010), finding
significantly greater mean HOMA-IR in girls of insulin-treated GD mothers (2.36) compared to both
normal (1.58) and impaired glucose tolerance during pregnancy (1.56); no significant difference in
mean HOMA-IR was found between daughters of mothers with normal and impaired glucose
tolerance during pregnancy (Egeland and Meltzer 2010). The potential confounding effect of
maternal pre-pregnancy BMI was not examined in this study.

74
Table 7. Summary of longitudinal studies examining the influence of maternal glucose tolerance status during pregnancy on child obesity and insulin resistance

Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of maternal glucose tolerance in pregnancy
size in outcome glucose measurement confounders
analyses (years) intolerance in
Unadjusted Adjusted
pregnancy

Boerschmann Germany Caucasian Prospective, Children German Diabetes BMI ≥ 90th centile BMI in early Prevalence of obesity Exacerbated by BMI in
et al. 2010 German (>98%) combining two of GD Association (German pregnancy, smoking, significantly higher in early pregnancy in OGD
GD studies: mothers: NGT reference) GD therapy, birth size OGD group
offspring 89 8 8 y: 20.2% 8 y: 36.4%
GD
study 74 11 11 y: 31.1% 11 y: 45.8%
T1D relative to OT1D
&
8 y: 11% (p=0.03)
Baby-Diab 11 y: 15.8% (p=0.01)
study
and OND
8 y: 10.3% (p=0.02)
11 y: 15.5% (p=0.005)
751 Significantly higher Not reported
HOMA-IR
(74 OGD HOMA-IR in OGD
425 OT1D compared to OT1D
252 OND) (p=0.03) and ONDM
(p=0.04) (no estimates
presented – graph only)
Boney et al. USA White Prospective 179 6, 7, 9, National Diabetes BMI>85th centile Maternal pre- No difference among the
2005 (>94%) with a 11 Data Group pregnancy BMI, 4 groups with regard to
retrospective NGT GWG, SES, birth size the prevalence of obesity
component GD or HOMA-IR
88 11 4 groups: HOMA-IR At 11 y, trend towards a Interaction of LGA and
LGA / GD FGIR <7 higher incidence of GD (but no independent
AGA / GD FGIR<7 in LGA / GD effect) on FGIR
LGA / NGT group (p=0.08) OR=10.4 (1.5, 74.4)
AGA / NGT
75
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of maternal glucose tolerance in pregnancy
size in outcome glucose measurement confounders
analyses (years) intolerance in
pregnancy Unadjusted Adjusted

Catalano et al. USA Caucasian, Prospective 89 6-11 National Diabetes BMI z-score Pre-pregnancy BMI, Child BMI z-score in Not presented, but
2009a African Data Group (2000 CDC) family history of NGT: 0.31 concluded that pre-
American, - NGT diabetes, smoking, GD: 0.90 (p=0.03) pregnancy obesity was a
Hispanic, education, age, better predictor for child
Asian - GD parity, GWG obesity
- elective waist No assoc with waist
caesarean circumference, circumference (p=0.36)
section central to peri- or ratio SFT (p=0.13)
pheral SFT ratio
63
%BF(DXA) + %BF at birth Child %BF in
NGT:27.8
GD:31 (p=0.14)
47
HOMA-IR HOMA-IR in
NGT: 1.89
GD: 2.55 (p=0.08)
Chandler- USA African Retrospective 27 5-10 1-h glucose level BMI centile (CDC) Ethnicity, resting r=0.445 (p<0.05) Not reported
Laney et al. American at OGCT (from energy expenditure,
2011 (78%), medical records) Waist physical activity, diet r=0.469 (p<0.05) Not reported
European circumference
American
FM (DXA) +FFM r=0.418 (p<0.05) r=0.443 (p<0.05)

FFM (DXA) Height, sex r=0.122 (p>0.05) r=0.371 (p=0.07)


76
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of maternal glucose tolerance in pregnancy
size in outcome glucose measurement confounders
analyses (years) intolerance in
pregnancy Unadjusted Adjusted

Crume et al. USA Multiethnic Retrospective 461 6-13 National Diabetes BMI BW, gestational age, 20.2 vs. 18.9 (p=0.02) 19.9 vs. 19.1 (p=0.2)
2011a Exploring Data Group age, sex, race,
Perinatal Tanner stage,
waist circumfer. maternal age, SES, 69.6 vs. 65.4 (p=0.004) 68.7 vs. 65.5 (p=0.05)
Outcomes GD
among smoking,pre-
Children NGT VAT (MRI) pregnancy BMI 27.8 vs. 24.2 (p=0.1) 26.4 vs. 25.3 (p=0.7)
SAT (MRI) 156.4 vs. 121.7 (p=0.01) 147.1 vs. 121.0 (p=0.08)

Subscapular to 0.83 vs. 0.76 (p=0.01) 0.82 vs. 0.77 (p=0.2)


triceps SFT ratio
Egeland and Canada Caucasian, Prospective 189 15 National Diabetes BMI centile (2000 BW z-score, ethnicity, 71.4 vs. 60.2 (p<0.01) Attenuated (p<0.10)
Meltzer 2010 non- Data Group CDC) physical activity, diet,
Caucasian mother’s 15 year
- NGT %BF (BIA) weight gain, mother’s 31.5 vs. 28.7 (p<0.01) Annulated (p<0.10)
- IGT (1 abnormal current BMI, %BF,
value at OGTT) waist circumference
-GD (≥2 abnormal waist circumfer. 81.2 vs. 73.5 (p<0.001) Attenuated (p<0.05)
values at OGTT)

GD/IGT=cases HOMA-IR 1.94 vs. 1.58 (p<0.05) Not presented


NGT=controls
Gillman et al. USA Nationwide, Linkage 14,881 9-14 Self-reported GD BMI centile BW, breastfeeding, GD vs. NGT: GD vs. NGT:
2003 Growing but homo- status (2000 CDC) SES, diet, physical
Up Today geneous in >85th activity, age, Tanner OR>85th =1.2 (0.9, 1.5) OR>85th =1.0 (0.7, 1.3)
Study & terms of GD >95th stage, maternal OR>95th =1.4 (1.1, 2.0) OR>95th =1.2 (0.8, 1.7)
Nurses’ race / NGT current BMI
Health ethnicity
Study II
77
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of maternal glucose tolerance in pregnancy
size in outcome glucose measurement confounders
analyses (years) intolerance in
pregnancy Unadjusted Adjusted

Hillier et al. USA Multiethnic Data-linkage 9,439 5-7 Glucose levels at Weight (age-,sex- GWG (self-reported), Not reported Highest vs. lowest quartile
2007 (electronic OGCT (universal specific centiles) age, parity, ethnicity, of glucose at OGCT:
medical screening)- >85th BW, sex OR>85th =1.22 (1.03, 1.45)
records) quartiles >95th OR>95th =1.28 (1.02, 1.60)

Relative to -OGCT:
National Diabetes
Data Group (likely +OGCT, -OGTT
treated if GD) / no significant change in
Carpenter and OR>85th, OR>95th
Coustan (likely +OGCT, +1 abnormal value
untreated if GD) at OGTT
+OGCT, -OGTT OR>85th =1.37 (1.01, 1.84)
+OGCT, +1 OR>95th =1.30 (0.89, 1.90)
abnormal value at +OGCT, + ≥2 abnormal
OGTT values at OGTT (GD -
+OGCT, + ≥2 untreated)
abnormal values OR>85th =1.89 (1.30, 2.76)
at OGTT (GD – OR>95th =1.82 (1.15, 2.88)
untreated)
+OGCT, + ≥2 abnormal
+OGCT, + ≥2 values at OGTT (GD -
abnormal values treated)
at OGTT (GD – OR>85th =1.29 (0.85,1.97)
treated) OR>95th =1.38 (0.84, 2.27)
Jeffery et al. UK Caucasian Retro- and 249 8 3rd trimester Weight z-score Not specified r=-0.09 (p=0.66) Not reported
2006 EarlyBird (mainly) prospective fasting glucose in (1990 UK)
women with
random glucose > Current weight Not reported r=0.23 (p=0.29)
6.5 mmol/l (n=26) HOMA-IR
78
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of maternal glucose tolerance in pregnancy
size in outcome glucose measurement confounders
analyses (years) intolerance in
pregnancy Unadjusted Adjusted

Krishnaveni et India Indian Prospective 455 9.5 OGTT BMI (continuous) Parental BMI GD vs. NGT:
al. 2010 Carpenter and Greater BMI in girls
Coustan 16.4 vs. 14.3 (p<0.001)

SFT Greater SFT in girls


triceps, 14.9 vs. 10.5 (p<0.001)
subscapular 14.1 vs. 7.6 (p<0.001)

HOMA-IR Current BMI Greater HOMA-IR Girls: not significant


Girls 1.2 vs. 0.9 Boys: significant but not
(p=0.006) presented
Boys:1.1 vs. 0.6
(p=0.002)
Lawlor et al. UK White, Prospective 6,842 9-11 GD screened for BMI z-score Pre-pregnancy BMI, Mean difference (95%CI) Mean difference(95%CI)
2010 Avon Non-white in high risk (1990 UK) age, SES, parity, GD: 0.32 (0.01, 0.63) GD: 0.01 (-0.30, 0.63)
Longitudinal women smoking, Glycosuria: 0.14 (0.01, Glycosuria: 0.09 (-0.05,
Study of 0.27) 0.22)
Parents ≥2 episodes of
and glycosuria Waist GD: 0.04 (0.00, 0.08) GD: 0.00 (-0.04, 0.06)
Children circumference z- Glycosuria: 0.02 (0.00, Glycosuria: 0.01 (0.00,
score 0.03) 0.03)

FM z-score (DXA) + height, height GD: 0.07 (-0.21, 0.36) GD: -0.16 (-0.43, 0.11)
squared Glycosuria: 0.18 (0.06, Glycosuria: 0.12 (0.01,
0.30) 0.23)
79
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of maternal glucose tolerance in pregnancy
size in outcome glucose measurement confounders
analyses (years) intolerance in
pregnancy Unadjusted Adjusted

Pirkola et al. Finland Caucasian Prospective 4,186 16 Selective BMI ≥ 85th centile Pre-pregnancy BMI, In Nwt women, relative In Nwt women, relative to
2010 Northern screening for GD (IOTF) smoking, paternal to women without risk women without risk
Finland with 75g 2-h BMI, size at birth factors for GD: factors for GD:
Birth OGTT women with GD women with GD OR=0.73
Cohort of OR=0.67 (0.24, 1.89) (0.26, 2.08)
1986 3 categories: women with normal women with normal
- GD if one OGTT OGTT
abnormal value at OR=1.18 (0.90, 1.56) OR=1.13 (0.83, 1.54)
OGTT
- normal OGTT Waist-to-height In Nwt women, relative In Nwt women, relative to
- control (no risk ratio > 0.5 to women without risk women without risk
factors for GD) factors for GD: factors for GD:
women with GD women with GD OR=1.22
OR=1.09 (0.39, 3.06) (0.43, 3.48)
women with normal women with normal
OGTT OGTT
OR=1.23 (0.89, 1.72) OR=1.16 (0.80, 1.67)
Silverman et al. USA Caucasian, Prospective 124 6-8 2nd and 3rd Symmetry index = Pre-pregnancy PIBW: r=0.28 (p<0.005) Not reported
1991 Diabetes in Black, trimester fasting relative weight / weight to calculate Amniotic fluid insulin:
Pregnancy Hispanic, plasma glucose relative height PIBW r=0.24 (p<0.05)
Study at other
GD vs. pre-gestational
North- diabetes: not significant
western
University
in Chicago
80
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of maternal glucose tolerance in pregnancy
size in outcome glucose measurement confounders
analyses (years) intolerance in
pregnancy Unadjusted Adjusted

Silverman et al. USA Caucasian, Prospective Not 14-17 2nd and 3rd BMI (continuous) Pre-pregnancy Children exposed to
1998 Diabetes in Black, clearly trimester fasting weight to calculate diabetes (T1D, T2D, GD)
Pregnancy Hispanic, reported plasma glucose PIBW in utero vs. control
Study at other 24.6 vs. 20.9 kg/m2
North- (p<0.001)
western In children exposed to
University diabetes (T1D, T2D, GD)
in Chicago PIBW: r=0.38 (p<0.001)
Amniotic fluid insulin: Amniotic fluid insulin:
r=0.29 (p=0.04) r=0.29 (p<0.05)
Whitaker et al. USA Non- Retrospective 524 5-10 Control (-OGCT) BMI z-score > Maternal pre- Mean BMI z-score in GD No effect of GD on BMI
1998 Hispanic (medical GD (+OGCT, ≥2 1.036 pregnancy BMI, vs. control offspring z-score
whites records) +values OGTT) – (US reference) paternal BMI at 0.39 vs. 0.45, p=0.40
(94%) Carpenter and delivery
Coustan (diet
treated) Maternal obesity Mothers with higher No longer significant
BMI≥27.3 kg/m2 quartiles of glucose after adjustment for
Non-fasting Paternal obesity levels at OGCT had parental obesity
glucose level at BMI≥27.8 kg/m2 children with lower
OGCT (quartile) obesity rates (p=0.05)
81
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of maternal glucose tolerance in pregnancy
size in outcome glucose measurement confounders
analyses (years) intolerance in
pregnancy Unadjusted Adjusted

Wright et al. USA White, Prospective 1,238 3 NGT (-OGCT) BMI z-score Maternal age, parity, No effect on BMI z-score Effect on BMI z-score
2009 Project Black, IGT (+OGCT, 0/1 (2000 CDC) ethnicity, education, NGT: 0.44 relative to NGT:
Viva Hispanic, + value at OGTT) smoking, family IGT: 0.52 IGT: 0.002 (-0.17, 0.17)
other GD (+OGCT, ≥2 history of diabetes, GD:0.47 GD: -0.08 (-0.37, 0.22)
+values OGTT) maternal pre-
pregnancy BMI (self-
Sum of reported weight and No effect on sum of SFT Effect on sum of SFT
subscapular SFT height), paternal BMI NGT: 16.6 relative to NGT:
and triceps SFT (self-reported weight IGT: 17.1 IGT: 0.25 (-0.48, 0.99)
and height), GWG GD:17.5 GD: 1.31 (0.08, 2.55)
(serial gestational
weights), BW

Ratio of + child current BMI z- No effect on ratio of SFT Effect on ratio of SFT
subscapular SFT score NGT: 0.64 relative to NGT:
and triceps SFT IGT: 0.67 IGT: 0.03 (-0.001, 0.06)
GD:0.65 GD: 0.01 (-0.04, 0.05)

Non-fasting As above As above Not reported No effect on:


glucose level at BMI z-score (-0.01)
OGCT (+10mg/dl) sum of SFT (0.03)
ratio of SFT (0.003)
%BF – percentage body fat, AGA – appropriate-for-gestational age, BIA – bioelectrical impedance analysis, BW – birth weight, CDC – Centers for Disease Control and Prevention, DXA – dual energy X-ray absorptiometry, FFM – fat-free mass, FGIR – fasting glucose to insulin
ratio, FM – fat mass, GD – gestational diabetes, GWG – gestational weight gain, HOMA-IR – homeostasis model assessment of insulin resistance, IGT – impaired glucose tolerance, IOTF – International Obesity Task Force, LGA – large-for-gestational age, MRI – magnetic
resonance imaging, NGT – normal glucose tolerance, OGCT – oral glucose challenge test, OGD – offspring of mothers with gestational diabetes, OGTT – oral glucose tolerance test, OND – offspring of non-diabetic mothers, OT1D – offspring of mothers with type 1 diabetes
PIBW - percentage of ideal body weight relative to the Metropolitan Life Insurance Company tables, r – Pearson’s correlation coefficient, SAT – subcutaneous (abdominal) adipose tissue, SES – socioeconomic status, SFT – skinfold thickness, T1D – type 1 diabetes, T2D – type 2
diabetes, VAT – visceral adipose tissue
82
Theoretical framework

Critique of longitudinal studies investigating the influence of maternal glucose intolerance


during pregnancy on child obesity and insulin resistance

The summarised studies assessing the relationship between various degrees of maternal glucose
intolerance during pregnancy and child obesity, percentage body fat, fat pattern and insulin
resistance have been limited by the way exposure and outcome variables were defined, confounding
adjustment, sampling frames and representativeness. These limitations are discussed below,
highlighting the gaps that the current project aims to address.

Exposure definition

The focus of most studies was contrasting the long-term effects of GD relative to normal glucose
tolerance during pregnancy (Boerschmann et al. 2010; Boney et al. 2005; Catalano et al. 2009a;
Crume et al. 2011a; Gillman et al. 2003; Krishnaveni et al. 2010; Pirkola et al. 2010; Whitaker et al.
1998), while only three studies separately considered milder degrees of glucose intolerance during
pregnancy (Egeland and Meltzer 2010; Hillier et al. 2007; Wright et al. 2009). There was a wide
variation in the criteria employed for diagnosing GD. For instance, Catalano et al. (2009a) assigned
the diagnosis of GD based on the stricter NDDG criteria (National Diabetes Data Group 1979),
whereas Whitaker et al. (1998) and Krishnaveni et al. (2010) on the more inclusive Carpenter and
Coustan criteria (Carpenter and Coustan 1982), and Boerschmann et al. (2010) on local German
criteria, with even lower thresholds for glucose concentration. Only one of the studies with lower
thresholds for glucose levels did not find an association between maternal GD and child BMI
(Whitaker et al. 1998). However, strong conclusions regarding associations between abnormal
glucose tolerance during pregnancy and obesity-related outcomes in children are precluded.
Moreover, extrapolating findings from these studies to settings where different diagnostic criteria are
in use might not yield accurate estimates.

As evidence is accumulating on the increasing risk of adverse outcomes across the entire spectrum
of maternal glucose tolerance during pregnancy (HAPO Study Cooperative Research Group 2008)
and not only in children of mothers with overt GD, several authors have considered milder degrees of
glucose intolerance during pregnancy in addition to GD, in relation to child outcomes. Inconsistent
categorisation makes interpretation and comparison of results difficult. Wright et al. (2009), for
example, defined a separate category of impaired glucose tolerance based on a positive OGCT and
none or one abnormal value at OGTT, while Egeland and Meltzer (2010) combined women with
impaired glucose tolerance and GD into one group (‘cases’), a combination which might have

83
Theoretical framework

weakened the true association between maternal GD and child obesity. In addition to separate
categories of glucose intolerance during pregnancy, Hillier et al. (2007) were able to differentiate the
effect of GD treatment on child outcomes, emphasising the role of variable degrees of
hyperglycaemia during pregnancy. Along the same lines of continuous risk across the range of
maternal hyperglycaemia during pregnancy, Chandler-Laney et al. (2011) found a positive
correlation between plasma glucose levels at OGCT (as a continuous variable) and metabolic
consequences in the child.

Other authors used more crude measures of glucose dysregulation during pregnancy than those
provided by the OGCT or OGTT. In the absence of universal screening in the UK at the time of
recruiting pregnant women for ALSPAC, glycosuria was used as a proxy for impaired glucose
tolerance during pregnancy (Lawlor et al. 2010); however, some of these women may have had
undiagnosed GD as supported by the very low prevalence of GD in this cohort (0.4%). Older
investigations considered second or third trimester fasting plasma glucose as an indicator of glucose
homeostasis during pregnancy (Silverman et al. 1991; Silverman et al. 1998). In an even less
reliable manner, the Growing Up Today Study considered maternal self-reported GD status and
assessed its influence on birth weight and BMI centile in adolescence (Gillman et al. 2003), which
might introduce information bias with subsequent implications on the interpretation of the results.

Outcome definition

The definitions of childhood global obesity were heterogeneous, making comparisons between
studies difficult. In some instances, analyses were based on BMI as a continuous variable (with no
reference to a specific population) (Crume et al. 2011a; Krishnaveni et al. 2010; Silverman et al.
1998), age- and sex-specific BMI centiles (relative to CDC (Chandler-Laney et al. 2011; Egeland and
Meltzer 2010; Gillman et al. 2003) or German (Boerschmann et al. 2010) reference data), BMI z-
scores (Catalano et al. 2009a; Lawlor et al. 2010; Wright et al. 2009), or symmetry index, another
measure of weight relative to height used in the past (Silverman et al. 1991). The precision of
measurement for BMI could be debated in the case of the Growing Up Today Study, which relied on
weight and height reported by the children themselves (Gillman et al. 2003). When data on
children’s height was not available, age- and sex-specific weight centiles (Hillier et al. 2007) or
weight z-scores (Jeffery et al. 2006; Whitaker et al. 1998) were used to define overweight status in
children despite the fact that these measures based on weight only are poor indicators of fatness
(Cole et al. 1995). Of these studies, only the large-scale one by Hillier et al. (2007) found an
association between maternal GD and child overweight status in younger children (5-7 year-old).

84
Theoretical framework

Similarly, different measures of adiposity were examined in relation to intrauterine exposure to


glucose intolerance, with variable accuracy of outcome assessment, ranging from the simple
measurement of subscapular and triceps SFT (Krishnaveni et al. 2010; Wright et al. 2009), to the
estimation of %BF by BIA (Egeland and Meltzer 2010) or DXA (Catalano et al. 2009a). One study
quantified visceral and subcutaneous abdominal adipose tissue by MRI (Crume et al. 2011a). The
conflicting findings of intrauterine programming of fatness by exposure to maternal glucose
intolerance could thus be attributed, at least in part, to differences in the methods of assessment. It
appeared that studies with less accurate measurements (SFT or BIA), which are the only practical
measures for large studies, were more likely to detect an effect (Egeland and Meltzer 2010;
Krishnaveni et al. 2010; Wright et al. 2009).

In terms of fat pattern, influences of intrauterine exposure to maternal glucose intolerance were
evaluated relative to offspring waist circumference (Catalano et al. 2009a; Chandler-Laney et al.
2011; Crume et al. 2011a; Egeland and Meltzer 2010), waist circumference z-score (Lawlor et al.
2010), or a ratio between central and peripheral SFT (Catalano et al. 2009a; Crume et al. 2011a;
Wright et al. 2009), but not relative to offspring waist-to-height ratio, which is a more robust indicator
of fat distribution.

Most studies in non-high risk populations focussing on child insulin resistance as an outcome of in
utero exposure to a diabetic environment estimated it by HOMA-IR (Boerschmann et al. 2010;
Catalano et al. 2009a; Egeland and Meltzer 2010; Jeffery et al. 2006; Krishnaveni et al. 2010) and
one study considered the ratio between fasting glucose and fasting insulin as a marker for insulin
resistance (Boney et al. 2005); however, in the latter study only FGIR (and not HOMA-IR) increased
the with exposure to GD in children born large-for-gestational age.

Confounders

Adjustment for several covariates was considered in the majority of these studies, but none of them
addressed the full set of potential confounders I have identified for the current study. The variables
adjusted for in different studies included maternal age, parity, ethnicity, smoking in pregnancy, family
history of diabetes, GD therapy, GWG, education (SES), size at birth, breastfeeding duration, and
parental BMI (either at the time of pregnancy or at follow-up). Importantly, maternal pre-pregnancy
BMI (or its proxy measure, early pregnancy BMI), a factor that is known to be associated both with
glucose intolerance during pregnancy and child obesity, was only adjusted for in a relatively limited
number of studies (Boerschmann et al. 2010; Catalano et al. 2009a; Lawlor et al. 2010; Silverman et

85
Theoretical framework

al. 1991; Silverman et al. 1998; Whitaker et al. 1998; Wright et al. 2009). When this adjustment was
undertaken, the effect was not consistent, ranging from attenuation of the GD effect on child BMI
(Catalano et al. 2009a; Crume et al. 2011a; Lawlor et al. 2010; Whitaker et al. 1998), to revealing an
effect not apparent without adjustment (Wright et al. 2009). The latter was noted in relation to the
sum of subscapular and triceps SFT, in the sense that only after full adjustment, exposure to GD
increased this sum by 31% relative to offspring whose mothers maintained normal glucose tolerance
during pregnancy (Wright et al. 2009). Other authors adjusted for maternal current BMI (Gillman et
al. 2003; Krishnaveni et al. 2010), which whilst not fitting the standard criteria for confounders
(discussed in Section 3.4.3) may have been included as a proxy for pre-pregnancy BMI.

Some of the studies evaluating the effect of GD on child insulin resistance only addressed potential
mediating role of current body size (weight or relative weight) (Jeffery et al. 2006; Krishnaveni et al.
2010), which attenuated the initial effect. However, none of these studies considered pregnancy-
related factors, such as maternal pre-pregnancy BMI or GWG, as potential confounders.

Sample representativeness and external validity

Sample representativeness in longitudinal studies is influenced by factors such as sample size,


sampling frame, and participation rates at follow-up. Several studies comprised large numbers of
participants (thousands) (Gillman 2004; Hillier et al. 2007; Lawlor et al. 2010; Pirkola et al. 2010),
whereas others were confined to samples smaller than 100 subjects (Catalano et al. 2009a;
Chandler-Laney et al. 2011). Larger scale studies were less likely to report an association between
maternal glucose intolerance during pregnancy and child obesity, which suggests that the
association identified in small-scale studies might be a result of chance.

Most samples included in these studies were representative of the general population, but a few of
them were not recruited at random, with potential implications for external validity. For instance,
studies conducted by Silverman et al. (1991, 1998) or Crume et al. (2011a) included a mixture of
women with GD and pre-gestational diabetes, and although an association with child weight relative
to height was found, separating the effect of hyperglycaemia induced by pregnancy from the
hyperglycaemia of pre-gestational diabetes was impossible. Although the Growing Up Today Study
comprised a large sample, it included only offspring of mothers who were nurses (Gillman et al.
2003), who, due to their health oriented profession, might have followed intensive treatment for GD
with likely effects on the degree of hyperglycaemia at which the fetus was persistently exposed and
subsequent effects on the post-natal outcomes. Another example of non-random sample came from

86
Theoretical framework

studies which included only families with private health insurance (Crume et al. 2011a; Hillier et al.
2007), thus not covering the entire socio-economic spectrum and potentially introducing selection
bias. Also, the finding of a positive association reported by Egeland and Meltzer (2010) might not be
generalisable to male children as their sample comprised girls only.

Participation rate, another factor that may alter sample representativeness over time, varied across
studies, ranging between 36% at 5-7 years in the data linkage study reported by Hillier et al. (2007)
and 81% in the prospective longitudinal study conducted in India (Krishnaveni et al. 2010). Potential
sources of bias were considered by some of the authors. In ALSPAC, with a participation rate of
50% at 9-11 years, it was found that participant mothers were older, more educated and had a
higher SES compared to those who did not take part at the follow-up (Lawlor et al. 2010). Similarly,
mothers from Project Viva participating at the 3 year follow-up (58% of the original sample) were
more likely to be older, white, had a higher income and lower pre-pregnancy BMI (Wright et al.
2009), which might explain the lack of an association between maternal GD and child global obesity
(however, there was an association with child SFT). Higher participation rates at 15 years were
identified among offspring of women with GD (51.6%) relative to those of mothers who maintained
normal glucose tolerance (24.4%) in the prospective study conducted by Egeland and Meltzer
(2010), which might contribute to the identification of an association between maternal GD and child
obesity and related outcomes. The response rate in the study conducted by Boney et al. (2005) was
not only differential (42.1% among women with GD and 10% among women with normal glucose
tolerance), but also low, with further potential effects on selection bias.

Children’s age also varied across studies from early childhood (Wright et al. 2009), pre-pubertal
(Boerschmann et al. 2010; Catalano et al. 2009a; Chandler-Laney et al. 2011; Hillier et al. 2007;
Jeffery et al. 2006; Krishnaveni et al. 2010; Lawlor et al. 2010; Whitaker et al. 1998), pubertal
(Egeland and Meltzer 2010; Pirkola et al. 2010), or combining multiple age groups (Crume et al.
2011a; Gillman et al. 2003). It was suggested that the long-term effects of GD on child obesity might
become apparent only during puberty (Crume et al. 2011a), but no clear trend by child age was
evident, as for a given age group some studies reported an association, while others did not.

Conclusion

This critical appraisal of studies addressing intrauterine programming of child obesity and insulin
resistance by maternal glucose intolerance during pregnancy shows several gaps in the literature,
that I aim to address with this thesis using a contemporary population of mother-child pairs living in

87
Theoretical framework

Australia. To the best of my knowledge, no study has been conducted in an Australian sample,
using the ADIPS criteria to define GD, looking at how maternal glucose tolerance across the entire
spectrum of severity (from borderline to GD) influences child BMI, percentage body fat, waist-to-
height ratio or HOMA-IR before puberty, with thorough adjustment for confounders. Although some
of these relationships have been previously reported in different populations (albeit with conflicting
findings), it is unlikely these estimates could be extrapolated to the Australian population, where a
different set of criteria to define categories of gestational glucose intolerance has been in use. In the
face of emerging evidence that any degree of maternal chronic hyperglycaemia during pregnancy
has detrimental effects on the child at birth, of particular importance is assessing the longer-term
influences of milder degrees of gestational glucose intolerance (which occur in a larger proportion of
women compared to GD). Studying these effects in pre-pubertal children is important given that
puberty may influence the outcomes of interest (Lee 2006) and interfere with the relationships
identified. While there is a suggestion that maternal gestational glucose intolerance may increase
the risk of obesity in the offspring, it is unclear whether it influences fat free mass or, conversely, the
adiposity amount and distribution, which are more sensitive markers of adverse long-term metabolic
health than BMI alone.

2.4.3 Intrauterine programming of obesity and insulin resistance by maternal gestational


weight gain

In recent years, it has been suggested that, in addition to maternal obesity and glucose intolerance
during pregnancy, gestational weight gain may affect the intrauterine milieu and thus fetal growth,
with subsequent offspring obesity risk later in life and potentially obesity-related metabolic disorders.
A summary of the longitudinal studies addressing this association in mother-child pairs is presented
in Table 8 and discussed below.

All but one of these studies have provided evidence of a positive, albeit relatively weak, association
between maternal weight gain during pregnancy and obesity in children of different ages (Fraser et
al. 2010; Moreira et al. 2007; Oken et al. 2007; Oken et al. 2008; Schack-Nielsen et al. 2005;
Sharma et al. 2005; Wrotniak et al. 2008) and only one study in young children did not (Whitaker
2004). Of the studies supporting an association between maternal GWG and child obesity, three
(Oken et al. 2007; Oken et al. 2008; Schack-Nielsen et al. 2010) reported a linear association, while
three (Fraser et al. 2010; Sharma et al. 2005; Wrotniak et al. 2008) a non-linear one. Two of the
latter studies (Sharma et al. 2005; Wrotniak et al. 2008) identified an interaction between GWG and

88
Theoretical framework

maternal pre-pregnancy BMI in relation to child obesity, with the association being strongest in the
group of underweight women. It is unclear why the studies reported different types of associations,
but may partly be due to heterogeneous exposure and outcome definitions.

The evidence is very limited with regard to more specific measures of adiposity. Only one study
addressed the influence of maternal GWG on the sum of subscapular and triceps SFT in the
offspring and found no association, despite reporting an influence on child global obesity risk (Oken
et al. 2007). The two studies investigating the relationship between maternal weight gain during
pregnancy and fat pattern yielded conflicting results, with Fraser et al. (2010) reporting a positive
association with child waist circumference at 9 years, and Oken et al. (2007) finding no association
with the ratio of central to peripheral SFT in 3 year old children. Besides the difference in children’s
age, there were differences in the measurement of weight gain in pregnancy, which might contribute
to the inconsistent results.

No studies have been published to date on obesity-related metabolic disorders, such as insulin
resistance, in children in relation to maternal GWG.

89
Table 8. Summary of longitudinal studies examining the influence of maternal gestational weight gain on child obesity

Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of GWG
size in outcome GWG measurement confounders
analyses (years) Unadjusted Adjusted

Fraser et al. UK White, non- Prospective 5,154 9 GWG=last weight BMI continuous Pre-pregnancy BMI Mean difference (95% CI) Mean difference (95% CI)
2010 Avon white measured in WC continuous (predicted pre- relative to adequate relative to adequate
Longi- pregnancy – first pregnancy weight GWG based on IOM GWG based on IOM
FM (DXA)
tudinal weight measured using multilevel recommendations: recommendations:
Study of in pregnancy models, self-reported
Parents (obstetric records) height), maternal age, Inadequate GWG: Inadequate GWG:
and parity, smoking, GWG BMI: -0.29 (-0.47, -0.12) BMI: -0.33 (-0.50, -0.15)
Children in previous period, WC: -0.83 (-1.31, -0.35) WC: -0.90 (-1.38, -0.42)
Inadequate / SES (parental FM: -217 (-497, 63) FM: -260 (-540, 21)
adequate (ref) / occupation)
excessive Excessive GWG: Excessive GWG:
(2009 IOM) BMI: 0.78 (0.59, 0.97) BMI: 0.74 (0.55, 0.94)
WC: 1.97 (1.46, 2.49) WC: 1.93 (1.41, 2.45)
FM: 1162 (860, 1464) FM: 1075 (773, 1378)

Not reported Relative to adequate GWG:


BMI >85th centile
Inadequate GWG:
(IOTF)
ORBMI>85th = 0.80 (0.67,
0.96)
WC ≥ 90th centile ORWC≥90th = 0.79 (0.69,
(British reference) 0.90)
Excessive GWG:
ORBMI>85th = 1.73 (1.45,
2.05)
ORWC≥90th = 1.36 (1.19,
1.57)
90
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of GWG
size in outcome GWG measurement confounders
analyses (years) Unadjusted Adjusted

Oken et al. USA White, Prospective 1,044 3 Total GWG = last BMI centile Maternal pre- OR≥95th =1.30 (1.04, 1.62) OR≥95th =1.52 (1.19, 1.94)
2007 Project Black, recorded weight <50th (reference) / pregnancy BMI, age,
Viva Hispanic, before delivery – 50-84th / 85-94th / parity, smoking, SES
other self-reported pre- ≥95th (2000 CDC) (income, education),
pregnancy weight ethnicity, glucose BMI z-score mean BMI z-score mean
BMI z-score (2000 tolerance status in difference (95% CI) for difference (95% CI) for
CDC) pregnancy, gestation each +5 kg in GWG: each +5 kg in GWG:
length, BW z-score 0.10 (0.04, 0.16) 0.11 (0.05, 0.17)
breastfeeding
Sum of duration, sex, paternal Sum SFT mean Sum SFT mean
subscapular SFT BMI difference (95% CI) for difference (95% CI) for
and triceps SFT each +5 kg in GWG: each +5 kg in GWG:
0.18 (-0.06, 0.42) 0.25 (0.00, 0.50)

Ratio of Ratio SFT mean Not reported


subscapular SFT difference (95% CI) for
and triceps SFT each +5 kg in GWG:
0.003 (-0.01, 0.01)

Inadequate (ref) / Not reported Relative to inadequate


adequate / GWG:
excessive Adequate GWG:
(1990 IOM) OR≥95th =3.77 (1.38, 10.27)
BMI z: 0.47 (0.37, 0.57)

Excessive GWG:
OR≥95th =4.35 (1.68, 11.24)
BMI z: 0.52 (0.44, 0.61)
91
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of GWG
size in outcome GWG measurement confounders
analyses (years) Unadjusted Adjusted

Oken et al. USA Nationwide, Linkage 11,994 9-14 Total GWG BMI z-score Maternal pre- Prevalence of BMI ≥95th Effect of 5-lb increase in
2008 Growing but homo- (self-reported) (self-reported pregnancy BMI (self- had a U-shaped GWG:
Up Today genous in weight and height) report), age, smoking, distribution across GWG OR85-94th =1.05 (1.03,
Study terms of SES (income, (highest among offspring 1.08)
race / 5-lb increments paternal education), of mothers with GWG OR≥95th =1.08 (1.05, 1.13)
& ethnicity BMI centile GD, ethnicity, BW, <14 lb or >45 lb and BMI z: 0.03 (0.02, 0.04)
Nurses’ <85th (ref) / 85-94th breastfeeding lower in between)
Health / ≥95th duration, child’s age Relative to adequate
Study II in 1996, sex, Tanner GWG based on IOM
(2000 CDC)
stage, diet, physical recommendations:
Inadequate / activity
adequate (ref) / Inadequate GWG:
excessive OR85-94th=0.97 (0.84,
1.12)
(1990 IOM) OR≥95th=0.91 (0.74, 1.13)
BMI z:-0.06 (-0.10, -0.01)

Excessive GWG:
OR85-94th=1.27 (1.12,
1.44)
OR≥95th=1.42 (1.19, 1.70)
BMI z: 0.14 (0.09, 0.18)
Schack-Nielsen Denmark Caucasian Prospective 2,034 1 Total GWG BMI z-score (1990 Maternal pre- Not reported BMI z-score mean
et al. 2005; Copen- 1,408 3 (obstetric records) British reference) pregnancy BMI, age, difference (95% CI) for
2010 hagen parental SES, each ‘category’ of GWG:
1,070 6 5 categories smoking, edema
Perinatal 1 y: 0.03 (0.02, 0.05)
Cohort 1,037 8 assigned the during pregnancy, BW 3 y: 0.02 (0.00, 0.04)
940 14 interval middle 6 y: 0.02 (0.00, 0.04)
value: <6 (5.5), 6- 8 y: 0.03 (0.02, 0.05)
8 (7), 9-10 (9.5), 14 y: 0.03 (0.01, 0.05)
11-12 (11.5), 13-15
(14), >16 (16.5) kg
92
Reference Setting Population Study design Sample Age at Measurement of Outcome Covariates, Effect of GWG
size in outcome GWG measurement confounders
analyses (years) Unadjusted Adjusted

Sharma et al. USA Not Data linkage of 165,013 2-4 Routinely BMI ≥ 95th centile Age, sex, ethnicity, No estimates reported Attenuated by BW
2005 specified Pediatric & collected GWG (2000 CDC) state of birth, Non-linear association adjustment, but still
Pregnancy maternal age, height, (p<0.0001) significant
Nutrition smoking, education, Effect modification by
- low Surveillance BW
income maternal pre-pregnancy
Systems BMI (p<0.0001)
Whitaker 2004 USA White, Retrospective, 8,494 2-5 Net GWG=(self- BMI ≥ 95th centile Not specified for this Not reported Not significant
Black, data linkage reported GWG- (2000 CDC) particular exposure
Hispanic, BW) / length of
other gestation
For main analyses:
BW, gestational age,
- low- Quartiles gender, ethnicity,
income (lowest=ref) parity, age, smoking,
education, marital
status
Wrotniak et al. US Multiethnic Retrospective 10,266 7 Total GWG = BMI ≥ 95th centile Maternal pre- + 1 kg GWG: + 1 kg GWG:
2008 National measured weight (2000 CDC) pregnancy BMI, age, OR=1.02 (1.00, 1.03) OR=1.03 (1.02, 1.05)
Collabo- at delivery – self- race, smoking,
reported pre- gestational age, Inadequate GWG: Inadequate GWG:
rative
pregnancy weight child’s sex, age, first OR=0.82 (0.69, 1.01) OR=0.88 (0.68, 1.14)
Perinatal
Project born status
Excessive GWG: Excessive GWG:
Inadequate / OR=1.62 (1.25, 2.12) OR=1.48 (1.06, 2.06)
adequate (ref) /
excessive + BW + 1 kg GWG:
(1990 IOM) OR=1.03 (1.01, 1.05)

Inadequate GWG:
OR=0.93 (0.72, 1.21)

Excessive GWG:
OR=1.40 (1.00, 1.95)
BW – birth weight, CDC – Centers for Disease Control and Prevention, DXA – dual energy X-ray absorptiometry, FM – fat mass, GWG – gestational weight gain, IOM – Institute of Medicine, IOTF – International Obesity Task force, SES – socioeconomic status, SFT – skinfold
thickness, WC – waist circumference
93
Theoretical framework

Critique of longitudinal studies investigating the influence of maternal gestational weight gain
on child obesity

The longitudinal studies examining the association between maternal weight gain during pregnancy
and child obesity, body composition and fat pattern have been limited by the definition of exposure
and outcome variables, adjustment for potential confounding factors, sampling frame and
representativeness. Insulin resistance in children, a common consequence of obesity, has not been
studied in relation to intrauterine exposure to excessive GWG. These studies are critiqued below,
with a view to clarifying gaps that the present study aims to address.

Exposure definition

No consensus was adopted in defining GWG in the summarised studies (Table 8), with some
expressing it as total (Fraser et al. 2010; Oken et al. 2007; Oken et al. 2008; Schack-Nielsen et al.
2005; Sharma et al. 2005; Wrotniak et al. 2008), while others as net GWG (separating maternal from
fetal components) (Oken et al. 2007; Whitaker 2004). More importantly, the precision of GWG
measurement varied across studies, including routinely collected data (Schack-Nielsen et al. 2005;
Sharma et al. 2005) with minimal information bias, direct measurement of at least one weight during
pregnancy (Fraser et al. 2010; Oken et al. 2007; Wrotniak et al. 2008), or self-reported GWG (Oken
et al. 2008; Whitaker 2004) with potential information bias. In addition to continuous measures of
GWG, categories (‘inadequate’, ‘adequate’, or ‘excessive’ weight gain based on the Institute of
Medicine recommendations, 1990 or 2009) were considered as predictors for obesity risk in the
offspring. Of these, the ‘adequate’ GWG category was chosen as the reference by most authors
(Fraser et al. 2010; Oken et al. 2008; Wrotniak et al. 2008). In contrast, Oken et al. (2007)
considered the ‘inadequate’ GWG as the reference category, which likely explains the much higher
odds of obesity reported in offspring exposed to excessive GWG. Other authors used internal
quartiles (Whitaker 2004) or more arbitrary categories (Schack-Nielsen et al. 2005) of GWG as
exposure variables. These variations in measurements are likely to have influenced the associations
reported by different studies and made comparisons challenging.

Outcome definition

Various indices of childhood global obesity were used in these studies, based mainly on direct
measures of weight and height, one on measurements collected from school health records (Schack-
Nielsen et al. 2005). Oken et al. (2008), however, based BMI calculation on child self-reported
weight and height, with potential impact on the validity of the results due to the established tendency

94
Theoretical framework

of adolescents to under-report body weight (Elgar et al. 2005). BMI was expressed as a continuous
variable, not standardised for age and sex (Fraser et al. 2010), as a z-score (Oken et al. 2007; Oken
et al. 2008), or as a centile relative to a specific population (CDC - (Oken et al. 2007; Oken et al.
2008; Sharma et al. 2005; Whitaker 2004; Wrotniak et al. 2008), or IOTF - (Fraser et al. 2010)). In a
further point of distinction between studies, different centiles were employed as reference categories
when calculating the odds of childhood obesity – e.g., 50th centile by Oken et al.(2007) and 85th
centile by Oken et al.(2008). As a result of these different definitions, the reported adjusted ORs for
childhood overweight/obesity in relation to maternal excessive GWG ranged from 4.35 in 3 year-old
children from Project Viva (Oken et al. 2007), to 1.73 in 9 year-old children enrolled in ALSPAC
(Fraser et al. 2010), and 1.42 in 9-14 year-old children from the Growing Up Today Study (Oken et
al. 2008). In addition, as the association between maternal GWG and the risk of obesity in the
offspring seems to decrease as children grow older, it could be argued that the different ages of the
children included in these studies also contributed to the wide variation in odds of childhood obesity
reported by different authors.

The amount of adipose tissue in children was investigated in relation to exposure to maternal GWG
in two studies. In one of them, fat mass measured by DXA was increased by about one kg in 9 year-
old children if the mother gained weight excessively during pregnancy, robust to confounder
adjustments (Fraser et al. 2010). In the other study, the sum of subscapular and triceps SFT in 3
year-old children was positively associated with GWG after adjustment for a series of covariates,
including parental BMI, but further adjustment for birth weight z-score attenuated the association
(Oken et al. 2007). The same two studies evaluated the influence of maternal GWG on child fat
pattern and reported opposite effects: positive association of waist circumference at 9 years with
GWG (Fraser et al. 2010) and no relationship between GWG and the ratio of central-to-peripheral
SFT in 3 year olds (Oken et al. 2007).

Confounders and effect modification

Each of the studies on the relationship between maternal GWG and child obesity adjusted for a
different set of covariates, including pre-pregnancy BMI, maternal age, parity, smoking, various
indicators of SES, ethnicity, gestational length, first born status, birth weight, breastfeeding duration,
or pubertal (Tanner) stage. Glucose intolerance during pregnancy, a factor considered to be on the
causal pathway between maternal GWG and child obesity, was taken into account only by Oken et
al. (2007; 2008).

95
Theoretical framework

Discordant findings were reported on a potential interaction between GWG and pre-pregnancy BMI
in relation to child obesity. Of the studies addressing interactions, four supported an effect
modification by pre-pregnancy BMI (Fraser et al. 2010; Oken et al. 2007; Oken et al. 2008; Schack-
Nielsen et al. 2005) while another two did not (Sharma et al. 2005; Wrotniak et al. 2008). This
difference would not seem to arise from a lack of statistical power as the studies with no interactions
were large.

Sample representativeness and external validity

Although published recently, two of the studies reviewed in this section comprised historical cohorts
established between 1959 and 1972 (Schack-Nielsen et al. 2005; Wrotniak et al. 2008), when the
prevalence of ‘inadequate’ GWG was higher (65%) and that of ‘excessive’ GWG was lower (11%)
(Wrotniak et al. 2008) compared to contemporary Western society (National Center for Health
Statistics 2010). Despite this, the associations are still likely to be relevant.

All the samples included in these analyses were large, with thousands of participants from various
ethnic backgrounds. Despite the large sample sizes, two of these studies were not representative of
the general population. As already mentioned, mothers of children enrolled in the Growing Up Today
Study were nurses, who, as a consequence of their health oriented profession, might have had a
higher rate of ‘adequate’ GWG (Oken et al. 2008). Second, the Copenhagen Perinatal Cohort
comprised mainly women with current or previous pregnancy complications (such as edema,
hypertension, proteinuria, pre-eclampsia) (Schack-Nielsen et al. 2005), which might have increased
disproportionately the water component of GWG, which is less likely to have direct metabolic effects
on the offspring, as opposed to the fat component.

Attrition, which may affect original sample representativeness, varied across studies and was highest
in the historical cohort reported on by Schack-Nielsen et al (2005) (in this study, of the 9,125
participants from the original sample, about 1,000 children were assessed at each follow-up beyond
infancy). Among the contemporary cohorts, participation rates ranged between 44.8% at 9-14 years
in Growing Up Today Study (Oken et al. 2008) and 57% at 9 years in ALSPAC (Fraser et al. 2010).
Details regarding the likelihood of bias due to attrition were reported only by Oken et al.(2007) who
noted that participants (49.1% of original cohort) were more likely to be white and to have lower pre-
pregnancy BMI, but had similar GWG as non-participants; thus, in principle their results should not
be affected by this relatively low participation.

96
Theoretical framework

Conclusion

This critical appraisal of studies on intrauterine programming of child obesity by maternal GWG
indicates a need for a study comprising a representative, contemporary Caucasian population of
children. Australia is a site where such research should be undertaken due to the adequacy of
routine data collection, relative affluence and potential generalisability to other developed Western
nations. The basis for conducting such a study is that, while there seems to be some evidence of
GWG effects on offspring BMI and very limited evidence on fat mass or waist circumference, no
studies have examined how maternal GWG across the entire spectrum influences child percentage
body fat, waist-to-height ratio and HOMA-IR before puberty. These are the gaps identified in the
existing literature that the current project aims to address. Examining maternal GWG as a continuous
variable has the advantage of removing arbitrary cut-offs (preferred in clinical approaches to data
analysis, but less acceptable from an epidemiological perspective), thus allowing the description of
the effect of incremental increases in GWG on child outcomes. It is possible that maternal GWG
influences offspring overall body size, but it is still uncertain whether it affects offspring adiposity or
fat distribution, measures that are more closely linked to long-term metabolic perturbations than
global obesity. Hence, the current study includes these additional outcome measures. If maternal
GWG appears to influence child insulin sensitivity (relationship that has not been previously
investigated), this might represent a step in identifying modifiable early life factors that contribute to
insulin resistance prior to puberty. In addition, in existing studies the adjustment for confounders has
often been incomplete, which may have led to reporting false associations.

2.5 Summary

From a public health perspective, identifying specific gestational factors with a potential to influence
body size, body composition, fat distribution and glucose-insulin homeostasis in the next generation
appears increasingly appealing. Once these maternal factors are identified, attention could be
directed to means by which they might be modified, thus contributing to the prevention of later
metabolic disorders in the child, such as obesity and insulin resistance.

97
Chapter 3 Methodology

This chapter outlines a brief description of the Generation 1 study in which my research project is
nested (Section 3.1), relevant aspects of data collection in the 9-10 year old children, including
details about approaching the families, procedures undertaken with participants, and ethical
considerations (Section 3.2). Means of managing the data are presented in Section 3.3, followed by
a clear definition of variables of interest (Section 3.4). The chapter concludes with a description of
the analytical framework adopted (Section 3.5).

3.1 The Generation 1 cohort

Given that my proposed research project was based on an already established birth cohort of
children, I consulted and drew upon three documents written by Associate Professor Vivienne
Moore, the principal investigator in the Generation 1 Study: a paper describing in detail eligibility
criteria for women and baseline data collection (Moore et al. 2004), an Ethics application (number H-
167-2006) and a National Health and Medical Research Council Grant application (number 453655),
the latter two outlining data collection for this project, which was largely designed prior to the start of
my candidature.

3.1.1 Study design and overall aim

The current project was an extension of the ongoing Generation 1 Study, a prospective cohort study
which commenced in 1998. At its inception, the overall aim of the study was to examine early life
origins of health and disease. Within the broader Generation 1 study, my project in particular aimed
to provide understanding of intrauterine environment influences (namely, maternal obesity prior to
pregnancy, glucose intolerance during pregnancy, and excessive gestational weight gain) on child
obesity, body composition, fat pattern and insulin resistance at the age of 9-10 years.

98
Methodology

Up to this stage, extensive data about pregnancy and growth of the children during infancy and in the
first 5.5 years of life have been collected over ten waves. In this phase of follow-up, at 9-10 years of
age, my focus was on describing the growth of this group of South Australian children (using
anthropometric measurements, namely, BMI z-score), assessing their body composition (in particular,
percentage body fat), their fat pattern (by waist-to-height ratio) and glucose-insulin homeostasis
(using a fasting blood sample).

3.1.2 Baseline sampling process and selection criteria

The Generation 1 cohort was established prospectively between 1998 and 2000. In order to provide
representative recruitment across the entire socioeconomic spectrum and cover various dietary
patterns, a sampling strategy was devised in which women were enrolled in early pregnancy through
the antenatal clinic at a public hospital (Lyell McEwin Hospital, Adelaide, South Australia) and the
rooms of three privately practising obstetricians (centrally located in the Adelaide area, with large
caseloads). At the public hospital, pregnant women were approached following a random schedule
(about 5 women per week, in order to limit the number of recruited women within a manageable load
for the fieldworkers). At the private practices all eligible women were approached (1-2 women per
week in each practice). Approximately half of the study participants were identified through the
public hospital, similar to the proportion of South Australian women who received antenatal care in a
public hospital clinic at that time (Chan et al. 2000).

To be eligible to join the study (Moore et al. 2004), women had to:

 be in the first 16 weeks of gestation (so that detailed information about early pregnancy,
specifically dietary intake, could be recalled accurately);

 have a singleton pregnancy (because there is an increased risk for low birth weight in multiple
pregnancies (Shinwell and Blickstein 2007; Taylor et al. 1998));

 have spontaneously conceived (since intrauterine growth is often restricted in pregnancies


occurring through assisted reproductive technology (Allen et al. 2006; Fisch et al. 1997));

 be free from certain medical conditions known to severely affect fetal growth (e.g., diabetic
mothers have increased risk of delivering macrosomic babies (Koukkou et al. 1997; Platt et al.
2002));

 be at least 18 years of age (as teenage mothers are more likely to give birth to growth
restricted babies (Chandra et al. 2002; Cooper et al. 1995));
99
Methodology

 be Caucasian (as assessment could only be done for Caucasian diets); and

 be sufficiently fluent in English to give informed consent and to complete the study
questionnaires.

Among the eligible women who were approached, 65% consented to participate in the study. Of
them, 557 women completed the pregnancy phase (representing 92% of women who enrolled in the
study) and gave birth to a baby that survived the neonatal period (Figure 1). Women who did not
complete the pregnancy phase were not invited to continue in the study. The ethics committee
required exclusion of women who gave birth to babies with serious congenital abnormalities, but
there were none.

3.1.3 Baseline data collection

Pregnant women who joined the study were interviewed by two research nurses, face-to-face, on
two occasions: in early pregnancy (between 8 and 20 weeks of gestation) and in late pregnancy
(between 27 and 42 weeks of gestation). A rich set of data was collected from these women both
through the interviews and from antenatal records. This information included, but was not limited to,
women’s age, results of screening for gestational diabetes (taken from the antenatal records), pre-
pregnancy weight (self-reported), height and weight measured in early and late pregnancy, parity,
personal and family medical history, blood pressure, diet, physical activity, smoking, education and
other social circumstances (employment status, household income, relationship status, family size),
and details regarding complications of the pregnancy (such as pregnancy-induced hypertension,
preeclampsia or gestational diabetes).

Children in this study were born in five hospitals in Adelaide. Special arrangements were made for
measurements of the babies and placentas for study purposes by midwifes. In addition, routinely
recorded data and birth details were abstracted from the hospital records.

100
Methodology

Figure 1. Flowchart of participation

605 women
joined the study

8% 48 women
withdrew from the study

27 (56%) – miscarriage or termination of pregnancy


21 (44%) – other reasons: moving away from Adelaide,
92% changing the hospital in which they intended to deliver, or
no longer wishing to be involved

557 women
completed the pregnancy
phase and delivered a baby

2.15% * 12 families withdrew from the cohort

0.36% *
2 children died

1.07% *
6 families lost to follow-up

537 children
approachable at 9-10 years

*% of women who completed the pregnancy phase and delivered a baby

3.1.4 Participation until this follow-up

After birth, detailed information was collected about family circumstances and growth of the
Generation 1 children every 3 months in the first year and at 2, 3.5 and 5.5 years of age. A
comprehensive description of data collected on every follow-up is not warranted here, but some
aspects that have set the scene for the current phase at 9-10 years and played a critical role to the
success of this project are outlined below.

Periodic tracing of participants, both manual and electronic, was an essential process for maintaining
the cohort as intact as possible and thus preventing attrition, which is an important methodological
issue for longitudinal studies. Irrespective of the form attrition takes (either as a premature

101
Methodology

discontinuation of participation or declining some follow-ups but then returning to the study at future
stages (Twisk and de Vente 2002)), it may compromise the internal and external validity of the study
(Garcia et al. 2005), result in a loss of statistical power and may lead to selection bias (Kristman et
al. 2004).

In order to minimise attrition, a series of strategies were implemented from the inception of the study
and for each follow-up (Table 9). As a result of these strategies, a trusting relationship was built
between the families and the research team. This has contributed towards the high participation rate
achieved (over 90% at each face-to-face follow-up). Only 20 families of the original 557 (less than
4%) had withdrawn before the latest follow-up, leaving 537 families to approach at this round. The
attrition that did occur in the Generation 1 study was due to the researchers’ inability to relocate
subjects after they had changed addresses (n=6), family’s decision of no longer wishing to be
involved (n=12) or subject’s death (n=2) (Figure 1).

Table 9. Main strategies to prevent loss to follow-up and enhance response rates

Enrolment Inform participants of requirements of the study


Collect extensive contact details (address, phone number, date and place of
birth, contact details of personal contacts)
Regular contact Newsletters with study updates, holiday cards
Research team Well trained, enthusiastic, warm, communicative, respectful, stable, flexible with
scheduling
Incentives Small tokens of appreciation, parking tickets, reimbursement for petrol

3.2 Data collection for the follow-up at 9-10 years of age

It is worth noting that the current thesis is one part of a much more extensive project. In short, the
main outcomes of interest for this research are child obesity (defined by child BMI z-score), %BF
(derived from BIA), fat pattern (indicated by WHtR), and IR (estimated by HOMA-IR). A detailed
description of data collection for these parameters is presented below, along with an outline of other
data collected that was not included in my analyses, but could offer a better understanding of the
entire Generation 1 follow-up at 9-10 years of age.

102
Methodology

3.2.1 Approaching participant families

All Generation 1 families who had indicated that they would be willing to consider future involvement
in the study were contacted by letter. In some cases, retracing of the families was needed, including
approaching alternate contact persons, and using electronic databases such as the electronic
telephone directory and the Electoral Roll. An information sheet was sent to each family, explaining
the procedures involved in this phase of the follow-up and also repeating their rights as study
participants. The information sheet was followed by a phone call within a week (where possible) to
answer queries and see if families wished to participate. Of the 537 mother-child pairs that were
approached, 443 (82.5% of current cohort members, 80% of the original cohort) agreed to take part
in this follow-up.

3.2.2 Procedures undertaken with participant children

The follow-up of Generation 1 participants at the 9-10 year round comprised three main parts: a visit
either at home or in one of the offices assigned by the research team (for interview and
anthropometric measurements), assessment of body composition, and a fasting blood sample. Each
home/office visit was attended by two trained fieldworkers, a research nurse to conduct the interview
with the mother, and myself or another PhD student to work with the child.

3.2.2.1 Interview

After renewing the written consent for agreement to participate in the study by the mother, one
fieldworker conducted an interview with the child, assessing their physical activity and diet. Consent
will be described in more detail in Section 3.2.3. The child’s physical activity was examined by
‘Multimedia Activity Recall for Children and Adolescents’ (MARCA) (Ridley et al. 2006). This is a
validated computer-delivered, self-report proforma which collects data on a wide range of physical
activities and sedentary behaviours from the previous day, providing an estimation of energy
expenditure over 24 hours (Ridley et al. 2006). The child’s diet in the past 6 months was assessed
by a Food Frequency Questionnaire entitled ‘Australian Children’s Eating Survey’ (Watson et al.
2006). Information on the child’s physical activity and diet was not considered in this project, but is
readily available for future research.

103
Methodology

While the child was being interviewed, the other fieldworker administered a questionnaire to the
mother covering aspects of her health, the child’s health, health-related behaviours, family socio-
economic circumstances, and parenting practices.

3.2.2.2 Anthropometric measurements and assessment of body composition

The study protocol included a range of anthropometric measurements in both mothers and children
who agreed to take part: standing height, sitting height, skinfold thickness (triceps, biceps,
subscapular, and suprailiac sites), and body circumferences (head, waist, hip, mid-arm and mid-calf).
These measurements were done based on a protocol (Norton and Olds 1996) endorsed by the
International Society for the Advancement of Kinanthropometry (ISAK) (Marfell-Jones and Olds
2007), by one of the three research nurses involved in this study, who had been previously trained in
performing them with an acceptable interpersonal variability. The measurements relevant to the
current project are presented below:

 Standing height to the nearest 0.1 cm was measured with the participant in bare feet, having
their feet together, with the heels, buttocks and the upper part of the back touching the scale,
and the plane going through the lower margin of the eye socket and the notch above the tragus
of the ear in a horizontal position (Norton and Olds 1996). Height measurements were
performed twice for each subject using a portable stadiometer, and the mean value was used in
the analyses.

 Weight to the nearest 0.1 kg was determined with the participant wearing minimal clothing and
no shoes (Norton and Olds 1996), using Tanita body composition analyser TBF-300 (Tanita
Corp., Tokyo, Japan) for a standard body build.

 Waist circumference to the nearest 0.1 cm was measured at the level of the narrowest point
between the lowest rib margin and the iliac crest (or the mid-point between these two references
if no narrowing was obvious) in a horizontal plane, at the end of a normal expiration, with the
participant in standing position, balanced on both feet, with arms extended comfortably (relaxed)
by their side (Norton and Olds 1996), using a flexible, non-extensible tape measure. Two to
three measurements were obtained for waist circumference and the mean value was used in
the analyses.

These anthropometric measurements were performed in all (n=443) participating children, except for
waist circumference measurements, which had one missing value.

104
Methodology

Body composition was assessed in all mothers and children who participated in this round of
follow-up, using a foot-to-foot bioelectrical impedance analyser Tanita TBF-300 (Tanita Corp., Tokyo,
Japan). The estimates of percent body fat, fat mass, fat free mass and total body water were
derived from specific equations embedded in the machine (undisclosed by the manufacturer), based
on weight, height, age and sex (all children were considered as having a ‘standard’ body build). This
technique was chosen as it is portable, non-invasive, easy to perform, safe (Hosking et al. 2006) and
inexpensive.

3.2.2.3 Fasting blood samples

Fasting blood samples were required in order to assess the child’s glucose-insulin metabolism, one
of the main outcomes for this project. At the end of the interview, the research nurse discussed with
the family the possibility of the child having a fasting blood test (for research purposes only). If the
mother’s written consent and the child’s verbal assent were obtained, the interviewers arranged for a
domiciliary nurse from a private blood collection facility to visit the family on an agreed day, at 8 am.
Fasting for 12 hours before the blood sample collection (nothing to eat or drink other than water from
8 pm the previous night) was requested. The phlebotomist checked that the child had fasted and
was still willing to provide a blood sample. Families were given a local skin anaesthetic (EMLA
cream) to apply about 1 hour prior to the appointment, in order to minimise pain associated with
needle insertion. A fasting blood sample (12 ml) was collected from 164 children (37% of children
who took part at this follow-up). Blood was initially processed by the collection laboratory, then
plasma and serum aliquots were frozen and subsequently couriered in batches to the University of
Adelaide Medical School for the assays and storage.

For the purpose of the current research study, the assays of interest were plasma glucose and
plasma insulin. Plasma glucose was quantitatively determined with a Hitachi 912 automated sample
system using the Glucose HK assay kit (Roche Diagnostics, NSW, Australia). The mean coefficient
of variation was around 3.3%. The reference range for the assay used was 3.33–5.55 mmol/l.
Plasma insulin was measured using a commercially available radioimmunoassay kit specific for
human insulin (Linco, Millipore). The mean coefficient of variation was around 5%. The reference
range for the assay used was 5-15 μU/ml.

105
Methodology

3.2.3 Ethical considerations

From its inception, the study has complied with the ethical standards for medical research involving
human participants (World Medical Association Declaration of Helsinki 1997). At baseline, ethics
approval was obtained from all hospitals and antenatal clinics involved in subject recruitment for the
study. For this phase of the follow-up, ethics approval was obtained from the University of Adelaide
Human Research Ethics Committee.

Obtaining consent and co-operation from participants

Participants were provided with written information about the study and also the opportunity to
discuss any study-related queries with a research nurse. As in previous rounds, written consent for
the interview and physical measurements was obtained from the mother and verbal assent from the
child, at the beginning of the visit. At the end of the interview, the possibility of the child having a
fasting blood test was discussed with the mother and the child. Given the fragility of any cohort, it
was emphasised that both the blood test was optional and the family could continue to be part of the
cohort even if they declined this component.

Potential risks for participants

Although the interviews did not contain obviously sensitive questions, participants were informed at
the beginning of the interview that they were free not to answer any questions they would find
uncomfortable (although they had been asked for similar information previously, mainly by the same
research nurses).

The anthropometric measurements and assessment of body composition by BIA were not invasive
or painful. The callipers for SFT measurement exerted some degree of pressure on a fold of skin,
uncomfortable at times, but normally not painful. The protocol was to discontinue measurements if
any child was distressed or found them painful and this did occur in one case.

With respect to the blood sample, overnight fasting for 12 hours before collection was not a harmful
procedure. In order to prevent pain associated with needle insertion, a local skin anaesthetic (EMLA
cream) was applied about one hour prior to blood collection. Also, the blood sample was collected
by an experienced phlebotomist, who had been trained to reassure participants before, during and
after the sample was taken, and to respond to any signs of faintness.

106
Methodology

Preserving confidentiality of participants

Confidentiality of participants has been totally preserved. They were identified on data collection
forms by a study ID number and not by name. The forms have been stored in lockable filing
cabinets in an area with restricted access and computerised data in password-protected files. Data
analyses have concerned groups of participants, so women and their children could not be
individually identifiable in any reports or papers. However, study participants were allowed to
request access to their individual data and results.

Payment, reimbursement or other inducements

Families were not paid for participating in this study. If they had to travel to the appointment,
reimbursement of car parking and petrol was offered. There was no clinical relationship with the
subjects, so they did not feel under any obligation to assist with the research in regard to their health
care. The costs of this project were covered by a National Health and Medical Research Council
Strategic Award for which Associate Professor Vivienne Moore, University of Adelaide, is Chief
Investigator A.

3.3 Data management

Data collected from the Generation 1 participants were entered into a Microsoft Office Access 2007
database by three trained data entry staff. De-identified data were stored as numerical responses
and text fields. A set of codes was created to code open-ended text questions. Data files relating to
this round of follow-up were linked to past data sets based on mother and child study ID numbers. A
data dictionary comprising a description of each variable was added to the existing Generation 1
data dictionary.

Prior to data analysis, raw data was thoroughly examined. This preliminary process involved data
cleaning and data screening through preparatory data analysis, in order to detect and correct any
potential errors, observe trends (patterns) within the data, and direct towards appropriate statistical
tests.

Data cleaning, an essential determinant of the validity of any study, was undertaken both during and
after entry, following a structured cleaning framework which involved detecting, diagnosing, and
editing suspected inaccuracies (Van den Broeck et al. 2005). During data entry, valid ranges were
107
Methodology

checked using ‘input edits’, known as an error prevention process (e.g., 2007 was not accepted for
an interview date at this follow-up, as all interviews took place in 2008-2010). After entry, data
cleaning was conducted by a statistician who performed a random 10% check of the first-stage
dataset (in which 10% of the observations were completely checked against the raw ‘paper’ data).
This procedure has been recognised as a good alternative to the double entry method (in which two
datasets with raw data are entered separately by two individuals and then compared for differences)
when resources are limited (Pryjmachuk and Richards 2007). It was particularly useful for identifying
erroneous inliers (incorrect data points falling within the expected range), which would have
otherwise escaped detection. If problem variables were detected, the checking for those variables
was extended to the full dataset.

Subsequently, frequency analysis was performed. Through this process values for each of the
variables were checked against the expected range. This step was particularly important for data
that had not been entered manually (e.g., blood test results). Discrepancies within a participant’s
characteristics were also sought and temporal consistency of data was checked (e.g., child’s height
and weight at 9-10 years was compared with their growth trajectory, using previously collected data
from 2, 3.5 and 5.5 years of age). Although some children were found to be shorter/taller, or
lighter/heavier than expected based on previous measurements, this was within physiological
possibility and all anthropometric measurements were judged to be accurate.

Preparatory data analysis consisted of examination of variables distributions and identification of


missing and outlying data, two common problems in epidemiological research.

Missing data was assessed in terms of extent (percentage of cases with missing data on a certain
variable) and pattern (missing completely at random, missing at random, and missing not at
random). Although there are no guidelines for what represents excessive missing data, it was
suggested that up to 10% missing data on a given variable would be acceptable (Fox-Wasylyshyn
and El-Masri 2005). Two aspects of missing data were encountered in this study: missing cases
(i.e., cohort members who did not participate in follow-up) and missing data items (i.e., certain data
not recorded for participants who had other data available). In order to determine the pattern of
missing cases and assess sample representativeness, participants were compared with non-
participants in terms of baseline characteristics, using independent samples t-tests (for continuous
variables) and chi-squared tests (for categorical variables).

108
Methodology

A series of methods for dealing with missing data have been proposed, such as complete case
analysis (exclusion of incomplete records, with some loss of statistical power), overall mean
substitution (replacement of the missing value with the average in the dataset), missing indicator
method (inclusion of a newly created indicator variable on missing in multivariable analysis along
with the original variable), or imputation (prediction of the missing value using other characteristics of
the subject) (Donders et al. 2006). Among these techniques, imputation has the advantage of
yielding the least biased results, for both missing at random and missing not at random data
(Donders et al. 2006). In this study, single and multiple imputation techniques were applied only for
two of the exposure variables (i.e., maternal pre-pregnancy weight, maternal weight just before
delivery) and are explained in greater detail in Section 3.4.1. Imputation techniques were not used
for outcome variables.

Outliers were defined in this study as observations that fell 1.5 times the interquartile range below
the first quartile or above the third quartile (Moore and McCabe 2003). They were assessed in terms
of leverage (deviation from the mean value) and influence (property of an observation whose
inclusion or exclusion from the analysis would be followed by major changes in the fitted models).
Although dealing with outliers is considered a matter of individual judgement (Moore and McCabe
2003), it has been suggested that extreme values on both exposure and outcome variables would be
more concerning than those on solely one of these categories, and should be discarded (Pryjmachuk
and Richards 2007). A similar approach has been recommended for true extreme values resulted
from unanticipated external processes (Pryjmachuk and Richards 2007) (e.g., not fasting when
fasting blood sample was required from children). My approach was to retain outliers wherever
possible, as they generally appeared to be genuine data points.

3.4 Defining variables of interest

3.4.1 Intrauterine exposures

Maternal pre-pregnancy BMI

Maternal pre-pregnancy nutritional status was measured by pre-pregnancy BMI (weight in kg divided
by height in m squared). This was calculated based on pre-pregnancy weight (self-reported) and
height (measured by the research staff at the interview in early pregnancy, which was considered
equal to the pre-pregnancy height).
109
Methodology

Maternal pre-pregnancy weight was self-reported at the time of the first interview, in early pregnancy.
It was collected as a response to the question: “How much did you weigh before pregnancy?”
followed by interviewer’s probe for more accuracy if a rounded weight was given as an answer.
However, there was no specification regarding woman’s clothing, time of the day or fasting status
when the measurement had been done, all these aspects potentially contributing to the inaccuracy of
self-reported pre-pregnancy weight.

Despite the wide acceptability of self-reported weight data in epidemiological studies (Spencer et al.
2002), it is also known that there is a great degree of discordance between subjective and objective
weight measures, with an overall tendency to underestimate it by self-report, especially in overweight
or obese individuals (Gorber et al. 2007; Rowland 1990). For this reason, the agreement between
self-reported pre-pregnancy weight and directly measured early pregnancy weight required
assessment in the Generation 1 women.

A majority of the women (n=502) in the Generation 1 cohort reported their pre-pregnancy weight. Of
them, 21 women reported a pre-pregnancy weight that seemed unlikely when compared to the
standardised estimates that corrected for weight gain during early pregnancy (these women either
lost more than 0.37 kg/week or gained more than 0.80 kg/week of gestation before the first interview
in early pregnancy). These 21 incongruent pre-pregnancy weight values were set to missing, and
were then imputed along with the other 55 missing values in the women who did not report their pre-
pregnancy weight. Both single and multiple imputations were performed. With single imputation,
pre-pregnancy weight was estimated from a linear regression against maternal measured early
pregnancy weight, height and age, which, based on the women without missing data, explained 90%
of the variance in predicted variable; therefore, imputation was deemed appropriate. Subsequently,
the multiply imputed pre-pregnancy weight was based on a regression of the singularly imputed pre-
pregnancy weight, early pregnancy weight and height, and maternal age. The average of the means
of pre-pregnancy weight from the five imputed data sets, the average standard deviation, minimum
and maximum values were identical to the singularly imputed variable. As a result, the singularly
imputed pre-pregnancy weight was used in all analyses.

Maternal glucose tolerance status during pregnancy

The definition of gestational glucose tolerance status in Generation 1 women was based on the
Australasian Diabetes in Pregnancy Society (ADIPS) guidelines (Hoffman et al. 1998). These
guidelines, which were made available for routine use just before enrolment of pregnant women for

110
Methodology

the Generation 1 study (1998-2000), recommended universal screening for GD at 26-28 weeks of
gestation (Hoffman et al. 1998). According to these criteria, the non-fasting oral glucose challenge
test (OGCT) was considered positive if 1 hour plasma glucose level was ≥ 7.8 mmol/l after a 50 g
glucose load (Hoffman et al. 1998). Pregnant women with a positive challenge test were further
referred for a fasting oral glucose tolerance test (OGTT). The cut-off points recommended by ADIPS
for GD were fasting plasma glucose level of ≥ 5.5 mmol/l and/or plasma glucose level of ≥ 8.0
mmol/l at 2 hours post-75 g glucose load (Hoffman et al. 1998); this definition was imposed by the
health service attended by the women and were not at the discretion of the researchers. Women
with negative OGTT following a positive OGCT were included in the group of borderline gestational
glucose intolerance, for the purpose of this study. There was no possibility of using a continuous
form of plasma glucose levels at OGTT due to the relatively small number of women with data
available from this test.

In the Generation 1 Study, OGCT and OGTT results of pregnant women were abstracted from the
antenatal records. OGCT was performed at about 28 weeks of gestation (range 23-30 weeks), and
OGTT at 29.5 weeks (range 20-36 weeks). Of the 557 women who completed the pregnancy phase,
535 women (96.1%) had available information regarding their glucose tolerance status; the other 22
women were excluded from the analyses involving this variable. Imputation was not used for missing
data since there is no known equation for predicting glucose tolerance status with acceptable
reliability on the basis of other non-metabolic data.

Maternal gestational weight gain

Gestational weight gain (kg) was calculated by subtracting pre-pregnancy weight from maternal
weight shortly before delivery. Measurement of maternal weight was part of the study protocol for
the second interview (which was intended to occur at 30-34 weeks of gestation). For some women
the second interview weight thus occurred several weeks before delivery. Of the 557 women who
gave consent to be part of the cohort, 550 (98.7%) had available data on measured weight in the last
trimester of pregnancy taken during the late pregnancy interview (between 27.3 and 40.6 weeks of
gestation) (Table 10). The seven women without a third trimester measure of weight were excluded
from the analysis of gestational weight gain as their weight shortly before delivery could not be
accurately predicted.

111
Methodology

Table 10. Timing of last available measure of weight during pregnancy

Time during pregnancy Number of women (%)


3rd trimester 550 (98.7%)
Within 7 days pre-delivery 41 (7.36%)
8-14 days pre-delivery 41 (7.36%)
15-21 days pre-delivery 37 (6.64%)
22+ days pre-delivery 431 (77.4%)
2nd trimester 5 (0.90%)
1st trimester 2 (0.36%)

For the 550 women with weight measured in the third trimester of pregnancy, mean (SD) gestational
age at delivery was 39.4 (1.53) weeks, while mean (SD) gestational age when maternal weight was
last recorded was 33.4 (2.24) weeks. Given this difference in gestational age (mean of 6 weeks), it
was considered that the last recorded maternal weight was unlikely to be representative of pre-
delivery weight. Therefore, maternal weight shortly before delivery was derived based on a linear
regression against last recorded maternal weight (in the third trimester) and the number of weeks
between last measurement and delivery, assuming a weekly gestational weight gain in the last
trimester of pregnancy of 0.36 kg/week (Hytten 1991; Widschut 2006).

Total GWG was calculated by subtracting maternal self-reported pre-pregnancy weight (using
imputed data as described in the previous section) from the derived maternal weight shortly before
delivery. Weekly GWG was calculated as total GWG divided by gestational age at delivery. Eleven
women had outlying mean weekly weight changes compared to the other Generation 1 women (they
gained either less than 0.055 kg/week or more than 0.655 kg/week across the entire pregnancy).
These improbable weight changes were excluded from the analysis.

3.4.2 Outcomes in children at the age of 9-10 years

The outcomes of interest in children were continuous measures of child global obesity, percentage
body fat, fat pattern and insulin resistance. We chose to employ continuous measures as we were
interested in the entire spectrum of these outcome variables in relation to maternal pre-pregnancy
BMI, glucose tolerance status during pregnancy and gestational weight gain.

112
Methodology

Child BMI z-score

Child BMI was used as a global measure of obesity or overweight status. It was calculated as weight
(kg) divided by squared height (m2). BMI was further converted to standard deviation scores (BMI z-
scores) based on the 1990 age- and sex-specific British growth reference curves (Cole et al. 1995),
using the ‘zanthro’ Stata macro (Vidmar et al. 2004). The age- and sex-specific reference curves
were based on the lambda-mu-sigma (LMS) method (Cole 1988). In addition, for descriptive
purposes, children were classified as normal weight, overweight or obese using the International
Obesity Task Force (IOTF) age- and sex-specific BMI cut-offs (Cole et al. 2000), using the Stata
function ‘zbmicat’. Data on BMI and thus BMI z-score was available in all 443 children who
consented to participate in the 9-10 year follow-up.

Child percentage body fat

Child %BF (including both subcutaneous and visceral fat) was regarded as a more specific measure
of obesity and was estimated using BIA (Tanita TBF-300). Data on %BF was available for 442
children (one child had a weight of 94 kg, which was greater than the value accepted as valid for this
age by the analyser and therefore could not have %BF estimated by BIA).

Child waist-to-height ratio

Fat distribution in children was derived from the WHtR, as a continuous variable. Data on waist and
height measurements were available for 442 children (one participant declined waist circumference
measurement).

Child insulin resistance

Child IR was quantified by HOMA-IR, calculated as the product of fasting plasma insulin (μU/ml) and
fasting plasma glucose (mmol/l) divided by 22.5 (Matthews et al. 1985). Data were available for 164
children (37% of participants at the 9-10 year follow-up) who gave consent for a fasting blood sample
collection. One of these children had extremely high values for plasma glucose and insulin, possibly
due to non-fasting, and these values were excluded from the analyses. Therefore, the analyses
were based on 163 participants with complete data on fasting plasma insulin and glucose levels.
Missing data for children who did not have a blood test are excluded from the relevant analyses.

113
Methodology

3.4.3 Potential confounders

It is well known that simple associations between exposures and outcomes can be in fact falsely
overestimated or, less frequently, obscured (MacKinnon et al. 2000) by the effect of a third variable.
This third variable can be either a confounder (if it is associated with both the exposure and the
outcome, without being a consequence of the exposure (Hennekens and Buring 1987)) (Figure 2), or
a mediator (if the variable is on the causal pathway). In order to prevent distortion of the true
association of interest and maintain validity of inferences, potential third variable effects need to be
accounted for in the analysis of any causal process. Confounding and mediating factors are treated
identically in statistical analysis (MacKinnon et al. 2000) and the distinction between them is based
on accumulated knowledge about the relationship of interest.

Figure 2. Confounders in relation to exposures and outcomes

relationship
Exposure Outcome
of interest

Confounder

There is limited scientific literature on potential confounding effects in the relationships of interest for
this project. Several factors have been suggested as being associated on one hand with an increased
risk of pre-pregnancy obesity, gestational glucose intolerance (detailed in Section 2.2.2.4), and/or
excessive GWG (detailed in Section 2.2.3.1), and on the other hand with an increased risk of
childhood obesity and insulin resistance. Of these factors, potential confounders considered in this
thesis were maternal age at the time of pregnancy, parity, smoking during pregnancy, presence of
pregnancy-induced hypertension, and maternal highest level of education completed at the time of
pregnancy (described below). In addition, each main exposure was included in turn as a potential
confounder for the relationships between the other two predictor variables and the outcomes of
interest. Child current BMI z-score was considered as a potential pathway variable and adjusted for
in the models with child HOMA-IR as an outcome. In this thesis, multivariate analysis was performed
to investigate and adjust for potential confounding and mediating effects.

Maternal age at the time of pregnancy was obtained by subtracting the woman’s date of birth from
the child’s date of birth (both abstracted from antenatal records) and dividing this difference by
365.25; it was treated as a continuous variable. Parity at the time of study pregnancy was defined

114
Methodology

as a binary variable (primiparity or multiparity), derived from the answer to a question regarding the
number of children previously delivered. Maternal smoking was considered as a binary variable
indicating if the mother had reported (at the early or late pregnancy interview) any smoking during
the study pregnancy (yes or no). Data on pregnancy-induced hypertension were abstracted from
antenatal casenotes and referred to any degree of high blood pressure with onset during pregnancy,
including preeclampsia (treated as a binary variable). Maternal education was chosen as a
surrogate marker for socio-economic status and was defined by the highest level of education
completed by the mother at the time of pregnancy: partial high-school, high-school, Training and
Further Education (TAFE) or college, and university degree.

Of note, factors such as child’s size at birth, breastfeeding, diet or physical activity have also been
associated both with maternal pre-pregnancy obesity, GD or GWG on one hand, and child obesity on
the other, but these variables are likely to be on the causal pathway between maternal exposures
and child outcomes of interest. Given that they are mediators (not confounders) and their potential
influence was not within the scope of this thesis, these variables were not included in analyses.

3.5 Analysis plan

Statistical analyses included in this thesis covered both descriptive and inferential aspects.
Descriptive statistics focused on summary measures for all variables considered in this project
(exposures, outcomes and potential confounders), as well as on the non-participation assessment in
relation to the original Generation 1 cohort. For descriptive purposes, summary measures are
reported as means, standard deviations (SD), medians, ranges, and interquartile ranges (IQR) for
continuous variables (i.e., maternal pre-pregnancy BMI, maternal plasma glucose levels at OGCT
and OGTT, maternal GWG, maternal age at the time of pregnancy, child BMI z-score, child %BF,
child WHtR, and child HOMA-IR), and as numbers and proportions for categorical variables (i.e.,
maternal glucose tolerance status during pregnancy, categories of maternal pre-pregnancy BMI,
categories of maternal GWG, parity, smoking, pregnancy-induced hypertension, and level of
educational attainment at the time of pregnancy).

In practice, different summary measures of central tendency and variability are relevant according to
the distribution of the variable of interest: mean and SD for normally distributed variables and median
and IQR for variables with a skewed distribution. For the latter type of data, the most appropriate

115
Methodology

graphical representation of the summary measures is the box and whisker plot (an example is further
presented in Chapter 4, Figure 6). The box contains 50% of the data, the middle line is the median,
the bottom and the top lines of the box are the 1st and 3rd quartiles (IQR), the whiskers extend to 1.5
of IQR and the dots represent the outliers. Normality was inspected for each outcome variable and a
transformation was sought for those with a skewed distribution (e.g., HOMA-IR). Non-participation
was examined using independent samples t-tests (for continuous variables) and chi-squared tests
(for categorical variables) and is presented in Section 4.1.4.

Inferential statistics followed a component-based approach, with statistical models being


constructed to investigate associations between each main intrauterine exposure (i.e., maternal pre-
pregnancy BMI, categories of glucose tolerance during pregnancy or GWG), and the development of
obesity in terms of BMI z-score, percentage body fat and central adiposity, and IR in children.
Briefly, the main components of the inferential analyses were:

1. maternal pre-pregnancy BMI and child BMI z-score, %BF, WHtR and HOMA-IR;

2. maternal glucose tolerance during pregnancy and child BMI z-score, %BF, WHtR and HOMA-IR;

3. maternal GWG and child BMI z-score, %BF, WHtR and HOMA-IR;

4. interactions between each pair of main exposures in relation to child BMI z-score, %BF, WHtR
and HOMA-IR.

After examining each bivariate relationship (using Pearson’s correlation), several covariates
identified in previous studies as independently predicting the exposures and the outcomes were
tested and adjusted for in each model, in a stepwise manner. The potential confounders controlled
for in this project included maternal age, parity, smoking during pregnancy, pregnancy-induced
hypertension and highest level of education completed by the time of pregnancy. Confounders were
retained in the models if p<0.1. Given the well known positive association between obesity and
insulin resistance, child BMI z-score was further considered as a potential mediating variable on the
pathway between each of the early origin factors addressed in this project and child HOMA-IR.

Potential interactions between each two main exposures (i.e., maternal pre-pregnancy BMI and
glucose tolerance status during pregnancy, maternal pre-pregnancy BMI and GWG, and glucose
tolerance status during pregnancy and GWG) were identified a priori and tested in relation to all
outcomes of interest, with the significant potential confounders included in the models. Two
exposures are considered to interact in relation to a certain outcome when their observed joint effect

116
Methodology

is greater (synergy) or smaller (antagonism) than the expected joint effect based on their
independent effects alone (Greenland 1983).

Models with continuous and (relatively) normally distributed outcomes (i.e., child BMI z-score, %BF
and WHtR) were run using simple and forward stepwise multiple linear regression.

Traditionally, the logarithmic transformation is applied to positively skewed dependent variables,


such as HOMA-IR, in order to obtain normal distributions. However interpreting the results of
analyses based on log transformations raises a number of issues. First, this type of transformation
often leads to symmetry of the dependent variable distribution and does not equate to normality
(Moran et al. 2007). Second, appropriate ‘back transformation’ to the original scale of the dependent
variable involves more than a simple exponentiation. When the dependent variable is log
transformed in an ordinary least squares regression model, the regression coefficients estimate the
expected geometric mean of the original variable and not the arithmetic mean, which would be
estimated using the untransformed dependent variable (and which is of primary interest) (UCLA:
Academic Technology Services and Statistical Consulting Group). In fact, the effect on the original
dependent variable is in terms of percentage change, quantified as 100 * [exp(β) – 1], where β is the
regression coefficient of the independent variable (regardless of whether it is continuous or
categorical) (Cole 2000). The above mentioned limitations of log transformation of a dependent
variable may be overcome by applying generalised linear modes (GLM), which synthesize the
general techniques used for continuous and discrete data into a unified conceptual framework with a
more flexible approach (Breslow 1996). In this thesis analyses with child HOMA-IR as the outcome
of interest were assessed using GLM with log link function and Gaussian family.

Assumptions underpinning each statistical analysis were tested and appeared to be met. Given that
the power was fixed by the sample size (established cohort), no power calculation was performed.
Throughout this thesis, the cut-off for statistical significance was set at 0.05, except for the criterion
of retaining confounders, in which case the cut-off was 0.1. Simultaneously testing multiple
hypotheses on a single dataset may inflate the overall type I error (the probability of the observed
associations being attributable to chance), thus increasing the likelihood of a false-positive
conclusion (Schulz and Grimes 2005). To minimise this type I error, adjustments for multiple
comparisons have been suggested, which usually involve multiplying the p-value of the observed
association by the number of hypotheses tested (Schulz and Grimes 2005), thereby reducing the
probability of mistakenly rejecting a true null hypothesis. However, randomly observed associations
are still possible (Rothman 1990), albeit less likely. Moreover, these random associations in nature,

117
Methodology

as opposed to simple random number sets, may be significant and are worthy of further investigation
(Rothman 1990). Therefore, adjustment for multiple comparisons was considered unnecessary and
not performed in the current study. All data analyses were performed using Stata IC version 10
statistical software (StataCorp, Inc., Texas USA).

3.6 Summary

This chapter outlined a brief history of the Generation 1 cohort, a detailed description of the methods
of data collection for the 9-10 year wave, along with the methods of analysis for each component.
The next chapter presents the results of statistical analyses.

118
Chapter 4 Results

This chapter presents the main findings of statistical analyses conducted in order to investigate
potential associations between three maternal pregnancy-related factors likely to contribute to fetal
overnutrition (maternal obesity prior to pregnancy, gestational glucose intolerance and excessive
gestational weight gain), on one hand, and child obesity and insulin resistance, on the other.

4.1 Descriptive statistics

This first section of results focuses on the description of variables of interest for this project, including
the intrauterine exposures (Section 4.1.1), the outcomes in children (Section 4.1.2), and the potential
confounders identified from previous research (Section 4.1.3). Sample representativeness at this
follow-up in relation to the original study group (which was considered representative of the wider
population of South Australian women who gave birth in 1998-2000) is presented in Section 4.1.4.

4.1.1 Intrauterine exposures

The three main exposures for this study were maternal pre-pregnancy BMI, glucose tolerance status
during pregnancy, and gestational weight gain; their summary measures are presented below.

4.1.1.1 Maternal pre-pregnancy BMI

As described in Section 3.4.1, maternal nutritional status prior to pregnancy was measured by BMI,
using self-reported pre-pregnancy weight and height measured in early pregnancy. Pre-pregnancy
weight was reported by 502 women (90.1%). Given the well known tendency of women to under-
report their weight (Gorber et al. 2007), self-reported pre-pregnancy weight was compared to the
directly measured weight in early pregnancy. Early pregnancy weight was measured between 7.7
and 20.4 weeks of gestation (mean (SD) for gestational age 14.4 (2.0) weeks) and was available for

119
Results

all 557 women. Overall, there was a very strong correlation between self-reported pre-pregnancy
weight and weight measured in early pregnancy (r=0.96, p<0.001) (Figure 3).

Figure 3. Correlation between maternal self-reported pre-pregnancy weight and measured weight in
early pregnancy
120
Measured weight in early pregnancy (kg)
60 80 40 100

40 60 80 100 120
Self-reported pre-pregnancy weight (kg)

However, for 21 women, the difference between measured weight in early pregnancy and self-
reported pre-pregnancy weight for gestational age fulfilled the criteria for outliers within Generation 1
cohort (these women either lost more than 0.37 kg/week or gained more than 0.80 kg/week of
gestation, assuming a constant rate of weight gain in early pregnancy). In order to maximize the
dataset, the 21 improbable values of pre-pregnancy weight (likely due to poor recall or reporting) were
set to missing and then imputed along with the other 55 missing values (as detailed in Section 3.4.1).
Anthropometric measurements of Generation 1 women before pregnancy are presented in Table 11.

Table 11. Maternal body size before pregnancy (n=557)

Pre-pregnancy measure Mean SD Range Median IQR

Weight (kg) 66.9 15.0 40.0 – 126.0 63.0 56.0 - 74.6

Height (cm) 163.6 6.16 146.5 – 185.0 163.5 159.5 - 167.6

BMI (kg/m2) 25.0 5.59 14.7 – 46.6 23.6 21.2 - 27.5


IQR – interquartile range, SD – standard deviation

Based on WHO classification of BMI (World Health Organization 1995), 39.3% of Generation 1
women were overweight or obese prior to pregnancy (Table 12 and Figure 4).

120
Results

Table 12. Maternal pre-pregnancy BMI status

Categories of pre-pregnancy BMI Number (%) of women

Underweight (<18.5 kg/m2) 27 (4.85)

Normal weight (18.5-24.9 kg/m2) 311 (55.8)

Overweight (25-29.9 kg/m2) 130 (23.3)

Obese (≥30 kg/m2) 89 (16.0)

Total 557 (100)

Figure 4. Maternal pre-pregnancy BMI status

27 (4.9%)
89 (16%)

130 (23.3%)

311 (55.8%)

underweight (<18.5 kg/m2) normal weight (18.5-24.9 kg/m2)


overweight (25-29.9 kg/m2) obese (>= 30 kg/m2)

4.1.1.2 Maternal glucose tolerance status during pregnancy

As described in Section 3.4.1, the guidelines proposed by the Australasian Diabetes in Pregnancy
Society were used to classify maternal glucose tolerance status during pregnancy. Briefly, the
screening for GD was a two stage process, comprising a non-fasting OGCT which, if positive, was
followed by a fasting OGTT, which is a diagnostic test. Women with normal OGTT following an
abnormal OGCT were considered to have BGGI. The plasma glucose levels from the OGCT and
OGTT are summarised in Table 13.

121
Results

Table 13. Summary of plasma glucose levels at OGCT and OGTT during pregnancy for Generation 1
women

Test Number (%) of Plasma glucose level (mmol/l)


(oral cohort members
glucose who underwent
load) the test Mean SD Range Median IQR

OGCT 538 (96.6%) 1 h post- load 6.3 1.8 2.8-21.6 6.2 5.2-7.2
(50 g)
OGTT 90 (16.2%) Fasting 4.7 0.6 3.5-6.9 4.7 4.4-5
(75 g)
2 h post-load 7.0 1.3 4.5-10.8 6.9 6.1-8
IQR – interquartile range, OGCT – oral glucose challenge test, OGTT – oral glucose tolerance test, SD – standard deviation

OGCT was performed in 538 women (96.6% of pregnant women enrolled in the Generation 1 cohort)
as part of their standard care (universal screening for GD was introduced in 1998, when recruitment
for the cohort started). Of the 538 women who underwent OGCT, 452 women (81.2%) had normal
plasma glucose level (<7.8mmol/l) and 86 women (15.4%) had a positive screening test (≥7.8 mmol/).

Of the 86 women with an abnormal OGCT, 80 (93.0%) had OGTT done. This test was positive in 27
women (fasting plasma glucose level >5.5 mmol/l and/or 2h-post 75 oral glucose load >8.0 mmol/l),
who were thus diagnosed with gestational diabetes; OGTT was negative in 53 women who, for the
purpose of this project, were included into the borderline gestational glucose intolerance category.
For the six women with abnormal OGCT and without OGTT available results, a degree of glucose
intolerance could be suspected (either borderline gestational glucose intolerance, or gestational
diabetes), but it was not possible to clearly define their glucose tolerance status. These women were
not included in further analyses regarding this exposure.

In addition to the 80 women undergoing an OGTT following an abnormal OGCT, six women had
OGTT done without being preceded by the screening test; OGTT was positive in three of them, hence
they were classified as GD, while the remaining three were not classifiable and were excluded from
further analyses regarding this exposure. An additional four women underwent OGTT despite having
normal OGCT, which was not common practice; OGTT was positive in one of them and this woman
was classified as GD, while the remaining three were classified as having normal glucose tolerance.

In summary (Table 14 and Figure 5), 31 women (5.57%) of the Generation 1 cohort were diagnosed
with GD, 53 women (9.52%) were included into the BGGI category, 451 women (81%) were
considered as having normal glucose tolerance, and for 22 women (3.95%) glucose tolerance status
could not be determined.
122
Results

Table 14. Glucose tolerance profiles during pregnancy in Generation 1 women

Number of
OGCT OGTT Glucose tolerance category Criteria or assumptions
women

Normal 452

Normal 3 NGT OGTT is a diagnostic test

Abnormal 1 GD OGTT is a diagnostic test

Missing 448 NGT Assume OGTT not recommended

Abnormal 86

Normal 53 BGGI

Abnormal 27 GD ADIPS

Missing 6 Non-classifiable (BGGI or GD)

Missing 19

Normal 3 Non-classifiable (NGT or BGGI)

Abnormal 3 GD OGTT is a diagnostic test

Missing 13 Non-classifiable

Total 557
ADIPS - Australasian Diabetes in Pregnancy Society, BGGI - borderline gestational glucose intolerance, GD - gestational diabetes,
NGT - normal glucose tolerance, OGCT- oral glucose challenge test, OGTT - oral glucose tolerance test

123
Results

Figure 5. Results of prenatal screening for gestational diabetes in Generation 1 women

557 – enrolled and


completed the pregnancy phase

19 – OGCT results
not available (3.4%)

538 – OGCT results


available (96.6%)

452 – normal OGCT 86 – abnormal OGCT


(81.1%) (15.4%)

448 – OGTT results 6 – OGTT results 13 –OGTT results


not available, as test not available (1.08%) not available (2.33%)
not required (80.4%)

4 – OGTT results 80 – OGTT results 6 – OGTT results


available (0.72%) available (14.4%) available (1.08%)

1 – abnormal OGTT 27 – abnormal OGTT 3 – abnormal OGTT


(0.18%) (4.85%) (0.54%)

3 – normal OGTT 53 – normal OGTT 3 – normal OGTT


(0.54%) (9.52%) (0.54%)

Normal glucose tolerance


Borderline gestational glucose intolerance
Gestational diabetes
Non-classifiable – excluded from analysis

124
Results

4.1.1.3 Maternal gestational weight gain

As described in Methods Section 3.4.1, gestational weight gain was calculated by subtracting
maternal self-reported pre-pregnancy weight (using reliable and imputed data) from maternal weight
shortly before delivery. After excluding 22 (3.23%) outliers, mean (SD) total GWG in Generation 1
women (of those with weight available during the third trimester of pregnancy) was 14.0 (4.54) kg
(range 2.29 to 26.4 kg); the mean (SD) weekly gestational weight gain was 0.35 (0.11) kg/week
(range 0.06 to 0.65 kg/week).

Across all categories of pre-pregnancy BMI, about a third of the women gained weight adequately
throughout pregnancy and almost a half gained weight excessively compared to the 2009
recommendations of Institute of Medicine (Rasmussen and Yaktine 2009) (Table 15). The highest
proportion (over 60%) of excessive GWG was observed in overweight and obese women.

Table 15. Total gestational weight gain across categories of pre-pregnancy BMI in Generation 1
women, relative to 2009 Institute of Medicine (IOM) recommendations

IOM recommendations 2009 for GWG


Pre-pregnancy BMI category n Insufficient GWG Adequate GWG Excessive GWG
n (%) n (%) n (%)

Underweight (<18.5 kg/m2) <12.7 kg 12.7 – 18.2 kg >18.2 kg

26 10 (38.5%) 10 (38.5%) 6 (23.0%)

Normal weight (18.5-24.9 kg/m2) <11.4 kg 11.4 – 15.9 kg >15.9 kg

304 57 (18.8%) 128 (42.1%) 119 (39.1%)

Overweight (25-29.9 kg/m2) <6.8 kg 6.8 – 11.4 kg >11.4 kg

127 8 (6.3%) 32 (25.2%) 87 (68.5%)

Obese (≥30 kg/m2) <5 kg 5 – 9.1 kg >9.1 kg

82 12 (14.6%) 18 (22.0%) 52 (63.4%)

Total 539 87 (16.1%) 188 (34.9%) 264 (49.0%)

125
Results

4.1.1.4 Interrelations between maternal pre-pregnancy BMI, glucose tolerance status during
pregnancy and gestational weight gain

Previous research (outlined in Section 2.2.4) has shown that the three exposures of interest for this
project are interrelated mainly through the amount of glucose provided to the fetus, with subsequent
impact on intrauterine growth. The interrelations observed between maternal pre-pregnancy BMI,
glucose tolerance status during pregnancy and GWG in Generation 1 women are reported below.
They are further examined as potentially interacting with each other in relation to child obesity and
insulin resistance, analyses that are presented in Section 4.2.4.

Maternal pre-pregnancy BMI and glucose tolerance status during pregnancy

As illustrated in Table 16 and Figure 6, pre-pregnancy BMI (mean and median, respectively) was
lowest among women who maintained normal glucose tolerance during pregnancy and highest
among those who developed gestational diabetes. The fact that obese women are at increased risk
of glucose intolerance during pregnancy is well established; however, not only overweight or obese
women develop this pregnancy-related condition. In this cohort, a quarter of the women with GD and
45.3% of those with BGGI were not overweight or obese. Even underweight women presented
glucose tolerance across the entire spectrum, albeit only one developed GD.

Table 16. Maternal pre-pregnancy BMI across the spectrum of glucose tolerance during pregnancy

Glucose Normal
Mean (SD) Underweight Overweight Obese
tolerance weight
pre-pregnancy
status during n n (%) n (%) n (%)
BMI (kg/m2) n (%)
pregnancy

NGT 451 24.5 (5.08) 22 (4.88) 271 (60.1) 98 (21.7) 60 (13.3)

BGGI 53 27.1 (7.02) 3 (5.66) 21 (39.6) 15 (28.3) 14 (26.4)

GD 31 29.2 (6.84) 1 (3.23) 7 (22.6) 12 (38.7) 11 (35.5)

Total 535 25.0 (5.56) 26 (4.86) 299 (55.9) 125 (23.4) 85 (15.9)
BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

126
Results

Figure 6. Maternal pre-pregnancy BMI across the spectrum of glucose tolerance during pregnancy

50 40
Maternal pre-pregnancy BMI
20 30 10

NGT BGGI GD

BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

Maternal pre-pregnancy BMI and gestational weight gain

Similar to other studies (Cedergren 2006; Schack-Nielsen et al. 2010), there was a moderate
negative correlation between pre-pregnancy BMI and GWG (r=-0.32, p<0.001). This relationship is
presented in Figure 7.

Figure 7. Maternal gestational weight gain in relation to pre-pregnancy BMI


25
Total gestational weight gain (kg)
5 10 0 15 20

10 20 30 40 50
Pre-pregnancy BMI (kg/m2)

127
Results

As presented in Table 17 and Figure 8, underweight women gained the greatest amount of weight
overall during pregnancy, while obese women gained the least, approximately 3.6 kg less compared
to normal weight women.

Table 17. Maternal gestational weight gain across categories of pre-pregnancy BMI

Pre-pregnancy BMI category n Mean (SD) total GWG (kg)

Underweight (<18.5 kg/m2) 26 15.0 (4.51)

Normal weight (18.5-24.9 kg/m2) 304 14.8 (4.00)

Overweight (25-29.9 kg/m2) 127 13.5 (4.63)

Obese (≥30 kg/m2) 82 11.2 (5.10)

Total 539 14.0 (4.54)

Figure 8. Maternal gestational weight gain across categories of pre-pregnancy BMI


25
Total gestational weight gain (kg)
5 10 0 15 20

underweight normal weight overweight obese

Maternal glucose tolerance during pregnancy and gestational weight gain

Overall, women who developed GD gained a smaller amount of weight during pregnancy (up to
delivery) compared to women who maintained normal glucose tolerance during pregnancy (which
aligns with the fact that they were likely to be, though not exclusively, overweight or obese).
Maternal GWG across the spectrum of glucose tolerance during pregnancy is summarised in Table 18
and Figure 9.

128
Results

Table 18. Maternal gestational weight gain and glucose tolerance status during pregnancy

Glucose tolerance profiles during pregnancy n Mean (SD) total GWG (kg)

NGT 439 14.1 (4.39)

BGGI 49 13.8 (4.58)

GD 31 12.8 (5.42)

Total 519 14.0 (4.76)

BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, GWG – gestational weight gain, NGT - normal glucose
tolerance

Figure 9. Maternal gestational weight gain and glucose tolerance status during pregnancy
25
Total gestational weight gain (kg)
5 10 0 15 20

NGT BGGI GD

BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

4.1.2 Outcomes in children at the age of 9-10 years

The mean (SD) age of children taking part in this phase of the Generation 1 study was 9.65 (0.31)
years. The outcomes considered in this project are child global obesity (measured by BMI z-score),
percentage body fat (estimated by BIA), central adiposity (derived from the WHtR), and insulin
resistance (defined by HOMA-IR). Their definitions were presented in Section 3.4.2.

129
Results

4.1.2.1 Child BMI z-score

Child BMI was considered as a global indicator of obesity or overweight status. Summary measures
of children’s weight, height, and BMI are outlined in Table 19. Based on the age- and sex-specific
growth reference curves adopted by the International Obesity Task Force (Cole et al. 1995), BMI can
be expressed both as a continuous z-score (the distribution is presented in Figure 10) and as a
three-category variable (normal weight, overweight and obese). According to the IOTF classification,
341 (77.0%) Generation 1 children were normal weight, 76 (17.1%) were overweight and 26 (5.9%)
were obese (Figure 11).

Table 19. Anthropometric measurements in Generation 1 children at 9-10 years (n=443)

Measure Mean SD Range Median IQR

Weight (kg) 34.4 8.27 21.9 - 94.0 32.8 28.7 - 38.1

Height (cm) 138.0 6.46 121.1 - 157.4 137.8 133.1 - 142.0

BMI (kg/m2) 17.9 3.17 12.8 - 37.9 17.2 15.8 - 19.3

BMI z-score 0.41 1.17 -3.07 - 3.99 0.34 -0.41 - 1.18


IQR – interquartile range, SD – standard deviation

Figure 10. BMI z-score in Generation 1 children


0.4
0.3
Density
0.2
0.1
0

-4 -2 0 2 4
Child BMI z-score

130
Results

Figure 11. Distribution of BMI categories in Generation 1 children

26 (5.9%)

76 (17.1%)

341 (77.0%)

Normal weight Overweight Obese

4.1.2.2 Child percentage body fat

Percentage body fat was estimated by bioelectrical impedance analysis in 439 children (99.1% of
those participating at the 9-10 year follow-up). Summary measures for percentage body fat in
Generation 1 children are presented in Table 20.

Table 20. Summary measures for percentage body fat in Generation 1 children (n=439)

Measure Mean SD Range Median IQR

Percentage body fat 19.8% 7.87% 1.4-49.5% 18.2% 14-24.7%


IQR – interquartile range, SD – standard deviation

Percentage body fat distribution was slightly skewed to the right (the graph labelled ‘identity’ in
Figure 12). Of the transformation options generated by the ‘ladder’ Stata command, the square root
of percentage body fat provided the best approximation to a normal distribution (Figure 12).
However, after considering the options, for a simpler and more intuitive interpretation of the results,
identity function was kept (no transformation was used) in the analyses involving this outcome,
particularly since all assumptions of linear regression (Berry and Feldman 1985) using the
untransformed variable were largely met.

131
Results

Figure 12. Histograms by transformation for child percentage body fat

cubic square identity


1.0e-04 .08
.002
8.0e-05
.0015 .06
6.0e-05
.001 .04
4.0e-05
5.0e-04 .02
2.0e-05

0 0 0
0 50000 100000 150000 0 500 1000 1500 2000 2500 0 10 20 30 40 50

sqrt log 1/sqrt


1 8
.5
.8
.4 6
Density

.3 .6
4
.2 .4
2
.1 .2

0 0 0
0 2 4 6 8 0 1 2 3 4 -.8 -.6 -.4 -.2 0

inverse 1/square 1/cubic


15 50 80

40
60
10
30
40
20
5
20
10

0 0 0
-.8 -.6 -.4 -.2 0 -.5 -.4 -.3 -.2 -.1 0 -.4 -.3 -.2 -.1 0

Child percentage body fat

4.1.2.3 Child waist-to-height ratio

Waist-to-height ratio was used as a proxy measure for central adiposity. Waist circumference was
measured in 442 children (99.8% of those participating in this round of follow-up). Summary
measures for waist circumference and WHtR in Generation 1 children are presented in Table 21.

Table 21. Summary measures of central adiposity in Generation 1 children (n=442)

Measure Mean SD Range Median IQR

Waist circumference (cm) 64.9 7.62 52-110.3 63.0 59.5-68.8

Waist-to-height ratio 0.470 0.047 0.390-0.700 0.460 0.436-0.492


IQR – interquartile range, SD – standard deviation

Similar to percentage body fat, WHtR was positively skewed (skewness coefficient = 1.40). In this
case, the transformation consisting of the inverse of the square root technically provided the best
approximation of normality, but did not represent a great improvement over identity (i.e. untransformed)
(Figure 13). Again, assumptions of linear regression (Berry and Feldman 1985) using the
untransformed variable were largely met, except for a minimal violation of normality of the residuals,
but, given the sample size, this was unlikely to affect the estimation of the regression coefficients. In
132
Results

order to facilitate the interpretation of the results, it was decided to keep identity function, particularly
since linear regression modelling is known to be robust against some violations of normality (Berry
and Feldman 1985).

Figure 13. Histograms by transformation for child waist-to-height ratio

cubic square identity


20 15
10
15
10

10
5
5
5

0 0 0
0 .1 .2 .3 .4 .1 .2 .3 .4 .5 .4 .5 .6 .7

sqrt log 1/sqrt


6 8
15
Density

6
4
10
4

5 2
2

0 0 0
.6 .65 .7 .75 .8 .85 -1 -.8 -.6 -.4 -1.6 -1.5 -1.4 -1.3 -1.2

inverse 1/square 1/cubic


2.5 .6 .2

2
.15
.4
1.5
.1
1
.2
.05
.5

0 0 0
-2.5 -2 -1.5 -7 -6 -5 -4 -3 -2 -15 -10 -5 0

Child waist-to-height ratio

One fifth of the Generation 1 children seen at 9-10 years of age had a WHtR greater or equal to 0.5,
a cut-off that has been proposed as a marker of central adiposity (McCarthy and Ashwell 2006).

4.1.2.4 Child insulin resistance

Insulin resistance was defined by HOMA-IR, which was calculated as the product of fasting insulin
(µU/ml) and fasting glucose (mmol/l) divided by 22.5 (Matthews et al. 1985). Six observations of
HOMA-IR fulfilled the criteria for outliers (falling more than 1.5 times interquartile range above the 3rd
quartile or below the 1st quartile). A closer look at these extreme values showed higher fasting
insulin levels in all cases, but no difference to the other Generation 1 children with respect to age,
BMI, fasting glucose, triglyceride, HDL-cholesterol and LDL-cholesterol levels, except for one case
with HOMA-IR of 8.82 whose LDL-cholesterol was in the ‘borderline high’ range (3.83 mmol/l). The
mother of this child also had a pre-pregnancy BMI outside of the overall pattern of the distribution, as
well as BGGI. Only this most extreme outlier for HOMA-IR was dropped from further analyses, while

133
Results

the other outliers were retained. Summary measures for fasting glucose, fasting insulin, and HOMA-
IR after omitting this extreme value are presented in Table 22.

Table 22. Summary measures for fasting glucose, fasting insulin, and HOMA-IR in Generation 1
children (n=163)

Measure Mean SD Range Median IQR

Insulin (µU/ml) 11.3 3.53 3.02-27.69 10.7 8.99-12.8

Glucose (mmol/l) 4.96 0.42 3.50-6.21 4.91 4.70-5.23

HOMA-IR 2.51 0.88 0.61-6.71 2.34 1.92-2.91


IQR – interquartile range, SD – standard deviation

The distribution of HOMA-IR showed marked kurtosis and positive skewness even after excluding the
most extreme outlier (the graph labelled ‘identity’ in Figure 14). Using the ‘ladder’ Stata command, the
log transformation of HOMA-IR provided the best approximation to a normal distribution (Figure 14).

Figure 14. Histograms by transformation for child HOMA-IR

cubic square identity


.04 .15 .8

.03 .6
.1

.02 .4

.05
.01 .2

0 0 0
0 100 200 300 0 10 20 30 40 50 0 2 4 6 8

sqrt log 1/sqrt


2 2 5

4
Density

1.5 1.5
3
1 1
2
.5 .5
1

0 0 0
.5 1 1.5 2 2.5 -.5 0 .5 1 1.5 2 -1.2 -1 -.8 -.6 -.4

inverse 1/square 1/cubic


4 4 5

4
3 3
3
2 2
2
1 1
1

0 0 0
-1.5 -1 -.5 0 -3 -2 -1 0 -4 -3 -2 -1 0

Child HOMA-IR

Traditionally, the logarithmic transformation is applied to positively skewed dependent variables in


order to obtain normal distributions. However interpreting the results of analyses based on log
transformations raises a number of issues, which have been discussed in detail in Section 3.5.
These limitations of log transformation of a dependent variable may be overcome by applying

134
Results

generalised linear modes (GLM), which in the present case (of child HOMA-IR) were applied with log
link function and Gaussian family.

A summary of anthropometric measurements and fasting blood assays of interest is presented in


Table 23.

Table 23. Children’s anthropometric measurements and fasting measures of glucose homeostasis at
9-10 years (mean, SD)

p-value for
All children Girls Boys
girls vs. boys
Age (years) 9.65 (0.31) 9.64 (0.29) 9.65 (0.32) 0.688
Body size n=443 n=221 n=222
Weight (kg) 34. 4 (8.27) 33.5 (6.91) 35.35 (9.36) 0.019
Height (cm) 138.0 (6.46) 137.1 (6.33) 138.9 (6.48) 0.005
BMI (kg/m2) 17.9 (3.17) 17.7 (2.65) 18.1 (3.61) 0.120
BMI-z score* 0.41 (1.17) 0.25 (1.04) 0.57 (1.27) 0.004
Percentage body fat n=439 n=218 n=221
Based on bioelectrical
19.8 (7.87) 20.6 (7.76) 19.0 (7.92) 0.039
impedance analysis
Fat pattern n=442 n=220 n=222
Waist circumference (cm) 64.9 (7.62) 63.4 (6.34) 66.3 (8.47) 0.000
Waist-to-height ratio 0.47 (0.05) 0.46 (0.04) 0.48 (0.05) 0.000
Fasting measures of
n=163 n=89 n=74
glucose homeostasis
Insulin (μU/ml) 11.3 (3.53) 11.5 (3.46) 11.1 (3.64) 0.492
Glucose (mmol/l) 4.96 (0.42) 4.92 (0.43) 5.00 (0.40) 0.226
HOMA-IR index 2.51 (0.88) 2.54 (0.85) 2.49 (0.92) 0.731
* Based on the 1990 British growth reference chart (age- and sex-specific)

4.1.3 Potential confounders

A series of factors have been identified in previous research, albeit often inconsistently, as being
associated with maternal pre-pregnancy obesity, glucose intolerance during pregnancy, and/or
excessive gestational weight gain, and also independently increasing the risk of obesity and related
outcomes in the child. Those variables that met the criteria for potential confounders were described
in Section 3.4.3. They include maternal age at the time of pregnancy, parity, smoking during

135
Results

pregnancy, presence of pregnancy-induced hypertension, and maternal socioeconomic status


(indicated in this study by the highest level of education attained at the time of pregnancy).

Maternal age at the time of study pregnancy ranged between 18.8 and 42.2 years, with a mean (SD)
of 29.8 (5.02) years. Maternal pre-pregnancy BMI and maternal age were not correlated (r=0.005,
p=0.916). Women with impaired glucose tolerance during pregnancy were older than those without
(mean (SD) age in GD group 31.7 (5.49) years, BGGI group 30.9 (4.28) years, NGT group 29.6
(5.05) years; p=0.021). An association approaching significance was found between total GWG and
maternal age (r=-0.075, p=0.081). Maternal age at the time of pregnancy was not associated with
child BMI z-score (r=0.026, p=0.589), %BF (r=0.001, p=0.981), WHtR (r=0.003, p=0.949), or HOMA-
IR (r=-0.044, p=0.575).

Parity at the time of study pregnancy was considered as a dichotomous variable (primiparity or
multiparity). One third (33.6%) of Generation 1 women had not given birth to a child prior to the
study pregnancy. Women with previous births had a significantly higher mean BMI at the beginning
of the current pregnancy (25.5 (5.87) kg/m2) compared to primiparous women (24.0 (4.86) kg/m2)
(p=0.002). There was no statistically significant difference in the likelihood of developing GD
between primiparous (4.40%) and multiparous (6.52%) women (p=0.596). Multiparous women
gained significantly less weight throughout pregnancy (13.4 (4.64) kg) compared to women with no
previous births (15.1 (4.12) kg) (p<0.001). Parity was not associated with child BMI z-score
(p=0.091), %BF (p=0.149), WHtR (p=0.843), or HOMA-IR (p=0.316).

Maternal smoking was reported at the early and late pregnancy interview. A binary variable was
derived to indicate if the mother had reported any smoking during her pregnancy (yes or no). A total
of 119 (21.4%) Generation 1 women reported smoking during pregnancy. Maternal smoking was not
associated with pre-pregnancy BMI (p=0.730), glucose tolerance status during pregnancy (p=0.141),
or GWG (p=0.561). Similarly, maternal smoking during pregnancy was not associated with child BMI
z-score (p=0.450), %BF (p=0.972), WHtR (p=0.409), or HOMA-IR (p=0.222).

Pregnancy-induced hypertension occurred in 63 women (11.3%) of the Generation 1 cohort.


Women who developed this complication had a significantly higher mean (SD) pre-pregnancy BMI
(27.6 (6.89) kg/m2) compared to those who did not (24.7 (5.33) kg/m2) (p<0.001). Pregnancy-
induced hypertension was not associated with glucose tolerance status during pregnancy (p=0.283),
or GWG (p=0.833). Pregnancy-induced hypertension was not associated with child BMI z-score
(p=0.828), %BF (p=0.618), WHtR (p=0.538), or HOMA-IR (p=0.345).

136
Results

Maternal education was chosen as a surrogate marker for SES and was defined by the highest
level of education completed by the mother at the time of pregnancy: partial high-school, high-
school, TAFE or college, and university degree. One third of Generation 1 women did not complete
high-school (33.8%), while half of them had undergone tertiary education, either TAFE/college
(30.3%) or a university degree (18.3%) by the time of the study pregnancy. Maternal pre-pregnancy
BMI was associated with the educational attainment (p=0.018); women with a university degree had,
on average, the lowest mean pre-pregnancy BMI (23.7 (4.09) kg/m2) while those with partial high-
school had the highest mean BMI (25.7 (6.49) kg/m2). GD prevalence was double in women without
high-school completed (8%) compared to those with a university degree (4%); however there was no
significant overall association between maternal education and glucose tolerance status during
pregnancy (p=0.607). There was no significant association between GWG and maternal level of
education (p=0.818). BMI z-score and WHtR were significantly higher in children whose mothers
had undergone TAFE or college by the time of pregnancy (mean (SD) BMI z-score 0.66 (1.23) and
WHtR 0.48 (0.05)) compared to children of mothers who had not completed high-school (mean (SD)
BMI z-score 0.25 (1.20), p=0.003 and WHtR 0.47 (0.05), p=0.017). Child %BF was overall
associated with maternal education at the time of pregnancy (p=0.031), but the relationship was not
significant within each level of education. Child HOMA-IR was not associated with maternal
education at the time of pregnancy (p=0.308).

4.1.4 Non-participation assessment

As with any other prospective longitudinal study, the Generation 1 cohort inevitably had a degree of
non-participation. Although low (20%), non-participation could introduce attrition bias into the study
and may affect the cohort representativeness. Therefore, baseline characteristics of three groups of
subjects were compared: the original cohort (n=557); those taking part in the 9-10 year follow-up
(n=443); and participants in the 9-10 year follow-up who provided a fasting blood sample (n=163).
Additionally, current anthropometric measures of children who provided a fasting blood sample
(n=163) were compared with those taking part in the 9-10 year follow-up but did not provide a fasting
blood sample (n=280). These comparisons were examined using independent samples t-tests (for
continuous variables) and chi-squared tests (for categorical variables).

137
Results

Maternal and child characteristics in the Generation 1 participants at 9-10 year relative to the
original cohort

The key baseline characteristics compared are listed in Table 24. Both the sample of participants at
9-10 years and the subgroup of participants who provided a fasting blood sample included mothers
from across spectrum of glucose tolerance, weight and socioeconomic status (as defined by their
education levels). Mothers and children participating in the 9-10 year follow-up were largely similar
to the original Generation 1 cohort (Table 24). Mothers of children taking part in the study at 9-10
years were older (p=0.002), were less likely to have smoked during pregnancy (p=0.001), and overall
had a higher level of education (p=0.017) than those not participating. The subsample of
participants at 9-10 years who provided a fasting blood sample was largely similar to the original
cohort, except were marginally less likely to have been exposed to maternal smoking during
pregnancy (p=0.075).

138
Table 24. Maternal and child characteristics in the Generation 1 participants at 9-10 year follow-up relative to the original cohort

Original Generation 1 cohort Participants at 9-10 years Participants at 9-10 years


with fasting blood sample collected
n=557 n=443 n=163
Maternal characteristics at study pregnancy Mean (SD) n (%) Mean (SD) n (%) Mean (SD) n (%)
Age (years) 29.8 (5.02) 30.2 (4.93) 30.3 (5.14)
Glucose tolerance status
Normal glucose tolerance 451 (81.0%) 354 (79.9%) 128 (78.5%)
Borderline gestational glucose intolerance 53 (9.52%) 41 (9.26%) 19 (11.7%)
Gestational diabetes 31 (5.57%) 30 (6.77%) 9 (5.52%)
Missing 22 (3.95%) 18 (4.06%) 7 (4.29%)
Pre-pregnancy weight (kg) 66.9 (15.0) 67.0 (15.4) 66.5 (16.0)
Pre-pregnancy height (cm) 163.6 (6.16) 163.8 (6.12) 163.6 (5.43)
Pre-pregnancy BMI (kg/m2) 25.0 (5.59) 25.0 (5.65) 24.9 (5.89)
Gestational weight gain (kg) 14.0 (4.54) 13.7 (4.38) 13.7 (4.40)
Parity
Primiparous 187 (33.6%) 153 (34.5%) 47 (28.8%)
Multiparous 370 (66.4%) 290 (65.5%) 116 (70.2%)
Smoking during pregnancy
No 438 (78.6%) 361 (81.5%) 136 (83.4%)
Yes 119 (21.4%) 82 (18.5%) 27 (16.6%)
Pregnancy-induced hypertension
No 494 (88.7%) 395 (89.2%) 147 (90.2%)
Yes 63 (11.3%) 48 (10.8%) 16 (9.82%)
Education
Partial high-school 188 (33.8%) 137 (30.9%) 52 (31.9%)
Complete high-school 98 (17.6%) 79 (17.8%) 29 (17.8%)
TAFE / college 169 (30.3%) 137 (30.9%) 47 (28.8%)
University degree 102 (18.3%) 90 (20.3%) 35 (21.5%)
Child characteristics at birth Mean (SD) n (%) Mean (SD) n (%) Mean (SD) n (%)
Weight (g) 3422.5 (530.9) 3421.4 (507.5) 3424.0 (518.2)
Gestational age at birth (weeks) 39.4 (1.63) 39.4 (1.56) 39.4 (1.66)
Sex
Female 283 (50.8%) 221 (49.9%) 89 (54.6%)
Male 274 (49.2%) 222 (50.1%) 74 (45.4%)
139
Results

Child characteristics in the Generation 1 participants at 9-10 years who provided a fasting
blood sample relative to all participating children at 9-10 years

The key child current characteristics compared are listed in Table 25. Generation 1 children taking
part in the 9-10 year follow-up who provided a fasting blood sample were in general similar to those
who did not, except they were slightly younger (mean (SD) age 9.58 (0.19) years versus 9.65 (0.31)
years, p=0.003). In summary, both the anthropometric measurements and the blood samples
collected from the children at this follow-up could be considered reasonably representative of the
South Australian population of 9-10 year old children.

Table 25. Child characteristics in the Generation 1 participants at 9-10 years who provided a fasting
blood sample collection relative to all participating children at 9-10 years

Participants at 9-10 years Participants at 9-10 years p*


n=443 with fasting blood
sample collected
n=163
Child characteristics at 9-10 years Mean (SD) Mean (SD)
Age (years) 9.65 (0.31) 9.58 (0.19) 0.0003
BMI z-score 0.41 (1.17) 0.31 (1.10) 0.145
Percentage body fat 19.8 (7.87) 19.1 (7.22) 0.135
Waist-to-height ratio 0.47 (0.05) 0.47 (0.04) 0.127
*pvalue for comparing participants with fasting blood sample collected at 9-10 years (n=163) to participants without fasting blood
sample collected (with anthropometric measurements only, n=280)

4.2 Inferential statistics

This section presents the main findings from analyses exploring the long-term metabolic
consequences for the child of three maternal factors likely to be associated with fetal overnutrition.
The effects of maternal pre-pregnancy BMI on child obesity and insulin resistance in pre-pubertal
years are detailed in Section 4.2.1, those of glucose intolerance during pregnancy in Section 4.2.2,
and those of gestational weight gain in Section 4.2.3. Two-way interactions between these maternal
factors in relation to child obesity and insulin resistance are presented in Section 4.2.4. The following
potential confounders were investigated for each relationship of interest, using a step-wise approach:
maternal age, parity, smoking during pregnancy, pregnancy-induced hypertension, and maternal
educational attainment at the time of pregnancy (as an indicator of socio-economic status). Potential
confounders with a p<0.1 were considered potentially influential and retained in the model.

140
Results

4.2.1 Maternal pre-pregnancy BMI and child outcomes

A main focus of this research project was estimating the effect that maternal pre-pregnancy BMI (as
a continuous variable) had on four child outcomes: child BMI z-score (calculated based on IOTF
reference centiles), percentage body fat (as a continuous variable obtained from bioelectrical
impedance analysis), fat pattern defined by waist-to-height ratio (as a continuous variable) and
insulin resistance, defined by HOMA-IR (as a continuous variable). In addition to the above
mentioned maternal factors (age, parity, smoking, pregnancy-induced hypertension, education),
maternal glucose tolerance status during pregnancy and gestational weight gain were also
considered as potential confounders.

4.2.1.1 Maternal pre-pregnancy BMI and child BMI z-score

In bivariate analysis, a positive correlation was found between maternal pre-pregnancy BMI and child
BMI z-score at 9-10 years (Pearson’s correlation coefficient r=0.31, p<0.001); this relationship is
graphically presented in Figure 15. There was no indication of a non-linear relationship.

Figure 15. Child BMI z-score in relation to maternal pre-pregnancy BMI


4
2
0
-2
-4

10 20 30 40 50
Maternal pre-pregnancy BMI

Child BMI z-score Fitted values

The progressive increase in child BMI and BMI z-score across categories of maternal pre-pregnancy
BMI is also shown in Table 26.

141
Results

Table 26. Child BMI and BMI z-score across categories of maternal pre-pregnancy BMI

Mean (SD) Mean (SD) Normal weight Overweight Obese


Pre-pregnancy child BMI children children children
n child BMI
BMI categories
(kg/m2) z-score n (%) n (%) n (%)
Underweight 18 16.0 (1.66) -0.40 (0.96) 17 (94.4) 1 (5.56) 0
Normal weight 259 17.3 (2.63) 0.10 (0.96) 217 (83. 8) 33 (12.7) 9 (3.47)
Overweight 96 18.6 (3.08) 0.56 (0.99) 62 (64.6) 27 (28.1) 7 (7.29)
Obese 70 19.6 (4.37) 0.73 (1.02) 45 (64.3) 15 (21.4) 10 (14.3)
Total 443 17.9 (3.17) 0.28 (1.02) 341 (77.0) 76 (17.2) 26 (5.87)

The relationship between maternal pre-pregnancy BMI and child BMI z-score was assessed using
multiple linear regression models. Assumptions of linear ordinary least squares models were met.
Child BMI z-score was positively associated with maternal pre-pregnancy BMI (Model 1, Table 27).
This relationship was robust to adjustment for various potential confounders (Models 2-8, Table 27).
After adjustment for GWG (p=0.018), pregnancy-induced hypertension (p=0.046) and maternal
education (p=0.007), for each one kg/m2 increase in pre-pregnancy BMI, the child BMI z-score
increased by 0.08 (p<0.001) (Final model, Table 27).

Table 27. Estimated change in child BMI z-score in relation to maternal pre-pregnancy BMI (n=443)

Model BMI z-score change for one kg/m2


Model description
number increase in pre-pregnancy BMI (95% CI)
Model 1 Unadjusted 0.06 (0.05, 0.08)
Model 2 M1 + glucose tolerance during pregnancy 0.07 (0.05, 0.09)
Model 3 M1 + gestational weight gain† 0.07 (0.05, 0.09)
Model 4 M3 + maternal age at the time of pregnancy 0.07 (0.05, 0.09)
Model 5 M3 + parity 0.07 (0.05, 0.09)
Model 6 M3 + maternal smoking during pregnancy 0.07 (0.05, 0.09)
Model 7 M3 + pregnancy-induced hypertension† 0.07 (0.05, 0.09)
Model 8 M7 + maternal education at the time of pregnancy† 0.08 (0.06, 0.10)
Final Pre-pregnancy BMI + gestational weight gain +
0.08 (0.06, 0.10)
model pregnancy-induced hypertension + education (M8)
† p<0.1 for potential confounder, hence retained in the model

4.2.1.2 Maternal pre-pregnancy BMI and child percentage body fat

Similar to child BMI z-score, child %BF increased with maternal pre-pregnancy BMI (Pearson’s
correlation coefficient r=0.30, p<0.001) (Figure 16). There was no indication of a non-linear
relationship.

142
Results

Figure 16. Child percentage body fat in relation to maternal pre-pregnancy BMI

50
40
30
20
10
0

10 20 30 40 50
Maternal pre-pregnancy BMI

Child percentage body fat Fitted values

As evident from the graph and further described in Table 28, children of obese mothers had the
highest %BF.

Table 28. Child percentage body fat across categories of maternal pre-pregnancy BMI

Pre-pregnancy BMI categories n Mean (SD) percentage body fat

Underweight 18 15.4 (5.48)


Normal weight 257 18.3 (7.03)
Overweight 95 22.1 (7.80)
Obese 69 23.5 (9.25)
Total 439 19.8 (7.87)

The association between maternal pre-pregnancy BMI and child %BF was assessed using linear
regression models. Similar to the child BMI z-score, %BF was positively associated with maternal
pre-pregnancy BMI (Model 1, Table 29). This relationship remained statistically significant after
controlling for potential confounders (Models 2-8, Table 29). After adjusting for GWG (p=0.060) and
maternal education (p=0.100), for each one kg/m2 increase in pre-pregnancy BMI, child %BF
increased by 0.44 (p<0.001).

143
Results

Table 29. Estimated change in child percentage body fat in relation to maternal pre-pregnancy BMI
(n=439)

Model Percent body fat change for one kg/m2


Model description
number increase in pre-pregnancy BMI (95% CI)
Model 1 Unadjusted 0.42 (0.30, 0.55)
Model 2 M1 + glucose tolerance during pregnancy 0.45 (0.31, 0.58)
Model 3 M1 + gestational weight gain† 0.45 (0.31, 0.58)
Model 4 M3 + maternal age at the time of pregnancy 0.45 (0.31, 0.59)
Model 5 M3 + parity 0.45 (0.31, 0.58)
Model 6 M3 + maternal smoking during pregnancy 0.45 (0.31, 0.58)
Model 7 M3 + pregnancy-induced hypertension 0.46 (0.32, 0.60)
Model 8 M3 + maternal education at the time of pregnancy† 0.44 (0.31, 0.58)
Final Pre-pregnancy BMI + gestational weight gain +
0.44 (0.31, 0.58)
model education (M8)
† p<0.1 for potential confounder, hence retained in the model

4.2.1.3 Maternal pre-pregnancy BMI and child waist-to-height ratio

There was an increasing trend of child WHtR with pre-pregnancy BMI (correlation coefficient r=0.29,
p<0.001) (Figure 17) and across categories of maternal pre-pregnancy BMI (Table 30).

Figure 17. Child waist-to-height ratio in relation to maternal pre-pregnancy BMI


0.7
0.6
0.5
0.4

10 20 30 40 50
Maternal pre-pregnancy BMI

Child waist-to-height ratio Fitted values

144
Results

Table 30. Child waist-to-height ratio across categories of maternal pre-pregnancy BMI

Pre-pregnancy BMI categories n Mean (SD) waist-to-height ratio

Underweight 18 0.45 (0.03)


Normal weight 258 0.46 (0.04)
Overweight 96 0.48 (0.05)
Obese 70 0.50 (0.06)
Total 442 0.47 (0.05)

The relationship between maternal pre-pregnancy BMI and child WHtR was examined by employing
linear regression analysis. There was a positive association between maternal pre-pregnancy BMI
and child WHtR (Model 1, Table 31), which remained statistically significant after adjustment for
potential confounders (Models 2-8, Table 31). Of the potential confounders considered, pregnancy-
induced hypertension and maternal education at the time of pregnancy were considered influential
and retained in the model (p<0.1).

Table 31. Estimated change in child waist-to-height ratio in relation to maternal pre-pregnancy BMI
(n=442)

Model WHtR change for one kg/m2 increase


Model description
number in pre-pregnancy BMI (95% CI)
Model 1 Unadjusted 0.002 (0.002, 0.003)
Model 2 M1 + glucose tolerance during pregnancy 0.003 (0.002, 0.004)
Model 3 M1 + gestational weight gain 0.002 (0.002, 0.003)
Model 4 M1 + maternal age at the time of pregnancy 0.002 (0.002, 0.003)
Model 5 M1 + parity 0.002 (0.002, 0.003)
Model 6 M1 + maternal smoking during pregnancy 0.002 (0.002, 0.003)
Model 7 M1 + pregnancy-induced hypertension† 0.003 (0.002, 0.003)
Model 8 M7 + maternal education at the time of pregnancy† 0.002 (0.002, 0.003)
Final Pre-pregnancy BMI + pregnancy-induced
0.002 (0.002, 0.003)
model hypertension + education (M8)
† p<0.1 for potential confounder, hence retained in the model

In summary, maternal pre-pregnancy BMI was positively associated with child BMI z-score, %BF and
WHtR, and all these relationships were robust to adjustment for potential confounders.

145
Results

4.2.1.4 Maternal pre-pregnancy BMI and child insulin resistance

As described in Section 4.1.2.4, the distribution of HOMA-IR was positively skewed and the best
approximation to a normal distribution was provided by natural logarithm of HOMA-IR. No
correlation was observed between maternal pre-pregnancy BMI and natural logarithm of HOMA-IR in
the child (correlation coefficient r=0.04, p=0.608) (Figure 18). As shown in Table 32, the mean value
of HOMA-IR appeared highest among children of overweight, rather than obese, mothers.

Figure 18. Natural logarithm of child HOMA-IR in relation to maternal pre-pregnancy BMI
2
1.5
Natural logarithm of HOMA-IR
0.5 0
-0.5 1

10 20 30 40 50
Maternal pre-pregnancy BMI

Table 32. Child HOMA-IR across categories of maternal pre-pregnancy BMI

Mean (SD) natural


Pre-pregnancy BMI categories n Mean (SD) HOMA-IR
logarithm of HOMA-IR
Underweight 9 2.00 (0.56) 0.63 (0.43)
Normal weight 89 2.46 (0.72) 0.86 (0.29)
Overweight 38 2.81 (1.05) 0.97 (0.32)
Obese 27 2.49 (1.12) 0.82 (0.44)
Total 163 2.51 (0.88) 0.87 (0.34)

As child HOMA-IR was not normally distributed, in preference to applying a log-transformation,


generalised linear models with log link function were applied in order to examine the relationship
between pre-pregnancy BMI and child HOMA-IR. Results are presented as ‘back transformed’
percentage differences in HOMA-IR for one kg/m2 increase in pre-pregnancy BMI. No association
was found between maternal pre-pregnancy BMI and child HOMA-IR in the unadjusted model (Model
1, Table 33) or after adjustment for potential confounders (Models 2-8, Table 33). Maternal glucose
146
Results

tolerance status during pregnancy and education were included in the final model, as both met the
criterion for retention (p<0.1).

Table 33. Estimated change in child HOMA-IR in relation to maternal pre-pregnancy BMI (n=163)

Model % difference in HOMA-IR for one kg/m2


Model description
number increase in pre-pregnancy BMI (95% CI)

Model 1 Unadjusted 0.42 (-0.43, 1.28)

Model 2 M1 + glucose tolerance during pregnancy† 0.28 (-0.57, 1.14)

Model 3 M2 + gestational weight gain 0.28 (-0.62, 1.18)

Model 4 M2 + maternal age at the time of pregnancy 0.27 (-0.58, 1.12)

Model 5 M2 + parity 0.27 (-0.57, 1.13)

Model 6 M2 + maternal smoking during pregnancy 0.25 (-0.59, 1.11)

Model 7 M2 + pregnancy-induced hypertension 0.39 (-0.48, 1.26)

Model 8 M2 + maternal education at the time of pregnancy† 0.18 (-0.70, 1.04)

Final Pre-pregnancy BMI + glucose tolerance during


0.18 (-0.70, 1.04)
model pregnancy + education (M8)
† p<0.1 for potential confounder, hence retained in the model

Investigating whether any relationship between maternal pre-pregnancy BMI and child HOMA-IR
was independent of child current body size was deemed important at the outset. Therefore, although
no association was found between maternal pre-pregnancy BMI and child HOMA-IR (before and
after adjustment for potential confounders), child current BMI z-score was considered as a potential
mediating factor on the pathway between the early origin factor (pre-pregnancy BMI) and the later
outcome (child insulin resistance). When child BMI z-score was added to the final model, there was
a negative relationship between pre-pregnancy BMI and child HOMA-IR, so that for each one kg/m2
increase in pre-pregnancy BMI, HOMA-IR was reduced by 0.83% (95% CI -1.63, -0.02, p=0.044) for
a given BMI z-score.

As suggested by Lucas et al. (1999), the relationship between the outcome of interest and the ‘early
origin factor’ requires further examination in context with ‘later body size’. Therefore, an interaction
was explored between maternal pre-pregnancy BMI and child BMI z-score in relation to child HOMA-
IR. The effect modification of child current BMI z-score on maternal pre-pregnancy BMI in relation to
child HOMA-IR was not statistically significant in the fully adjusted model (% difference in child
HOMA-IR for the interaction term = -0.32, 95% CI -1.00, 0.37). This indicates that the negative

147
Results

relationship between maternal pre-pregnancy BMI and child HOMA-IR after adjustment for child BMI
z-score was consistent across child BMI z-score.

In summary, child HOMA-IR was not significantly associated with maternal pre-pregnancy BMI or
exposure to other prenatal factors known to influence fetal development and metabolism. However,
after controlling for child current BMI z-score, a variable likely to be a mediator on the causal
pathway, maternal pre-pregnancy BMI was inversely associated with child HOMA-IR.

4.2.2 Maternal glucose tolerance status during pregnancy and child outcomes

Maternal glucose intolerance during pregnancy was the second exposure of interest for this project.
As described in Section 3.4.1, maternal glucose tolerance status during pregnancy was defined as a
categorical, ordinal variable. The three obesity-related outcomes and insulin resistance in children
were examined as potential long-term consequences of intrauterine exposure to maternal glucose
intolerance. For each association examined, both unadjusted and adjusted models were run. The
same set of potential confounders was examined. Potential confounders with a p<0.1 were
considered potentially influential and were retained in the regression model.

4.2.2.1 Maternal glucose tolerance status during pregnancy and child BMI z-score

Inspecting the data, it appeared that mean BMI z-score in children increased with the level of
glucose intolerance during pregnancy, being highest among children whose mothers had GD (Table
34 and Figure 19). However, the differences among groups were not statistically significant.

Table 34. Child BMI and BMI z-score across the categories of maternal glucose tolerance during
pregnancy

Glucose Mean (SD) Normal


Mean (SD) Overweight Obese
tolerance child BMI weight
n child BMI
profiles in n (%) n (%)
(kg/m2) z-score n (%)
pregnancy
NGT 354 17.8 (2.94) 0.24 (1.00) 276 (78.0) 62 (17.5) 16 (4.52)
BGGI 41 18.6 (4.63) 0.32 (1.22) 31 (75.6) 3 (7.32) 7 (17.0)
GD 30 18.7 (3.58) 0.53 (0.98) 22 (73.3) 5 (16.7) 3 (10.0)
Total 425 17.9 (3.20) 0.41 (1.18) 329 (77.4) 70 (16.5) 26 (6.12)
BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

148
Results

Figure 19. Child BMI z-score across categories of maternal glucose tolerance during pregnancy

4
2
Child BMI z-score
0
-2
-4

NGT BGGI GD

BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

The association between maternal glucose tolerance status during pregnancy and BMI z-score in
childhood was assessed by linear regression. Assumptions of linear ordinary least squares models
were met. Simple linear regression was followed by multiple linear regression to include potential
confounders.

The unadjusted mean BMI z-score of children exposed to either GD or BGGI in utero was not
significantly different to the BMI z-score of children whose mothers had normal glucose tolerance
during pregnancy (Model 1, Table 35). Of the potential confounders examined, maternal pre-
pregnancy BMI, GWG, pregnancy-induced hypertension and maternal education were retained in the
model (p<0.001, p=0.052, p=0.046 and p=0.006, respectively), but their inclusion in the model did
not alter the relationship between maternal glucose tolerance status during pregnancy and child BMI
z-score at the age of 9-10 years (Models 2, 3, 7 and 8 Table 35).

In summary, glucose tolerance status during pregnancy was not associated with child BMI z-score at
the age of 9-10 years, in unadjusted or fully adjusted models.

149
Table 35. Estimated change in child BMI z-score in relation to maternal glucose tolerance status during pregnancy

NGT BGGI GD
(reference group) n=41 n=30
Model
Model description n=354
number
Adjusted mean BMI z-score change compared BMI z-score change compared
BMI z-score (95% CI) to the reference group (95% CI) to the reference group (95% CI)
Model 1 Unadjusted 0.37 (0.25, 0.49) 0.13 (-0.25, 0.51) 0.33 (-0.11, 0.77)

Model 2 M1 + pre-pregnancy BMI† 0.38 (0.26, 0.50) -0.06 (-0.43, 0.30) -0.01 (-0.44, 0.41)

Model 3 M2 + gestational weight gain† 0.40 (0.28, 0.51) -0.11 (-0.49, 0.27) -0.03 (-0.45, 0.40)

Model 4 M3 + maternal age at the time of pregnancy 0.40 (0.28, 0.51) -0.12 (-0.50, 0.26) -0.04 (-0.47, 0.39)

Model 5 M3 + parity 0.40 (0.28, 0.52) -0.09 (-0.47, 0.29) -0.04 (-0.46, 0.39)

Model 6 M3 + maternal smoking during pregnancy 0.39 (0.27, 0.51) -0.10 (-0.48, 0.28) -0.02 (-0.44, 0.41)

Model 7 M3 + pregnancy-induced hypertension† 0.39 (0.28, 0.51) -0.08 (-0.46, 0.30) -0.02 (-0.44, 0.41)

Model 8 M7 + maternal education at the time of pregnancy† 0.39 (0.27, 0.50) -0.06 (-0.44, 0.32) 0.02 (-0.40, 0.44)

Gestational glucose tolerance status + pre-pregnancy


Final
BMI + gestational weight gain + pregnancy-induced 0.39 (0.27, 0.50) -0.06 (-0.44, 0.32) 0.02 (-0.40, 0.44)
model
hypertension + education (M8)
BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance, † p<0.1 for potential confounder, hence retained in the model
150
Results

4.2.2.2 Maternal glucose tolerance status during pregnancy and child percentage body fat

Inspection of data suggested that mean %BF in children increased with the degree of maternal
glucose tolerance impairment during pregnancy (Table 36 and Figure 20) but the differences among
groups were not statistically significant.

Table 36. Child percentage body fat across the categories of maternal glucose tolerance during
pregnancy

Glucose tolerance profiles in pregnancy n Mean (SD) percentage body fat

NGT 350 19.3 (7.36)


BGGI 41 21.0 (10.7)
GD 30 23.1 (8.79)
Total 421 19.8 (7.90)
BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

Figure 20. Child percentage body fat across the categories of maternal glucose tolerance during
pregnancy
50
40
Child percentage body fat
20 10
0 30

NGT BGGI GD

BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

The effect of glucose tolerance status during pregnancy on child %BF at 9-10 years was analysed
based on linear ordinary least squares, both simple and multiple regression models. The unadjusted
mean %BF was 3.78 percentage points greater in children whose mothers developed GD while they
were in utero compared to children of mothers with normal glucose tolerance during pregnancy
(p=0.012) (Model 1, Table 37). The %BF in children of mothers with BGGI was not significantly
different relative to the reference group (Model 1, Table 37). After adjustment for maternal pre-

151
Results

pregnancy BMI the association between exposure to GD and increased %BF in children was no
longer statistically significant (Model 2, Table 37). Controlling for GWG in addition to pre-pregnancy
BMI did not make any further difference (Model 3, Table 37). Of the other potential confounders
considered, maternal level of education met the criterion for retention (p=0.073); however, this
adjustment did not modify the effect estimates (Model 8, Table 37).

In summary, an association between exposure to maternal GD and increased %BF in children was
found, but it was attenuated to the null after adjustment for maternal pre-pregnancy BMI.

152
Table 37. Estimated change in child percentage body fat in relation to maternal glucose tolerance status during pregnancy

NGT
BGGI GD
(reference group)
Model n=41 n=30
Model description n=350
number
Adjusted mean % body fat change compared % body fat change compared
% body fat (95% CI) to the reference group (95% CI) to the reference group (95% CI)
Model 1 Unadjusted 19.3 (18.5, 20.1) 1.73 (-0.82, 4.27) 3.78 (0.84, 6.71)

Model 2 M1+ pre-pregnancy BMI† 19.4 (18.6, 20.2) 0.48 (-1.97, 2.93) 1.64 (-1.23, 4.50)

Model 3 M2 + gestational weight gain 19.5 (18.7, 20.3) 0.05 (-2.50, 2.60) 1.62 (-1.24, 4.49)

Model 4 M2 + maternal age at the time of pregnancy 19.4 (18.6, 20.2) 0.50 (-1.96, 2.97) 1.67 (-1.21, 4.56)

Model 5 M2 + parity 19.4 (18.6, 20.2) 0.55 (-1.91, 3.01) 1.62 (-1.25, 4.49)

Model 6 M2 + maternal smoking during pregnancy 19.4 (18.6, 20.2) 0.49 (-1.97, 2.95) 1.65 (-1.23, 4.53)

Model 7 M2 + pregnancy-induced hypertension 19.4 (18.6, 20.2) 0.54 (-1.92, 3.00) 1.66 (-1.21, 4.53)

Model 8 M2+ maternal education at the time of pregnancy† 19.4 (18.6, 20.2) 0.56 (-1.89, 3.00) 1.73 (-1.13, 4.59)

Final Gestational glucose tolerance status + pre-pregnancy


19.4 (18.6, 20.2) 0.56 (-1.89, 3.00) 1.73 (-1.13, 4.59)
model BMI + education (M8)
BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance, † p<0.1 for potential confounder, hence retained in the model
153
Results

4.2.2.3 Maternal glucose tolerance status during pregnancy and child waist-to-height ratio

No difference in mean WHtR of the children was observed among the three categories of maternal
glucose tolerance during pregnancy (Table 38 and Figure 21).

Table 38. Child waist-to-height ratio across categories of maternal glucose tolerance during
pregnancy

Glucose tolerance profiles in pregnancy n Mean (SD) WHtR

NGT 353 0.469 (0.044)


BGGI 41 0.477 (0.067)
GD 30 0.471 (0.050)
Total 424 0.470 (0.047)
BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

Figure 21. Child waist-to-height ratio across categories of maternal glucose tolerance during
pregnancy
0.7 0.6
Waist-to-height ratio
0.5 0.4

NGT BGGI GD

BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

The relationship between categories of maternal glucose tolerance during pregnancy and child
WHtR was examined using linear regression modelling. No association was found between
categories of maternal glucose tolerance during pregnancy and WHtR in the unadjusted (Model 1,
Table 39) or adjusted models (Models 2-8, Table 39). Of the potential confounders considered,
maternal pre-pregnancy BMI, pregnancy-induced hypertension and maternal educational level were
retained in the final model (p<0.1).

154
Table 39. Estimated change in child waist-to-height ratio in relation to maternal glucose tolerance status during pregnancy

NGT
BGGI GD
(reference group)
Model n=41 n=30
Model description n=353
number
Adjusted mean WHtR change compared WHtR change compared
WHtR (95% CI) to the reference group (95% CI) to the reference group (95% CI)
Model 1 Unadjusted 0.47 (0.46, 0.47) 0.008 (-0.007, 0.024) 0.003 (-0.015, 0.021)

Model 2 M1+ pre-pregnancy BMI† 0.47 (0.46, 0.47) 0.001 (-0.014, 0.015) -0.010 (-0.027, 0.007)

Model 3 M2 + gestational weight gain 0.47 (0.46, 0.47) -0.002 (-0.018, 0.013) -0.010 (-0.027, 0.007)

Model 4 M2 + maternal age at the time of pregnancy 0.47 (0.46, 0.47) 0.001 (-0.014, 0.015) -0.010 (-0.027, 0.007)

Model 5 M2 + parity 0.47 (0.46, 0.47) 0.000 (-0.014, 0.015) -0.010 (-0.027, 0.007)

Model 6 M2 + maternal smoking during pregnancy 0.47 (0.46, 0.47) 0.001 (-0.014, 0.016) -0.010 (-0.027, 0.008)

Model 7 M2 + pregnancy-induced hypertension† 0.47 (0.46, 0.47) 0.001 (-0.013, 0.016) -0.010 (-0.027, 0.007)

Model 8 M7+ maternal education at the time of pregnancy† 0.47 (0.46, 0.47) 0.003 (-0.012, 0.017) -0.009 (-0.026, 0.008)

Final Gestational glucose tolerance status + pre-pregnancy


0.47 (0.46, 0.47) 0.003 (-0.012, 0.017) -0.009 (-0.026, 0.008)
model BMI + pregnancy-induced hypertension + education (M8)
BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance, † p<0.1 for potential confounder, hence retained in the model
155
Results

4.2.2.4 Maternal glucose tolerance status during pregnancy and child insulin resistance

Children of GD mothers appeared to have the highest mean HOMA-IR, while the mean HOMA-IR of
children whose mothers developed BGGI appeared to be lower compared to those of mothers who
maintained normal glucose tolerance during pregnancy (Table 40 and Figure 22).

Table 40. Child HOMA-IR across categories of maternal glucose tolerance during pregnancy

Glucose tolerance profiles in pregnancy n Mean (SD) HOMA-IR

NGT 128 2.47 (0.80)


BGGI 19 2.11 (0.72)
GD 9 3.42 (1.38)
Total 156 2.48 (0.87)
BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

Figure 22. Child HOMA-IR across categories of maternal glucose tolerance during pregnancy
8
6
HOMA-IR index
42
0

NGT BGGI GD

BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance

The association between glucose tolerance during pregnancy and child HOMA-IR was assessed
using a generalised linear model with log link function and Gaussian family. Data are presented as
‘back transformed’ means for the reference group and percentage differences for BGGI and GD.

Overall, glucose tolerance during pregnancy was associated with child HOMA-IR (p<0.001). In
models unadjusted for confounders, mean HOMA-IR in children exposed to GD was 38.5% greater
(p<0.001) than in children whose mothers maintained normal glucose tolerance during pregnancy

156
Results

(Model 1, Table 41). The unadjusted mean HOMA-IR in offspring of women with BGGI was not
significantly different to the reference group (Model 1, Table 41). Of the potential confounders
investigated, maternal education met the criterion for retention (p=0.076), but adjustment for this
variable did not result in appreciable changes in the estimates of the association between glucose
tolerance status during pregnancy and HOMA-IR for children (Model 8, Table 41).

In addition to the potential confounders, the child current BMI z-score was assessed as a potential
mediating factor on the pathway linking maternal glucose tolerance status during pregnancy and
child HOMA-IR. The addition of child BMI z-score (p<0.001) to the final model slightly attenuated the
relationship between exposure to GD and HOMA-IR, but it remained statistically significant (39.8%,
95% CI 21.9, 60.4). In the model including confounding and mediating variables, children exposed
to BGGI in utero had significantly lower HOMA-IR compared to children of mothers with normal
glucose tolerance during pregnancy (-17.9%, 95% CI -29.9, -3.96). Following the recommendations
of Lucas et al. (1999), an assessment of whether the effect of exposure to maternal glucose
intolerance during pregnancy on child HOMA-IR was different according to child BMI z-score was
undertaken. A significant positive interaction was identified between GD (but not BGGI) and child
current BMI z-score in relation to HOMA-IR (12.7%, 95% CI 1.55, 15.0).

In summary, in utero exposure to GD was associated with a significantly higher HOMA-IR in children
at the age of 9-10 years compared to offspring whose mothers maintained NGT. This relationship
was robust to adjustment for potential confounders and the potential mediating effect of child current
BMI z-score. In addition to the independent effect of GD on child HOMA-IR, GD had a significant
synergistic effect with child BMI z-score on HOMA-IR. Exposure to BGGI was not associated with
child HOMA-IR in unadjusted models or after adjustment for potential confounders, but had a small
negative effect on child HOMA-IR after including child current BMI z-score in the model, which was
consistent across the range of child current BMI z-score.

157
Table 41. Estimated change in child HOMA-IR in relation to maternal glucose tolerance status during pregnancy

NGT GD
BGGI
(reference group) n=9
Model n=19
Model description n=128
number
Adjusted mean % difference in HOMA-IR compared % difference in HOMA-IR compared
HOMA-IR (95% CI) to the reference group (95% CI) to the reference group (95% CI)
Model 1 Unadjusted 2.47 (2.33, 2.62) -14.5 (-29.0, 3.00) 38.5 (16.9, 64.0)

Model 2 M1+ pre-pregnancy BMI 2.47 (2.33, 2.62) -15.2 (-29.8, 2.42) 37.2 (15.6, 62.9)

Model 3 M1 + gestational weight gain 2.48 (2.34, 2.64) -14.3 (-29.7, 4.13) 37.8 (16.1, 63.6)

Model 4 M1 + maternal age at the time of pregnancy 2.48 (2.34, 2.63) -14.3 (-28. 9, 3.27) 40.4 (18.2, 66.8)

Model 5 M1 + parity 2.47 (2.32, 2.62) -13.7 (-28.5, 4.12) 37.6 (16.1, 63.2)

Model 6 M1 + maternal smoking during pregnancy 2.49 (2.34, 2.63) -15.6 (-29.9, 1.68) 38.5 (17.0, 63.9)

Model 7 M1 + pregnancy-induced hypertension 2.47 (2.32, 2.62) -13.1 (-28.1, 5.05) 39.1 (17.5, 64.8)

Model 8 M1+ maternal education at the time of pregnancy† 2.46 (2.32, 2.60) -14.8 (-29.1, 2.46) 42.9 (20.9, 68.9)

Final
Gestational glucose tolerance status + education (M8) 2.46 (2.32, 2.60) -14.8 (-29.1, 2.46) 42.9 (20.9, 68.9)
model
BGGI - borderline gestational glucose intolerance, GD - gestational diabetes, NGT - normal glucose tolerance, † p<0.1 for potential confounder, hence retained in the model
158
Results

4.2.3 Maternal gestational weight gain and child outcomes

In this section, the focus was on investigating the relationships between maternal GWG and child
BMI z-score, percentage body fat, waist-to-height ratio and HOMA-IR in pre-pubertal years.
Following the unadjusted model for each relationship of interest, models were constructed with
consecutive adjustments for potential confounders.

4.2.3.1 Maternal gestational weight gain and child BMI z-score

Overall, no correlation was found between maternal GWG and child BMI z-score (correlation
coefficient r=0.01, p=0.863), as presented in Figure 23.

Figure 23. Child BMI z-score in relation to maternal gestational weight gain
4
2
Child BMI z-score
0
-2
-4

0 5 10 15 20 25
Gestational weight gain

In order to further assess the relationship between maternal GWG and child BMI z-score, linear
regression models were applied, including adjustments for potential confounders. There was no
association between GWG and child BMI z-score in the unadjusted model (Model 1, Table 42).
Adjustment for maternal pre-pregnancy BMI led to a statistically significant positive relationship
between GWG and child BMI z-score (Model 3, Table 42). After additional adjustments for
pregnancy-induced hypertension and maternal education (the potential confounders that met
criterion for retention, p<0.1), child BMI z-score increased by 0.032 for each one kg increase in
gestational weight gain (Final model, Table 42).

159
Results

Table 42. Estimated change in child BMI z-score in relation to maternal gestational weight gain
(n=430)

Model BMI z-score change for one kg


Model description
number increase in GWG (95% CI)
Model 1 Unadjusted 0.002 (-0.023, 0.028)
Model 2 M1 + glucose tolerance during pregnancy -0.001 (-0.027, 0.026)
Model 3 M1 + pre-pregnancy BMI† 0.030 (0.004, 0.055)
Model 4 M3 + maternal age at the time of pregnancy 0.031 (0.005, 0.056)
Model 5 M3 + parity 0.033 (0.007, 0.058)
Model 6 M3 + maternal smoking during pregnancy 0.029 (0.004, 0.055)
Model 7 M3 + pregnancy-induced hypertension† 0.032 (0.007, 0.057)
Model 8 M7 + maternal education at the time of pregnancy† 0.032 (0.007, 0.057)
Final Gestational weight gain + pre-pregnancy BMI + 0.032 (0.007, 0.057)
model pregnancy induced hypertension + education (M8)
† p<0.1 for potential confounder, hence retained in the model

4.2.3.2 Maternal gestational weight gain and child percentage body fat

Child %BF did not correlate overall with maternal weight gain during pregnancy (Pearson’s
correlation coefficient r= -0.01, p=0.917); this is shown graphically in Figure 24. The relationship
between GWG and child %BF was further examined using linear regression modelling and
considering the same set of potential confounders. Maternal GWG was not associated with child
%BF in the unadjusted model (Model 1, Table 43). However, after adjustment for pre-pregnancy
BMI, there was a borderline positive association between maternal GWG and child %BF (Model 3,
Table 43). Inclusion of other potential confounders in the models did not appreciably change the
effect size and the p-value was generally borderline (the exception was Model 5, Table 43, which
was statistically significant).

160
Results

Figure 24. Child percentage body fat in relation to maternal gestational weight gain

50
40
Child percentage body fat
20 10
0 30

0 5 10 15 20 25
Gestational weight gain (kg)

Table 43. Estimated change in child percentage body fat in relation to maternal gestational weight
gain (n=426)

Model Percentage body fat change for


Model description
number one kg increase in GWG (95% CI)
Model 1 Unadjusted 0.009 (-0.180, 0.162)
Model 2 M1 + glucose tolerance during pregnancy -0.034 (-0.211, 0.143)
Model 3 M1 + pre-pregnancy BMI† 0.164 (-0.007, 0.335)
Model 4 M3 + maternal age at the time of pregnancy 0.166 (-0.006, 0.338)
Model 5 M3 + parity 0.181 (0.007, 0.354)
Model 6 M3 + maternal smoking during pregnancy 0.165 (-0.006, 0.336)
Model 7 M3 + pregnancy-induced hypertension 0.170 (-0.002, 0.342)
Model 8 M3 + maternal education at the time of pregnancy† 0.157 (-0.014, 0.328)
Final Gestational weight gain + pre-pregnancy BMI +
0.157 (-0.014, 0.328)
model education (M8)
† p<0.1 for potential confounder, hence retained in the model

161
Results

4.2.3.3 Maternal gestational weight gain and child waist-to-height ratio

Bivariate analysis showed no association between maternal weight gain during pregnancy and child
WHtR (correlation coefficient r=0.04, p=0.385); this is presented in Figure 25. In multiple linear
regression, no relationship was found between maternal GWG and child WHtR, neither in unadjusted
(Model 1, Table 44) nor adjusted models (Models 2-8, Table 44).

Figure 25. Child waist-to-height ratio in relation to maternal gestational weight gain
0.7 0.6
Child waist-to-height ratio
0.5 0.4

0 5 10 15 20 25
Gestational weight gain (kg)

Table 44. Estimated change in child waist-to-height ratio in relation to maternal gestational weight
gain (n=429)

Model WHtR change for one kg


Model description
number increase in GWG (95% CI)
Model 1 Unadjusted 0.000 (-0.001, 0.001)
Model 2 M1 + glucose tolerance during pregnancy -0.001 (-0.002, 0.000)
Model 3 M1 + pre-pregnancy BMI† 0.000 (-0.001, 0.002)
Model 4 M3 + maternal age at the time of pregnancy 0.001 (-0.001, 0.002)
Model 5 M3 + parity 0.000 (-0.001, 0.002)
Model 6 M3 + maternal smoking during pregnancy 0.000 (-0.001, 0.002)
Model 7 M3 + pregnancy-induced hypertension† 0.001 (0.000, 0.002)
Model 8 M7 + maternal education at the time of pregnancy† 0.001 (0.000, 0.002)
Final Gestational weight gain + pre-pregnancy BMI + 0.001 (0.000, 0.002)
model pregnancy-induced hypertension + education (M8)
† p<0.1 for potential confounder, hence retained in the model

162
Results

4.2.3.4 Maternal gestational weight gain and child insulin resistance

The index of insulin resistance, HOMA-IR, in children did not correlate with maternal weight gain
during pregnancy (r=0.04), as illustrated in Figure 26.

Figure 26. Child HOMA-IR in relation to maternal gestational weight gain


8
6
HOMA-IR index
42
0

0 5 10 15 20 25
Gestational weight gain (kg)

The association between GWG and child HOMA-IR was investigated using a generalised linear
model with log link and Gaussian family. No significant relationship was identified between maternal
GWG and HOMA-IR in children, either in the unadjusted model (Model 1, Table 45) or after
adjustment for potential confounders (Models 2-8, Table 45).

Child current BMI z-score was considered as a potential mediator on the pathway between maternal
weight gain during pregnancy and child HOMA-IR, but its addition to the model did not have an effect
on the association of interest (-0.50, 95% CI -1.51, 0.51).

In summary, of the four outcomes of interest in children, only BMI z-score was associated with
maternal GWG after adjustment for pre-pregnancy BMI. Neither of the specific measures of
adiposity in children, nor the index of insulin resistance was associated with GWG in unadjusted or
fully adjusted models.

163
Results

Table 45. Estimated change in child HOMA-IR in relation to maternal gestational weight gain (n=157)

% difference in HOMA-IR
Model
Model description for one kg increase in p
number
GWG (95% CI)
Model 1 Unadjusted -0.32 (-1.60, 0.97) 0.626
Model 2 M1 + glucose tolerance during pregnancy† -0.17 (-1.42, 1.10) 0.970
Model 3 M2 + pre-pregnancy BMI -0.09 (-1.36, 1.21) 0.895
Model 4 M2 + maternal age at the time of pregnancy -0.19 (-1.43, 1.08) 0.772
Model 5 M2 + parity -0.11 (-1.39, 1.18) 0.863
Model 6 M2 + maternal smoking during pregnancy† 0.02 (-1.25, 1.31) 0.977
Model 7 M6 + pregnancy-induced hypertension 0.00 (-1.28, 1.29) 0.999
Model 8 M6 + maternal education at the time of pregnancy† 0.04 (-1.20, 1.29) 0.955
Final Gestational weight gain + glucose tolerance during pregnancy 0.04 (-1.20, 1.29) 0.955
model + maternal smoking in pregnancy + education (M8)
† p<0.1 for potential confounder, hence retained in the model

4.2.4 Two-way interactions between the three intrauterine exposures of interest in relation
to child outcomes

In the models described in Sections 4.2.1, 4.2.2, and 4.2.3, the relationship between each maternal
intrauterine exposure and child outcomes of interest was assumed to be consistent across different
levels of the other intrauterine factors. In this section, first order interactions between pairs of
intrauterine exposures were explored in relation to child outcomes.

Interaction between maternal pre-pregnancy BMI and glucose tolerance status during
pregnancy in relation to child outcomes

There was no significant interaction between maternal pre-pregnancy BMI and categories of glucose
tolerance during pregnancy, in relation to child BMI z-score, %BF, or HOMA-IR as outcomes. A
significant interaction was identified between maternal pre-pregnancy BMI and glucose tolerance
status when WHtR was the outcome (p=0.033). The association between pre-pregnancy BMI and
WHtR was greater in children of BGGI mothers than in those whose mothers had normal glucose
tolerance during pregnancy. For every one kg/m2 increase in maternal pre-pregnancy BMI, WHtR at
9-10 years increased by 0.003 more in children of mothers with BGGI than children of mothers with
normal glucose tolerance (p=0.013) with adjustment for pregnancy-induced hypertension and
maternal education. This interaction is presented in Figure 27.

164
Results

Figure 27. Interaction between maternal glucose tolerance status during pregnancy and pre-
pregnancy BMI in relation to child waist-to-height ratio

0.65
0.6
Child waist-to-height ratio
0.55
0.5
0.45

10 20 30 40 50
Pre-pregnancy BMI

Normal glucose tolerance Borderline gestational glucose intolerance


Gestational diabetes

Interaction between maternal pre-pregnancy BMI and gestational weight gain in relation to
child outcomes

No significant interaction was identified between maternal pre-pregnancy BMI and GWG in relation
to child BMI z-score, %BF or HOMA-IR. There was a significant interaction between maternal pre-
pregnancy BMI and GWG when child WHtR was the outcome. The association between GWG and
WHtR increased with increasing pre-pregnancy BMI. For every one kg/m2 increase in maternal pre-
pregnancy BMI, child WHtR increased by 0.0002 more per one kg increase in GWG (p=0.010). In
order to facilitate interpretation of results from this interaction, the relationship between child waist-to-
height ratio and maternal GWG is presented in Figure 28, separately according to maternal pre-
pregnancy BMI status. Due to the small number (n=18), underweight women were included in the
same category with normal weight women (n=259).

165
Results

Figure 28. Interaction between maternal pre-pregnancy BMI and gestational weight gain in relation to
child waist-to-height ratio

0.65 0.6
Child waist-to-height ratio
0.5 0.55
0.45

0 5 10 15 20 25
Gestational weight gain (kg)

Underweight and normal weight Overweight


Obese

Interaction between glucose tolerance status during pregnancy and gestational weight gain
in relation to child outcomes

There was no significant interaction between maternal glucose tolerance status during pregnancy
and GWG in relation to child BMI z-score, %BF, WHtR or HOMA-IR.

166
Chapter 5 Discussion and conclusion

The aim of the current project was to shed light on the impact of an intrauterine environment
characterised by overnutrition on obesity and insulin resistance in pre-pubertal offspring. This final
chapter provides a synopsis of the main findings (Section 5.1), an outline of the strengths and
limitations of this study (Section 5.2), an interpretation of the findings in the context of previous
research (Section 5.3), a discussion of potential implications and recommendations for public health
(Section 5.4), suggestions for future research (Section 5.5) and a concluding statement (Section 5.6).

5.1 Overview of major findings

This is the first prospective longitudinal study to concomitantly show in an analysis with thorough
adjustment for confounders a positive, continuous association between maternal body size at the
time of pregnancy, and a range of obesity measures in pre-pubertal children (from the global BMI z-
score, to the more specific percentage body fat and waist-to-height ratio). These results confirm
previous studies showing a positive association between maternal pre-pregnancy obesity and a
child’s risk of developing obesity. However, only one earlier study (Lawlor et al. 2007b) considered
maternal pre-pregnancy BMI and child BMI as continuous variables. Thus, the current study
contributes significantly to the description of progressive effects of maternal body size prior to
pregnancy on child overall weight and adiposity. Maternal pre-pregnancy weight for height was not
associated with child IR before or after adjustment for potential confounders.

This is the first study to report interactions between maternal body size prior to pregnancy and two
other aspects of intrauterine environment in relation to child central adiposity. The positive
association between maternal weight for height prior to pregnancy and child central adiposity was
amplified in children whose mothers had developed mild glucose intolerance during pregnancy
(compared to those whose mothers had maintained normal glucose tolerance) and in children whose
mothers gained more weight during pregnancy.
167
Discussion and Conclusion

Although maternal GWG has been repeatedly documented as a risk factor for obesity in the
offspring, this is the first study to investigate it as a potential contributor to child insulin resistance.
Maternal GWG was not associated with any of the child outcomes of interest (BMI z-score, %BF,
WHtR or HOMA-IR) in unadjusted models, but a positive association was found between maternal
GWG and child body size or adiposity after adjustment for pre-pregnancy weight for height.

This study advances the work in the area of intrauterine programming of child obesity and IR by
maternal gestational glucose intolerance, as it addresses not only GD, but also milder degrees of
gestational glucose intolerance as a potential influence. Exposure to maternal GD, but not to milder
degrees of glucose intolerance, was associated with higher IR in children, with this association being
robust to adjustment for potential confounders. In contrast, no association was found between
maternal mild or clinically defined gestational glucose intolerance and child body size or central
adiposity at 9-10 years. The positive association identified between maternal GD and child overall
adiposity was attenuated to null after adjustment for maternal pre-pregnancy body size.

As IR is strongly correlated with body size and given the debate regarding the appropriateness of
including offspring’s current body size when analysing intrauterine effects on later outcomes,
possible mediation by current body size was considered for all relationships between maternal
factors and child IR. Maternal pre-pregnancy body size was inversely associated with child IR after
considering child current body size as a potential mediator. The relationship between maternal GD
and child IR was only partly mediated by child current body size. Moreover, maternal GD and child
current body size synergistically increased child IR (there was a significant interaction). Exposure to
milder degrees of glucose intolerance during pregnancy had the opposite effect, being associated
with a reduction of child IR after including child current body size as a potential mediating variable.
Child current body size did not mediate the effects of maternal GWG on child IR.

Of the three maternal factors associated with fetal overnutrition examined in this study, maternal
body size prior to pregnancy appears to play the dominant role in programming child obesity,
interacting with at least two other intrauterine exposures (GWG and BGGI) to further increase the
risk of central adiposity. Maternal GD, on the other hand, appears to be an important, independent
intrauterine contributor to child insulin resistance prior to puberty.

168
Discussion and Conclusion

5.2 Study strengths and limitations

A number of strengths pertaining to the design of this study, the assessments performed in mothers
during pregnancy and children at the current follow-up, and the analyses presented in this thesis
warrant mention. At the same time, several limitations in the study design, methods and analyses
need to be acknowledged when interpreting the findings of this thesis. The strengths and limitations
relevant to this project are addressed below.

5.2.1 Study design

The three major methodological strengths of this study are its prospective longitudinal design, the
sample representativeness and the high participation rates at follow-up. The Generation 1 study
offered a sound dataset for the purpose of this project. The richness of the available pregnancy data
meant that it was possible to examine the role of specific intrauterine exposures on childhood obesity
and related metabolic outcomes, while taking into account the potential influences of other
covariates.

The prospective collection of the data minimised selection bias as all three exposures of interest
(maternal BMI prior to pregnancy, gestational glucose tolerance status and weight gain) were
ascertained at the time of pregnancy, well before the investigation of obesity and related outcomes in
children. Given that follow-up of children in relation to obesity, fat pattern, or insulin resistance was
not foreseen at the time of pregnancy, it is unlikely that the outcomes of interest would have
influenced the measurement and classification of exposures.

Sample representativeness across a range of demographic variables is important to ensure


minimal selection bias and increases the reliability of the results and statistical power (by not
attenuating exposure variables). For the Generation 1 cohort, sample representativeness at
baseline across SES and other demographic variables, a highly recommended but rarely achieved
feature in observational studies, was obtained by balanced recruitment from public and private
antenatal clinics, with high initial response rate (as described in Section 3.1.2). This is a strong
feature of the Generation 1 study given the well known poor representation of low SES participants
in observational studies (e.g., Project Viva cohort in the US was more likely to include well educated
women (Oken et al. 2005); 29.3% women had a graduate degree and only 3% had not completed
high school (Rich-Edwards et al. 2010) as opposed to 18.3% and 33.8%, respectively, in Generation
1 women). As previously reported (Moore et al. 2004), women who accepted the invitation to enrol
169
Discussion and Conclusion

in the current study were slightly older (29 years versus 28 years, p<0.05) and their mean area
disadvantage score based on postcode of residence was slightly higher (mean difference was 56,
p<0.05) than those who declined to participate. These differences are small and unlikely to alter
associations of interest unless the relationship between the exposure and outcome is different for
non-participants, which seems improbable.

The Generation 1 cohort included only Caucasian women, thus restricting the ability to generalise
the study findings to other ethnic groups in which susceptibility to obesity, glucose intolerance during
pregnancy, excessive GWG, and associations with childhood obesity and insulin resistance might be
different. Other ethnic groups were excluded since the initial focus of the Generation 1 study was
maternal dietary intake during pregnancy, requiring use of a food frequency questionnaire that was
appropriate and relevant only for Caucasian women (Moore et al. 2004).

High participation rates at each follow-up in the Generation 1 study can be largely attributed to the
good rapport built with the families by the stable research team (the chief investigators and study co-
ordinator remained the same from the inception of the cohort). At the most recent follow-up, the
participation rate for anthropometric measurements was 80% of the original cohort, which was higher
compared to any other prospective cohorts over similar length of time. For instance in ALSPAC,
among the 10,725 original participants with complete data on GD and maternal glycosuria during
pregnancy, 6,842 (64%) had available data on anthropometric measurements at the age of 9-11
years (Lawlor et al. 2010). Similarly 57% of the children had complete data on maternal GWG and
adiposity at the age of 9 years (Fraser et al. 2010). As described in Section 2.4, similar longitudinal
studies had even lower participation rates. Although impossible to directly assess selection bias, it
appears that baseline characteristics of participants differed to those of non-participants at the most
recent follow-up only with respect to mothers’ age (slightly older), education (higher level of
education) and smoking status during pregnancy (less likely to have smoked), as detailed in Section
4.1.4. Similar maternal characteristics were reported to diverge between participants and non-
participants in the 9 year follow-up of ALSPAC study (Fraser et al. 2010). Such differences do not
necessarily bias associations under investigation, as the biological relationships are likely to apply to
mothers of all ages, levels of education, and smoking status.

When Generation 1 children were 9-10 years old, despite the high participation rate for
anthropometric measurements, only 36.8% of participating children (29.3% of the original cohort)
gave consent for the collection of a fasting blood sample. Four factors might explain this lower
participation rate for blood collection: (1) children’s aversion to needles or needle phobia; (2) the

170
Discussion and Conclusion

burden of participation for children and their families (i.e., having to fast overnight and receive a
separate visit by a domiciliary nurse for blood collection in addition to participation in interviews and
anthropometric measurements was a considerable imposition, to which some of the families were
reluctant to consent); (3) some families having relocated interstate or in rural areas not covered by
the blood collection service that the Generation 1 team employed for the purpose of this study; and
(4) fieldworkers’ fatigue, which may have had an impact on their enthusiasm in asking for consent.
However, other than being slightly older (9.65 years versus 9.58 years, p=0.003), participating
children who did not consent for blood sample collection were similar to those who did (as shown in
Section 4.1.4). While no major differences were identified in the baseline characteristics of children
who took part in this follow-up and those who did not, or between those who gave consent for blood
sample collection and those who did not, a possibility still exists that other unmeasured or unknown
factors related to the outcomes might be different between the two subgroups. However, for this to
bias the reported associations the effects would need to be different in those assessed and those not
assessed, which seems implausible.

One potential limitation of this study design is the sample size. The sample size was dictated by the
size of the original cohort (n=557) and by the participation rates at follow-up, thereby preventing
changes. The relatively small sample size (n=443 mother-child pairs participating at the most recent
follow-up; n=163 fasting blood-samples collected from the children) could have constrained statistical
power and thus somewhat limited the ability to detect true effects. However, other studies with
similar or smaller sample sizes have been able to show significant associations between maternal
obesity prior to pregnancy and child obesity (Blair et al. 2007; Boerschmann et al. 2010; Catalano et
al. 2009a), or between maternal glucose levels during pregnancy and child obesity or insulin
resistance (Chandler-Laney et al. 2011; Crume et al. 2011a; Egeland and Meltzer 2010).

Sample size when conducting interaction analyses warrants a special mention. In order for the study
to have sufficient statistical power to detect an interaction effect, sample size needs to be larger than
the sample size required for estimating the main effects (Shieh 2009), hence it is plausible that this
study was underpowered to detect interactions.

171
Discussion and Conclusion

5.2.2 Measurements

Maternal self-reported pre-pregnancy weight

As described in the Methods chapter (Section 3.4.1), pre-pregnancy weight was reported by the
women in the first 16 weeks of gestation. It is well known that self-reported weight tends to
underestimate real weight, particularly among obese women (Gorber et al. 2007). It may also be
affected by recall bias, in turn influenced by age, education, or employment (Stommel and
Schoenborn 2010; Yu and Nagey 1992), by non-response bias, often due to social desirability
(Gorber et al. 2007), and sometimes by interviewer bias (Hennekens and Buring 1987), as a result of
inconsistencies in interview techniques. However, given that women were recruited during their
pregnancy, it was the only available indicator of pre-pregnancy weight in this study for this purpose.

A number of factors arguably minimise bias around self-reported pre-pregnancy weight in this study.
Firstly, the interviewers were not aware of the study hypotheses at the time of data collection and the
interviews were rigorously structured. Secondly, the time between the exposure (i.e., pre-pregnancy
weight) assessment and data collection was minimised by interviewing women early in the index
pregnancy. Thirdly, after data collection, checks were undertaken to identify those self-reported
weights that were not plausible when compared to the weights measured in early pregnancy. In
order to include data from women reporting unlikely pre-pregnancy weights (n=21) or missing data
on pre-pregnancy weight (n=55), single imputation was used, the limitations and strengths of which
will be discussed in Section 5.2.3. Moreover, while recognising limitations, self-reported weight is
widely used and accepted in epidemiological research and has been validated against direct
measures of weight (Oken et al. 2007).

An alternative approach to self-reported data may be given by estimating pre-pregnancy weight,


either by retrospective extrapolation or using standardised estimates that correct for weight gain
during early pregnancy. Retrospective extrapolation, while useful when at least two consecutive
weight measurements in early pregnancy are available (which is not often, and not in this study), is
limited by the fact that it does not account by short term fluctuations in body weight and, even more
importantly, erroneously assumes constant rate of weight gain throughout pregnancy (Harris and
Ellison 1998). Although mean weight gain in early pregnancy is considered minimal overall, wide
variations have been reported ranging between -3 kg at 9 weeks (Van Loan et al 1995) to 4.5 kg at
15 weeks (Clapp and Little 1995). In contrast to the retrospective extrapolation approach,
standardised estimates that correct for weight gain during early pregnancy have the only limitation

172
Discussion and Conclusion

that they do not take into account inter-individual variability or different weight gains for each woman
in consecutive pregnancies (Harris and Ellison 1998).

Screening for maternal gestational diabetes

Results of GD screening for Generation 1 women (i.e., plasma glucose levels at OGCT and/or
OGTT) were abstracted from antenatal records (details were presented in Section 3.4.1). Thus, they
can be considered objective and with minimal, if any, bias, particularly since the initial focus of the
study was not on GD but on maternal diet during pregnancy (Moore et al. 2004). Over 90% of
Generation 1 women underwent routine OGCT and, where required, OGTT. Given that the
screening for GD based on the two-step approach became universal in Australia in 1998, when
recruitment for this study had just commenced, some of the missing data on OGCT and/or OGTT
could be attributed to an initially less effective implementation of the screening. Another explanation
for the small amount of missing data on OGCT in women with available data on OGTT could be that
some women (n=6) were referred directly for the diagnostic test OGTT, skipping the OGCT, perhaps
due to a perceived high risk for GD (e.g., obesity, family history of diabetes).

By defining categories of glucose tolerance during pregnancy instead of using continuous plasma
glucose levels in analyses, there is some potential for misclassification and for some reduction in
statistical power (Ragland 1992; Selvin 1987), but this is believed to be unavoidable and minimal.

There is no standard international definition of BGGI. The BGGI category used in this study is
directly comparable only to the category defined on the basis of the same ADIPS criteria (Ju et al.
2008); it does not correspond with the definitions used in studies from other settings, such as the
study of Bonomo et al. (2005), which used NDDG criteria to define maternal glucose tolerance
status. Nonetheless, based on ADIPS criteria in use in Australia, this was the only possible way to
define an intermediate category of gestational glucose intolerance given that one plasma glucose
level above the thresholds at OGTT (i.e., either fasting plasma glucose ≥5.5 mmol/l or 2-hour plasma
glucose ≥8 mmol/l) is sufficient for the diagnosis of GD, as opposed to at least two plasma glucose
levels above the thresholds at OGTT required by other criteria (e.g., NDDG) used in other settings.
Thus comparing the findings of this thesis with other studies addressing health consequences of
milder degrees of gestational glucose intolerance requires prudence.

Although not common practice in Australia at the time when Generation 1 women were pregnant, it is
possible that some women with GD may have received dietary advice or even insulin therapy. If this
treatment succeeded in controlling maternal glucose levels, hence reducing fetal exposure to the

173
Discussion and Conclusion

nutrient-rich environment, the current study findings on the influence of maternal glucose intolerance
during pregnancy on child outcomes might underestimate the true association in a population where
treatment had not occurred. It is also possible that the GD group included women with unrecognized
pre-existing glucose intolerance, as they were not screened prior to pregnancy (not common
practice). However, this is unlikely to be driving the results presented in this thesis since the number
of women with this condition is likely to have been low.

Maternal gestational weight gain

In the current project, GWG was computed across the entire pregnancy (total GWG) from weight
measured shortly prior to delivery (a strength of this study) and self-reported pre-pregnancy weight
(a potential weakness, but the strategies undertaken to overcome the tendency to under-report pre-
pregnancy weight have been discussed). To maximise the dataset, single imputation was used for
women whose last recorded weight was measured early in the third trimester of pregnancy (details
were presented in Section 3.4.1). As described in Chapter 2, most previous studies used self-
reported GWG (Oken et al. 2008), which may also be problematic.

Child BMI

Body mass index, in particular age- and sex-specific BMI centiles or z-score, is widely used in
epidemiological studies as an indicator of global obesity as it is inexpensive, easy to compute, and
reproducible. In this study, child BMI was calculated based on direct measures of weight and height,
thus avoiding the potential bias introduced by self-reported measures. Ideally, the most stable
anthropometric measurements are done in the morning after overnight fasting and after voiding
(Norton and Olds 1996), to account for diurnal variation of body weight and height, both contributing
to a small increase of BMI during the day (Rodríguez et al. 2000). However, these strict conditions
were not feasible in a large-scale epidemiological study of a community sample like this, which
involved home visits at a time prioritised by the families’ availability.

Despite its advantages, BMI has some recognised limitations as a proxy measure of adiposity or of
fat distribution, as described in Section 2.3.1.3. Hence, in this thesis, obesity measures in children
were extended to include percentage body fat and fat distribution.

Child percentage body fat

For the purpose of this project, %BF, a measure of fat mass adjusted for body size, was estimated
by foot-to-foot bioelectrical impedance analysis (built into weight scales), a technique with multiple
174
Discussion and Conclusion

advantages for research in the field, and few limitations. Bioelectrical impedance analysis has the
advantage of being portable, simple, non-invasive and inexpensive. It is advantageous to skinfold
thickness measurements, a common alternative, as it is not prone to examiners’ error (Segal et al.
1991) and provides estimates for the whole body fat mass, including visceral fat (unlike the
equations based on skinfold thicknesses that take into account only subcutaneous fat). Ideally, the
assessment should have been undertaken in the morning after overnight fasting, given the small
diurnal decrease in fat mass estimate (Rodríguez et al. 2000) and that food and liquid intake are
known to influence total body water measurement (Kyle et al. 2004b). However, this strict condition
was impractical for the purpose of this study, as most children were assessed after school, to
minimise intrusion into their family life.

Child waist-to-height ratio

For the purpose of this study, fat pattern was indicated by the WHtR, a measure that has been
shown to correlate well with cardio-metabolic risk (Kahn et al. 2005; Savva et al. 2000). Using WHtR
rather than waist circumference alone or waist-to-hip ratio is a strength, as this index takes into
account an individual’s height. Another strength is that in this study only three research nurses were
involved in performing all anthropometric measurements in triplicate, thus minimising the inter-
examiner variability in measurements. This is important given that body girth measurements,
including waist circumference, are prone to examiner errors. As mentioned in Section 2.3.1.3,
although WHtR has been shown to reliably predict cardiovascular disease risk, it cannot differentiate
between the visceral and subcutaneous abdominal adipose tissue.

Imaging techniques (CT, MRI or DXA) would have provided more accurate estimates of fat
distribution, particularly the differentiation between intra-abdominal and subcutaneous adipose
depots, but their use was not feasible in this study. It is unclear how this extra precision could have
changed the results.

Child HOMA-IR

Homeostasis model assessment is a convenient index of insulin resistance that has been validated
in children against more invasive and accurate methods (Gungor et al. 2004). Although
advantageous for epidemiological studies, HOMA-IR is not without some limitations, including its
inability to differentiate between hepatic and peripheral insulin resistance, and its potential to
underestimate insulin resistance because the equation was originally calibrated to insulin assays
used in 1970s (Wallace et al. 2004). Care also needs to be taken in making comparisons of insulin

175
Discussion and Conclusion

resistance estimates based on data from different laboratories given the large insulin inter-assay
variation and the lack of standardisation of insulin assays between laboratories (Robbins et al. 1996).

It has been suggested that it would be best to use the mean of three fasting concentrations for
insulin collected at 5 minute-intervals in order to overcome the pulsatile nature of insulin secretion
and improve intra-subject variability (Wallace et al. 2004). However, this was not feasible in the
current study and, in fact, for practical reasons, most studies use one fasting insulin concentration to
calculate HOMA-IR (Monzillo and Hamdy 2003).

5.2.3 Analyses

In addition to the issue of sample size (mentioned previously as a design issue), the robustness of
findings presented in this thesis is directly influenced by the approach to the analyses, in particular
definition of variables, handling of missing data, choice of statistical models, and adjustment for
potential confounders and mediators.

A strength of this study is the way most exposure and outcome variables were included in analyses
as continuous variables. By using continuous measures for both exposures and outcomes as
opposed to categories or arbitrary cut-points, their relationships could be assessed in a more refined
manner across the spectrum, reducing the misclassification associated with categorical variables
and maximising statistical power (Ragland 1992; Selvin 1987).

Methods for dealing with missing data were discussed in Section 3.4.1. In this study, in order to
maximise the dataset, single imputation was applied for exposure variables only (i.e., maternal pre-
pregnancy weight, maternal weight just before delivery), because this could be done with high
reliability. Subsequently, complete case analyses were undertaken, which may have had two
limitations: potentially lower precision of estimates due to the reduced sample size and potentially
biased estimates. However, the drop in numbers was small (n=18) and results did not change
appreciably.

The index of insulin resistance, HOMA-IR in children, had a positively skewed distribution, for which
logarithmic transformation is often applied to obtain a normal distribution. In this study though, given
the limitations of log transforming a dependent variable (which were detailed in Section 4.1.2.4),
generalised linear models (GLM) with log link function and Gaussian family were applied in
analyses where HOMA-IR was the outcome. One main advantage of using GLM rather than log
transformation was that the cumbersome appropriate ‘back transformation’ to the original scale
176
Discussion and Conclusion

(which involves more than simple exponentiation) could be avoided (UCLA: Academic Technology
Services and Statistical Consulting Group). The second advantage was that the regression
coefficients directly estimated the expected arithmetic mean of the original variable (and not the less
informative geometric mean, as is the case with log transformed dependent variables) (UCLA:
Academic Technology Services and Statistical Consulting Group).

Thorough adjustment for potential confounding variables was a major strength of this study.
Covariates were decided a priori based on theoretical and empirical evidence for an association with
both the exposure and the outcome of interest, and were included in analyses using a stepwise
approach. The richness of the data collected during pregnancy allowed adjustment for all the
predefined potential confounders: maternal age, parity, smoking during pregnancy, pregnancy-
induced hypertension, and maternal education. However, as is the case for any observational study,
there could be residual confounders (unmeasured variables or unadjusted for in this project) that
might inflate or reduce the observed associations.

A potential limitation is that no adjustment for multiple testing was performed in this study, which may
have increased the likelihood that some of the significant findings, particularly those unexpected and
not reported previously, arose by chance. However, the appropriate approach to multiple
comparisons remains controversial. Some authors recommend adjustment to minimise the
likelihood of mistakenly rejecting the null hypothesis using, for instance, Bonferroni correction. This
approach is recommended mainly when random numbers are analysed, but with natural
observations (like in this study), adjustment would increase the likelihood of a false-negative
conclusion (Rothman 1990). Others argue that by reducing type I error for a null association, type II
error for associations that are not null is increased, thereby reducing statistical power (Perneger
1998).

Summing up Section 5.2, this study has benefited from a number of major strengths, in particular the
prospective nature of data collection, sample representativeness, high participation rates at follow-
up, and, in general, high quality data. Limitations detailed above are inevitable in a large-scale
longitudinal study like this. The most important one is likely to be the sample size, while the others
are considered unlikely to have affected the reliability of the results presented in this thesis.

177
Discussion and Conclusion

5.3 Relationship to studies on early origins of child obesity and insulin


resistance

This section contextualises the associations observed in this study with findings from previous
longitudinal studies on intrauterine influences of childhood obesity and insulin resistance. It starts
(Section 5.3.1) by comparing characteristics of the Generation 1 sample with other cohorts, thus
placing the observed findings in the context of other published data. It then presents a comparison
of the associations of interest with previous similar studies (Sections 5.3.2 to 5.3.6), as well as an
outline of potential mechanisms that might underlie the observed associations (Section 5.3.7).

5.3.1 Comparison of the Generation 1 sample with previous relevant studies

Mean pre-pregnancy BMI in Generation 1 women was comparable to that described in other
contemporary cohorts (Table 46). The prevalence of overweight and obesity prior to pregnancy was
in line with larger Australian surveys from 10 years ago, when this cohort was established (Cameron
et al. 2003), but higher relative to some studies from other settings (Table 46). The variation in the
prevalence estimates could be attributed to different rates of overweight and obesity in the
background population, sample representativeness, or measurement accuracy.

As there is no international consensus to define GD and BGGI, comparisons of the prevalence


estimates for GD and BGGI in this study with other studies are difficult. Prevalence in the current
study was slightly higher compared to those reported in the only other Australian study using the
same categories of glucose intolerance during pregnancy (Table 46). This is possibly due to the fact
that only primiparous women were included in the other study (Ju et al. 2008), who are more likely to
maintain normal glucose tolerance during pregnancy; these women were also younger compared to
Generation 1 women.

Overall, only about one third (34.9%) of Generation 1 women gained weight within the 2009
recommendations of the Institute of Medicine, while almost half (49%) gained excessive weight
during pregnancy. The amount of weight gained throughout pregnancy by the Generation 1 women
was comparable to that reported by other contemporary international studies (Table 46).

178
Table 46. Comparison of the Generation 1 sample with previous studies

Study Country Mean maternal Prevalence of Prevalence of GD Mean GWG (kg) Mean child BMI z-score Prevalence of child
(Reference) Year of pregnancy BMI prior to maternal and BGGI (SD) (SD) overweight and
data collection pregnancy (kg/m2) overweight and obesity
(SD) obesity prior to
pregnancy
Generation 1 Australia 25.0 (5.59) Overweight: 23.3% GD: 5.57% 14.0 (4.54) 0.41 (1.17) in 9 year-olds Overweight: 17.7%
1998-200 Obesity: 16% BGGI: 9.52% Obesity: 5.9%
Mater University Study of Australia 22.0 (4.1)
Pregnancy and Its 1981-1984
Outcomes
(Lawlor et al. 2007b)
Project Viva USA 24.6 (5.0) 15.6 (5.4)
(Oken et al. 2007) 1999-2002
Growing Up Today Study USA Overweight: 11.3% 14.3 (5.1)
(Oken et al. 2008) 1996 Obesity: 3.6%
ALSPAC UK Overweight or 12.1 (5.1) 0.36 (1.19) in 11 year-olds Overweight: 13.4%
(Brion et al. 2011; Fraser 1991-1992 obese: 21.3% Obesity: 15.8%
et al. 2010; Hughes et al.
2011)
Auckland Birthweight New Zealand Overweight or
Collaborative Study 1995-1997 obese: 23.4%
(Blair et al. 2007)
Australian Collaborative Australia GD: 4%
Trial of Supplements with 2001-2005 BGGI: 8%
antioxidants to pregnant
women to prevent
preeclampsia
(Ju et al. 2008)
178
Discussion and Conclusion

The prevalence of obesity in the Generation 1 children cohort was lower than the national rates
(Australian Bureau of Statistics 2009a) and those from ALSPAC (Hughes et al. 2011). In contrast,
the prevalence of overweight in Generation 1 children was similar to the rate identified by the
Australian National Health Survey (Australian Bureau of Statistics 2009a), but higher than the rates
reported in children from the ALSPAC cohort (Hughes et al. 2011) (Table 46).

Children participating in the current study appear largely similar in terms of %BF with other groups of
Caucasian children. There is limited published data regarding the reference values for %BF in
children and it is difficult to compare the mean found in children of the Generation 1 cohort (19.8% ±
7.87%) with those few other studies due to the different methods employed to estimate %BF.
However, a simple comparison of sex-specific means of %BF in the Generation 1 cohort to the
reference curves based on BIA data showed that Generation 1 girls’ %BF (20.6% ± 7.76%) was
between the 25th and 50th centiles, while boys’ %BF (19% ± 7.92%) was between the 50th and 75th
centiles (McCarthy et al. 2006).

Likewise, limited data exists with respect to reference values for WHtR in children. Relative to
children of similar age from a British cohort (McCarthy and Ashwell 2006), Generation 1 children had
higher mean WHtR. Given that the Generation 1 children were assessed more than 10 years after
the children from the above mentioned British cohort, this difference may reflect the increasing
prevalence of childhood obesity over time.

While current glucose assays provide similar plasma glucose concentrations irrespective of the
method used, no standardised method of measuring plasma insulin has been validated to date
(Staten et al. 2010). Thus, plasma insulin levels may vary widely according to the assay used,
mainly due to variations in assay specificity, calibration procedures or conversion factors (Manley et
al. 2007). Of the commercially available human insulin assays, radioimmunoassays are known to
give the highest insulin levels (Manley et al. 2007). Fasting plasma insulin levels measured in
Generation 1 children by a specific radioimmunoassay (mean 11.3 μU/ml) were similar to those
reported in Bogalusa study (Freedman et al. 2007a), but double those reported in other studies of
pre-pubertal non-obese children, which measured them by ELISA (Ong et al. 2004). Without a
reference method for insulin measurement, indices derived on the basis of the absolute value of
insulin (such as HOMA-IR) are affected by this variation. Therefore, using cut-offs to define insulin
resistance is not appropriate across different studies. This was another reason for considering
HOMA-IR as a continuous variable in this study.

180
Discussion and Conclusion

5.3.2 Maternal pre-pregnancy BMI and child outcomes

This study builds on previous research concerning the association between maternal obesity prior to
pregnancy and child obesity and related metabolic outcomes, by showing that the relationship is
continuous. The finding of a positive association between maternal pre-pregnancy BMI and child
global obesity, adiposity and fat pattern in pre-pubertal years is consistent with previous studies, as
critiqued in Section 2.4.1 and summarised in Table 6. Had Generation 1 data been analysed using
categorical variables for maternal pre-pregnancy BMI and child BMI, as per the approach taken in
the great majority of previous studies, the magnitude of the effects would have been comparable
(OR for child overweight or obesity was 1.96 (95% CI 1.19, 3.20) if the mother was overweight, and
2.19 (95% CI 1.26, 3.80) if the mother was obese prior to pregnancy). When pre-pregnancy BMI
was converted to an internal standard deviation in the Generation 1 study to facilitate comparison
with the only one other study which used both exposure and outcome variables in continuous format
(Lawlor et al. 2007b), the effect of a one standard deviation increase in maternal pre-pregnancy BMI
on offspring BMI z-score was of 0.43 (95% CI 0.31, 0.54) in fully adjusted model, similar to the
estimates reported by the previous study (Lawlor et al. 2007b).

Robust associations were identified between maternal pre-pregnancy BMI and child body size and
adiposity. While causality cannot be concluded, the consistency of this finding supports a
relationship. It is not clear if this is a direct contribution of the intrauterine environment or if it is due
to genetic factors or shared postnatal environment (i.e., similar diet and physical activity patterns).
(This is discussed further in Section 5.5.) Data obtained in this project do not allow for the explicit
untangling of these possibilities; however, they support the public health recommendations of
avoiding obesity at the beginning of pregnancy in order to reduce the obesity risk in the subsequent
generation.

No association was detected between maternal pre-pregnancy BMI and child HOMA-IR before or
after adjustment for potential confounders. This null result is in agreement with the findings of the
only previous prospective study of children of similar age to the Generation 1 children, but who had
all been exposed to GD while in utero (Boerschmann et al. 2010). In contrast, a retrospective study
in young adults reported a significantly higher degree of insulin resistance (calculated from the 2-h
OGTT) in offspring of mothers who were obese at the time of pregnancy compared to those of
normal weight mothers (Mingrone et al. 2008). Whilst the study of young adults was not without
limitation (pre-pregnancy weight and height were self-reported about 20 years after delivery), overall
these results suggest an effect of maternal pre-pregnancy BMI on offspring insulin resistance may

181
Discussion and Conclusion

not be evident until after childhood, when HOMA-IR takes on a different distribution with a greater
range of values (Li et al. 2003).

5.3.3 Maternal glucose tolerance status during pregnancy and child outcomes

The null result between GD or BGGI and child BMI z-score in both unadjusted and adjusted models
is in agreement with some (Boney et al. 2005; Gillman et al. 2003; Jeffery et al. 2006; Lawlor et al.
2010; Pirkola et al. 2010; Whitaker et al. 1998; Wright et al. 2009), but not all (Boerschmann et al.
2010; Catalano et al. 2009a; Chandler-Laney et al. 2011; Crume et al. 2011a; Egeland and Meltzer
2010; Krishnaveni et al. 2010) of the previous longitudinal studies summarised in Table 7. As
discussed in Section 2.4.2, discrepancy in findings could be due to differences in the consideration of
potential confounders, in particular maternal body size, which is known to be a strong predictor of
offspring weight status. Of the studies that did identify a crude association between exposure to
chronic hyperglycaemia in utero and child obesity, all reported an attenuation (sometimes to null)
after adjustment for maternal BMI prior to pregnancy (Crume et al. 2011a; Lawlor et al. 2010) or at
the time of follow-up (Egeland and Meltzer 2010; Gillman et al. 2003; Krishnaveni et al. 2010). This
indicates that the initial relationship may have been driven by maternal body size rather than
maternal GD. Similar to the current project, previous large-scale studies found no independent
association between GD and child obesity before and after adjustment for various confounders,
including maternal pre-pregnancy BMI (Pirkola et al. 2010; Whitaker et al. 1998; Wright et al. 2009).

The lack of an association between maternal GD and child BMI could reflect, at least in part, a
limited statistical power to detect true effects. In addition, as mentioned earlier, management of GD
may also have attenuated the associations between GD and child obesity, although this seems
unlikely for the Generation 1 study, as the intensive management of GD was not routinely
recommended at the time when the women were pregnant. Of note, in all the previous studies which
did not detect an association between maternal GD and child obesity, women with GD received
dietary advice and insulin if required (Lawlor et al. 2010; Pirkola et al. 2010; Whitaker et al. 1998;
Wright et al. 2009) or, due to their profession (nurses), were likely to have had a better glycaemic
profile during pregnancy (Gillman et al. 2003).

In contrast, data from the ALSPAC study (which has a very large sample size) showed an
association between milder degrees of hyperglycaemia during pregnancy (women presenting with
glycosuria and not receiving treatment) and an increased risk of obesity among offspring (Lawlor et
al. 2010). Similarly, an increasing risk of child overweight and obesity was shown to parallel

182
Discussion and Conclusion

increasing maternal glucose level at OGCT (Hillier et al. 2007). However, adjustment for maternal
body size at the time of pregnancy was not undertaken in that study, which might confound the
associations observed.

No association was identified in the current study between GD or BGGI and child body composition
or fat pattern. This finding is consistent with two previous studies which reported no associations
between maternal GD and percentage body fat (Catalano et al. 2009a) or fat pattern (Catalano et al.
2009a; Pirkola et al. 2010) in either children or adolescents. In contrast, exposure to GD was found
to be positively associated with child adiposity measured by the sum of subscapular and tricipital
skinfold thicknesses in 3 year old participants in Project Viva (Wright et al. 2009) or by subcutaneous
abdominal adipose tissue (from MRI measurements), as well as with fat distribution estimated from
waist circumference in 6-13 year-old children from the EPOCH study (Crume et al. 2011a). Two
other studies with smaller sample sizes to the present research, in which maternal BMI prior to
pregnancy was not considered as a potential confounder, also reported a positive association
between maternal 1-hour glucose level at OGCT and offspring fat mass at the age of 5-10 years
(Chandler-Laney et al. 2011), as well as greater waist circumference in female adolescent offspring
of mothers with GD or impaired glucose tolerance (Egeland and Meltzer 2010). In this context,
although no association was found in the current study between maternal glucose intolerance during
pregnancy and offspring body composition or fat pattern in pre-pubertal years, it is possible that such
an effect might be detectable during puberty or later in life (Singhal et al. 2003).

In contrast to child body size, maternal GD was a strong predictor of child insulin resistance,
independent of confounders. Whilst some of the previous studies have reported higher HOMA-IR in
children exposed to maternal GD compared to those whose mothers had normal glucose tolerance
during pregnancy (Boerschmann et al. 2010; Catalano et al. 2009a; Krishnaveni et al. 2010), this is
the first to present an effect after thorough adjustment for confounders.

Intrauterine exposure to maternal BGGI did not appear to be associated with child insulin resistance
in Generation 1 children. This is similar to the findings of the only other study that examined the
association between milder degrees of glucose intolerance during pregnancy, which indicated that
exposure to maternal impaired glucose tolerance during pregnancy had no significant effect on child
HOMA-IR (Egeland and Meltzer 2010).

In summary, while no significant association was identified between maternal glucose intolerance
during pregnancy (across the spectrum) and child obesity or adiposity, the findings of this thesis in

183
Discussion and Conclusion

conjunction with existing literature suggest that intrauterine exposure to maternal GD increases
insulin resistance in the child, independent of other maternal factors.

5.3.4 Maternal gestational weight gain and child outcomes

Previous research in the area of intrauterine programming by maternal GWG (critiqued in Section
2.4.3) has been dominated by studies addressing the effects on childhood obesity and their findings
have been inconsistent. Data on the effects of GWG on body composition, fat pattern and insulin
resistance in the child have been scarce. On the basis of Generation 1 data, no association was
found between maternal GWG and child BMI z-score at the age of 9-10 years until adjusting for
maternal pre-pregnancy BMI, when the association was positive, which highlights again the
importance of maternal body size prior to pregnancy for child weight status.

No association was found in the current study between maternal GWG and more specific features of
child adiposity. This is at odds with the one other study of similarly aged children from ALSPAC
(Fraser et al. 2010), which found an increase in both fat mass and waist circumference with
excessive weight gain during pregnancy. A potential explanation for the difference in findings might
be the different statistical power of the two studies.

This is the first study to investigate the association between maternal GWG and child insulin
resistance and report no significant independent effect of GWG on offspring HOMA-IR. While this
might again reflect a lack of statistical power, the fact that other associations between maternal
factors and child insulin resistance were observed suggests that GWG is relatively unimportant for
this outcome.

5.3.5 Interpreting mediation by current body size

Given the established strong correlation between an individual’s obesity and insulin resistance, it
was deemed important and appropriate in the current study to investigate whether any relationship
between maternal intrauterine conditions and child HOMA-IR was mediated by child current body
size. Examination of this mediating variable for all associations of interest was also dictated by the
existing controversy regarding the appropriate analysis and interpretation of associations between
pre-birth factors and later outcomes (Cole 2005; Lawlor et al. 2007a; Tu et al. 2005). In this section,
mediation results are interpreted based on recommendations in the literature.

184
Discussion and Conclusion

The association between maternal pre-pregnancy BMI and child IR changed from null to negative
with the addition of child current BMI z-score to the model. This suggests that amongst children who
grow to the same body size at age 9-10 years, those born to mothers with lower BMI prior to
pregnancy have, on average, higher IR or, conversely, those born to mothers with higher pre-
pregnancy BMI have, on average, lower IR. Similarly, there was no significant difference in the IR of
children of mothers with normal glucose tolerance and borderline gestational glucose intolerance
until child current BMI z-score was added to the model, when children of BGGI mothers had lower
IR. This can be interpreted as amongst children who grow to the same size at age 9-10 years, those
whose mothers had BGGI have, on average, lower IR than those whose mothers had normal
glucose tolerance during pregnancy. The positive association of maternal GD and the null
association of GWG with child IR did not change with adjustment for the child’s current body size.

These phenomena (where a positive or null association between an early life factor and later
outcome becomes negative after controlling for current size) have been described in statistical terms
as the ‘reversal paradox’ or the Simpson’s paradox, or within the generalised linear modelling
framework, as the ‘suppression effect’ (Tu et al. 2005). The interpretation of this result is contentious
(Cole 2005; Lawlor et al. 2007a; Tu et al. 2005). The reversal paradox has been investigated
principally in relation to effects of birth size and current body size on adult blood pressure (Lucas et
al. 1999). Rather than considering it as a ‘statistical artefact’ (Tu et al. 2005), it was suggested that
the inverse relationship between birth size and the later outcome with adjustment for current body
size should be interpreted as being driven by the change in body size (i.e., BMI centile crossing)
(Cole 2005; Lawlor et al. 2007a; Tu et al. 2005). I propose that this approach can be extrapolated
‘upstream’ from birth size to intrauterine exposures that influence birth size.

An alternate interpretation is that children born to mothers with low BMI had slower intrauterine
growth, followed by a period of accelerated postnatal growth (upward BMI centile crossing).
Evidence elsewhere suggests that periods of rapid weight gain in any stage of life are characterised
by a disproportionate gain in fat compared to lean body tissue (Dulloo et al. 2006), which could, in
turn, increase susceptibility to IR. As per recommendations by Lucas et al. (1999), an interaction
between maternal BMI and child current BMI was also tested for, which was not significant. It is also
possible that this result reflects a non-pathologic intergenerational tracking of BMI (whereby children
of high BMI mothers have high BMI themselves, but this higher BMI is not accompanied by
increased IR). In future, it would be interesting to examine whether the relationship between
maternal pre-pregnancy BMI and child HOMA-IR is further affected by including intermediate

185
Discussion and Conclusion

measures of child body size (e.g., BMI at 2 years or BMI at the age of adiposity rebound) in the
models or more specific measures of adiposity (e.g., %BF or WHtR).

The only comparable study to this has been in adult twins (aged 18-34 years), where low maternal
pre-pregnancy BMI was also associated with hyperinsulinemia and IR in the adult offspring after
adjusting for their current body size (Loos et al. 2002). However, the findings of the present study
are intriguing when compared to a study of infants. Catalano et al. (2009b) reported higher HOMA-
IR at birth in infants of obese mothers compared to those of lean mothers, which correlated strongly
with neonatal adiposity. These conflicting results could suggest, again, that postnatal factors,
including weight gain during childhood are on the pathway between maternal BMI and child IR.

The negative association between maternal BGGI and child IR after controlling for current body size
is more complex to interpret due to the use of categorical variables and the need to consider it in
context with the positive association between maternal GD and child HOMA-IR, which was only
partly mediated by current body size. Potential mediation of the association between maternal GD
and child IR by child current body size has been previously examined in a study, which showed that
adjusting for child current BMI reduced the association in boys and attenuated it to null in girls
(Krishnaveni et al. 2010), indicating perhaps a more important role of body size in determining IR,
compared to that of exposure to GD. This is supported by my finding that the effect of GD on child
HOMA-IR was amplified by child current BMI z-score (there was a positive interaction between
maternal GD and child current BMI z-score in relation to HOMA-IR).

One possible interpretation of the negative association between maternal BGGI and child IR is that
children of mothers with BGGI may have different allocations of fat and lean tissue and, while BMI z-
score did not vary significantly across the groups, there may be enough variation such that adjusting
for it reveals lower HOMA-IR for a given BMI z-score. It could be hypothesised that within this
cohort, fetuses of BGGI mothers were able to convert the surplus glucose into muscle growth, hence
being leaner for a given BMI. It is also prudent when interpreting this result to consider that BGGI
women formed a relatively small group in this cohort. Nonetheless, this is the first study to report this
interesting result and it warrants further investigation.

5.3.6 Two-way interactions between main exposures in relation to child outcomes

Generation 1 data revealed effect modification of maternal pre-pregnancy BMI on child central
adiposity by BGGI, but not GD. The fact that no interaction was detected in this study between

186
Discussion and Conclusion

maternal pre-pregnancy BMI and GD is in contrast to a large prospective cohort study in


adolescents, which has reported higher risk of obesity and central adiposity (based on waist
circumference) in offspring exposed concomitantly to maternal overweight at the time of pregnancy
and GD compared to those exposed only to maternal overweight (Pirkola et al. 2010). This
discrepancy in findings could reflect a lack of power in the current study.

The second interaction identified was between maternal pre-pregnancy BMI and GWG in relation to
child central adiposity. Similar interactions have been reported previously, but not universally, in
relation to child global obesity (Fraser et al. 2010; Oken et al. 2007; Oken et al. 2008; Schack-
Nielsen et al. 2005), but not specifically with central fat deposition.

These interactions suggest there is complexity in the way the intrauterine environment affects growth
and particularly the development of central adiposity. They once again highlight the important role of
maternal pre-pregnancy BMI for child obesity and suggest that by addressing only gestational
glucose intolerance or GWG in isolation, without optimising maternal BMI prior to conception, the
reduction of central obesity in the child would be limited.

5.3.7 Potential underlying mechanisms

The positive associations between maternal body size prior to pregnancy and child body size,
body composition and fat pattern observed in this study could be explained by genetic inheritance,
shared poor nutritional and physical activity behaviours, and/or intrauterine programming through
epigenetic changes. The current study does not allow for disentanglement of these three categories
of effects.

At least a part of the direct association between maternal and child body size is genetically
determined. Genetic predisposition to obesity rarely has a monogenic basis (i.e., a mutation in a
single gene that is sufficient to cause obesity, such as the genes on the leptinergic-melanocortinergic
pathway) and is mainly due to the combined effect of polygenic variants (the individual effect of each
polygenic variant on the phenotype is small) (Hinney et al. 2010). Genome wide association studies
have led to the detection of at least 17 confirmed polygenic variants involved in body weight
regulation, including the fat mass and obesity associated gene (FTO), but many more polygenic
variants await identification (Hinney et al. 2010). Based on longitudinal twin studies, it has been
shown that the genetic influence on BMI variance strengthens during childhood (Haworth et al.
2008). Heritability has been described to influence a number of factors involved in obesity

187
Discussion and Conclusion

development, including appetite, food preferences (Garcia-Bailo et al. 2008) and physical activity
levels (Perusse et al. 1989). However, the rapid rise in childhood obesity is unlikely to be accounted
for by genetic factors, and rather by changes in the environmental factors (Department of Health
Public Health Research Consortium et al. 2007).

Another potential explanation for the association between maternal weight status prior to pregnancy
and child body size could be the shared postnatal environment, including shared patterns of diet and
physical activity. Mothers play an important role in shaping their children’s eating behaviour through
the foods they make available (diet composition), child feeding practices, their own eating behaviour
and parental control (restriction to junk foods and pressure to eat healthy foods) (Scaglioni et al.
2008), with subsequent effects on body weight. It appears that patterns of eating and food
preferences established during childhood are likely to track into adulthood (Fisk et al. 2011), with
further potential impact on body weight. There is accumulating evidence of a direct association
between maternal and child quality of diet, which appears to follow a socioeconomic gradient
(McLeod et al. 2011). Women whose diets include more healthy foods are more likely to have
children with similar dietary patterns, and, conversely, those with high intakes of junk foods or large
portion sizes are more likely to have children with comparable diets (Robinson et al. 2007). In the
same way, mothers may influence their children’s physical activity through shared activities,
modelling, and support (Gustafson and Rhodes 2006). That said, it should be acknowledged that
there is inconsistent evidence regarding parent-child correlations of physical activity (Gustafson and
Rhodes 2006), with some studies indicating similar patterns of physical activity in children and their
parents especially for sedentary activity (Fogelholm et al. 1999; Simonen et al. 2002), while others
not (Anderssen et al. 2006).

While accepting the above alternatives may play a part, biological plausibility of intrauterine
programming of adiposity has been supported by experimental work in various animal models. It has
been postulated that programming of obesity by fetal overnutrition involves persistent abnormalities
in adipocytes (number and size) as well as dysfunction in the hypothalamic regulation of energy
balance (appetite control). Both mechanisms are likely to be mediated by fetal hyperinsulinism
secondary to maternal increased levels of plasma glucose. As suggested by animal studies
(Mühlhäusler and Smith 2009; Mühlhäusler 2007), it could be hypothesised that human fetal
adipocytes exposed to higher levels of glucose during early development may be programmed to
maintain an increased capacity to form new adipocytes and to accumulate lipids in the existing fat
cells.

188
Discussion and Conclusion

Another potential mechanism of intrauterine programming of child obesity may involve lifelong
altered programming of hypothalamic regulation of energy balance, with subsequent increased food
intake and reduced energy expenditure. Fetal insulin appears to play a key role for development of
neuronal networks in the hypothalamus that regulate energy balance (Bouret 2009). It is possible
that hypothalamic insulin resistance secondary to an abnormal hypothalamic development leads to
the release of high levels of orexigenic neurotransmitters (e.g. neuropeptide Y, the strongest
promoter of appetite identified so far) and reduced sensitivity to satiety signals, as indicated by
previous studies in rats (Plagemann et al. 1999). As a consequence, these children may have
exaggerated appetite, with subsequent predisposition to become overweight and accumulate
adipose tissue.

The negative association observed between maternal body size prior to pregnancy and child IR after
considering child’s current body size as a potential mediator might be in fact driven by rapid
postnatal growth in some of these children. The role of rapid postnatal growth in the first few years
of life in predicting IR during childhood or adolescence has been previously shown (Fewtrell et al.
2000; Singhal et al. 2003), with the highest level of IR observed in children born in the lowest birth
weight tertile who became obese (Ong et al. 2004). The mechanisms underlying rapid transition
from small size at birth to overweight or obesity during childhood (rapid catch-up growth) have not
been fully characterised, but they appear to involve increased appetite secondary to high levels of
leptin (Ong et al. 1999), increased number of receptors for insulin in the first years of life (Hales and
Barker 2001) and a suite of factors that promote growth, including insulin-like growth factors (IGF)
and their binding proteins, which regulate not only growth during childhood (Hill and Hogg 1989), but
also insulin sensitivity (Jones and Clemmons 1995, Singhal, 2003 #1612).

The fact that intrauterine exposure to maternal GD was associated with IR in the child, independent
of other maternal factors and only partially mediated by child current body size could be explained by
genetic inheritance of IR or intrauterine programming. Overall heritability for insulin sensitivity has
been estimated at 0.53-0.55 (Shaat and Groop 2007). A number of genes have been identified as
playing a role in cellular mechanisms of IR in GD (e.g., insulin receptor substrate 1, insulin receptor,
ectonucleotide pyrophosphatase/phosphodiesterase 1) (Shaat and Groop 2007). These genes could
be inherited by the offspring, who would be thus predisposed to develop IR.

Potential underlying mechanisms of the intrauterine programming of offspring obesity and IR by


maternal GD also appear to involve fetal hyperinsulinism secondary to the excessive glucose
passage from the mother, which has programming effects on adipocytes development (Mühlhäusler

189
Discussion and Conclusion

and Smith 2009) and on appetite-regulating neuronal network by stimulating the expression of
neuropeptide Y neurons (Plagemann et al. 1992), with subsequent weight gain and, potentially, IR.
(In contrast to previous animal studies, maternal GD in the current study was only associated with
child IR and not with child body size or body composition). The exact mechanisms through which
maternal GD affects child IR and not child body size or composition are not clear, but it could be
posited that defects in the insulin receptor or insulin signalling proteins induced by maternal GD
might be more prominent or longer-lasting compared to the imprinting effects on adipocytes and
hypothalamic regulation of energy balance.

The positive association between maternal GWG and child body size in pre-pubertal years (after
adjustment for maternal body size prior to pregnancy) might reflect genetic inheritance of the
potential to gain weight or programming by fetal overnutrition. Intrauterine programming is supported
by animal models of energy-rich diet during pregnancy, mimicking excessive GWG in humans. For
instance, lambs of sheep that were overfed in late pregnancy had an increased appetite,
hyperglycaemia, hyperinsulinemia, and greater deposition of subcutaneous fat in the postnatal
period, potentially related to reduced expression of the leptin receptor in the hypothalamus and of
appetite-inhibiting neuropeptides (such as cocaine- and amphetamine- regulated transcript) in
response to increases in fat mass (Mühlhäusler et al. 2006).

While acknowledging the interrelationships between the three maternal factors considered in this
study, maternal body size at the time of pregnancy appears to have a greater contribution to the
development of obesity in the next generation compared to gestational glucose intolerance during
pregnancy or GWG, as it profoundly shapes the intrauterine environment.

5.4 Implications and recommendations for public health

From a life course perspective, the intrauterine period is emerging as an important stage during
which maternal factors associated with fetal overnutrition, in particular maternal obesity and
gestational diabetes, appear to contribute to the programming of offspring obesity and insulin
resistance. In this view, the current research suggests that one potential means of preventing
childhood obesity is to optimise maternal body weight status before a woman conceives. Not only
might the child’s risk of global obesity be reduced if born to a mother who was not obese prior to
pregnancy, but other aspects of the child’s adiposity, including the central distribution of fat, might

190
Discussion and Conclusion

also be improved, with potential further benefits for the child’s metabolic health. Moreover, based on
this study, the risk of central adiposity appears even greater in children whose mothers were obese
prior to pregnancy and either developed borderline gestational glucose intolerance or gained more
weight during pregnancy. This latter finding has broader implications, as these maternal conditions
rarely occur in isolation and are rather interrelated. However, as responsibility for implementing
appropriate interventions rests with different sectors, preventative strategies need to be
comprehensive and address them in a co-ordinated multidisciplinary approach.

To address this issue, obesity-related health promotion needs to occur at both individual and
population levels, targeting three specific time windows in a woman’s life: pre-conception, during
pregnancy and postpartum (between pregnancies).

One recommendation flowing from the research presented in this thesis is that overweight and obese
women of reproductive age who are planning pregnancy should receive preconception counselling
with respect to the risks associated with excessive weight (both for the women themselves and their
future offspring). Achieving an optimal nutritional status should be encouraged through healthy diet
and regular physical activity (American Dietetic Association et al. 2009).

However, this focussed approach is limited by the fact that pregnancy is not always planned. In
Australia for instance, although there is no national data on the rates of unintended pregnancy
(Australian Government and Department of Health and Ageing 2009), it has been estimated that
about half (51%) of all women had experienced an unplanned pregnancy at some time during their
reproductive carrier (Marie Stopes International 2006). Similar rates of unintended pregnancy have
been reported in the USA (Finer and Henshaw 2006). Therefore, in order to also capture unintended
pregnancies (and reduce the rates of poor pregnancy outcomes) preconception care could be made
available to all women of reproductive age, irrespective of their intention to conceive. However, an
inadequate uptake of such a broad and inclusive approach could be a limiting factor for its success.

Nonetheless, current recommendations from the US Centers for Disease Control and Prevention and
its Select Panel on Preconception Care stipulate that preconception care should be offered to all
women with a potential to become pregnant (15-44 years), both before a first and a subsequent
pregnancy, aiming to identify and manage women’s medical, behavioural and social risks to their
own health and any pregnancy they may have (Centers for Disease Control and Prevention (CDC)
and Agency for Toxic Substances and Disease Registry (ATSDR) 2006). This primary prevention
approach needs to include, but not be limited to, developing a reproductive life plan, providing risk

191
Discussion and Conclusion

assessment through screening, promoting a healthy weight status, smoking and alcohol cessation,
periconceptional folic acid supplementation, appropriate vaccinations, and management of chronic
medical conditions (Centers for Disease Control and Prevention (CDC) and Agency for Toxic
Substances and Disease Registry (ATSDR) 2006, Royal Australian College of General Practitioners,
2009 #1544).

Only recently has preconception care (including that related to women’s weight status) been included
among the preventive activities in general practice in Australia, broadening the former focus on the
three months preceding a pregnancy (Royal Australian College of General Practitioners 2009). The
effectiveness of its implementation has not been comprehensively evaluated, but data from a
qualitative study have suggested that there is room for improvement in the delivery and uptake of
preconception care (in general) in Australia (Mazza and Chapman 2010). The main enabling factors
identified were women’s willingness and motivation to engage in healthier behaviours in order to
ensure the best possible intrauterine environment for the future child, while the major barriers
included women’s receptivity to preconception care (depending on their life stage), the view that
pregnancy is a normal event (which should not require intervention prior to it occurring), the limited
availability of preconception counselling and the perceived role of their doctors to be one of acute
care (Mazza and Chapman 2010).

More specifically, the outcomes of counselling obese women of childbearing age before becoming
pregnant are still largely unknown, but it seems that about half (46%) of such women report some
weight loss (Callaway et al. 2009), while fewer women manage to achieve normal weight before
conception (American Dietetic Association et al. 2009). This is not surprising given that long-term
weight loss and maintenance is difficult to achieve at any stage of life (Loveman et al. 2011).

A number of barriers in addressing overweight and obesity prior to pregnancy, which might impede
the effectiveness of preconception care, have been described in Australian women from an urban
setting. These include the insufficient uptake of routine pre-pregnancy health checks (about half of
women), the relatively high proportion of unplanned pregnancies (35%) and the inaccurate self-
categorisation of weight, with a tendency to underestimate weight status (and potentially an
inaccurate perception of the associated risks) (Callaway et al. 2009). An intriguing finding of this
study was that only one third of the overweight and obese women having preconception health
checks reported being advised to lose weight (Callaway et al. 2009). While the possibility exists that
some women might have forgotten their doctors’ advice, this finding may also suggest that some of
the doctors did not actually provide lifestyle advice to overweight and obese women prior to

192
Discussion and Conclusion

conception, indicating a need for healthcare providers to become more proactive in providing
preconception counselling. On the other hand, other overweight and obese women participating in
this study reported receiving advice to lose weight not only from their doctors, but also their partners,
family members and friends (Callaway et al. 2009). This highlights the importance of improving
knowledge regarding the benefits of healthy weight during pregnancy in the wider community.

It has been suggested that awareness of the relevance and uptake of preconception care could be
improved by healthcare professionals offering it as part of consultations for other matters, sending
invitation letters to all women of childbearing age, running preconception classes, or making
educational materials with relevant information available in waiting rooms (Mazza and Chapman
2010). Further benefits could arise from promoting those factors that have been identified by
community members themselves as enabling the uptake of preventive health care in general practice
(such as continuity of care, trust and rapport), and addressing barriers (such as the lack of
knowledge regarding the relevance of preventive care, time constraints and cost of consultations)
(Mazza et al. 2011).

However, structured as a clinical initiative only, preconception care cannot be expected to achieve
significant improvements in the women’s health (Moos 2010). Preconception health promotion
rather needs to encompass education and counselling to encourage individual behavioural change
by health professionals from various settings (e.g., general practice physicians, obstetricians,
gynaecologists, nurses, midwives), with further support from community-based public health
programs, effective policies at local and national levels, as well as involvement of financial systems
(Centers for Disease Control and Prevention (CDC) and Agency for Toxic Substances and Disease
Registry (ATSDR) 2006).

A response to the concern that “preconception health is pronatalist and has the potential to frame
women as nothing more than vessels for growing healthy offspring” is the recommendation to
healthcare providers to simultaneously offer non-pregnancy related reasons for adopting healthy
behaviours and emphasise their positive impact on the general health of women themselves, as part
of primary care (Moos 2010). All women deserve this care in their own right, irrespective of their
childbearing desire. Therefore any routine healthcare encounter (for instance, every time women are
screened for cervical cancer) should be regarded as an opportunity to address women’s health
needs and promote women’s own wellness as a whole, including the provision of lifestyle advice to
encourage healthy weight. This strategy would have a beneficial impact on women’s future health by

193
Discussion and Conclusion

reducing obesity-related complications, which in itself is an important outcome; for those women who
do become pregnant, the benefits are likely to extend to the next generation.

Some obese women may enter pregnancy without receiving preconception counselling or without
achieving significant changes in their weight. For these women, pregnancy may offer the opportunity
to intervene, so that their risk of developing complications related to excessive weight, both for
themselves and their children, is minimised. Obese pregnant women are recommended to restrain
the amount of weight they gain (American College of Obstetricians and Gynecologists 2005), but not
to lose weight, which could negatively impact on fetal growth (Dodd and Robinson 2011).

Despite the guidelines proposed by the Institute of Medicine (detailed in Section 2.2.3), there are still
large inconsistencies in the advice regarding optimal GWG according to pre-pregnancy BMI given by
health practitioners (Phelan 2010). Although obese women gain, on average, less weight during
pregnancy compared to normal weight women, they are more likely to exceed the GWG
recommended by IOM, as shown in previous studies (Althuizen et al. 2009; Olson et al. 2003) and
supported by data from the Generation 1 study (Section 4.1.1.4). Difficulty adhering to the
recommended GWG may partly stem from time and financial constraints, and these barriers need to
be taken into account in future health promotion strategies (Phelan 2010).

A recent systematic review of randomised controlled trials of dietary and exercise interventions in
pregnancy for overweight or obese women indicated no statistically significant differences in mean
GWG, large-for-gestational-age infant and other outcomes, between women who did and those who
did not receive the interventions (Dodd et al. 2010). Cognitive behavioural therapy for obesity, which
is known to be effective mainly in the short term (Van Dorsten and Lindley 2008), has not been
evaluated in pregnant women (Denison and Chiswick 2011), but could be a promising way forward
given that it is tailored on individual’s needs, providing practical strategies to address psychological
barriers to the long-term adherence to healthy weight behaviour (Cooper and Fairburn 2001).
Weight loss drugs (e.g., Sibutramine, Orlistat) have not been licensed for use in pregnancy (Denison
and Chiswick 2011) and are not likely to be an acceptable solution. The findings from the ongoing
Australian multicentre RCT evaluating the effectiveness of a combination of periodic dietary, exercise
and behavioural advice in pregnancy to limit gestational weight gain in overweight and obese women
(Dodd et al. 2011), are much awaited.

During pregnancy, many women tend to be motivated to adopt lifestyle changes for the benefit of
their future offspring, and these behaviours are likely to perpetuate postpartum (Phelan 2010).

194
Discussion and Conclusion

Although vigorous physical activity cannot be recommended to women during pregnancy due to the
potential adverse effects on fetal growth (Kennelly et al. 2002), some level of exercise is beneficial
for woman’s cardiovascular function and insulin sensitivity, as well as for fetal growth (Clapp 2006a;
Clapp 2000). However, a number of socio-demographic factors, medical (in particular obstetrical)
history, and pregnancy symptoms have been identified as predictors of lower physical activity level
among pregnant women, which could affect effectiveness of exercise interventions. They include
older age, lower educational attainment social disadvantage in general, obesity, and early pregnancy
symptoms (nausea, vomiting and lower back pain) (Foxcroft et al. 2011).

Therefore, exercise interventions during pregnancy tailored to take into account these factors (and
potentially others as they are identified) and offer greater support to women from lower SES groups
might prove more effective in controlling gestational weight gain.

Based on the current study, maternal GWG does not play an independent role in either child obesity
or insulin resistance. Gestational weight gain rather interacts with maternal pre-pregnancy BMI to
synergistically increase child central adiposity. This latter finding implies that optimising only GWG
would be insufficient to reduce central adiposity in the child and that in addition, achieving an optimal
pre-pregnancy BMI is required. That said, intervening in pregnancy to prevent excessive weight gain
has immediate benefits (e.g., reducing the risk of preeclampsia or caesarean section) and gives the
opportunity to reduce maternal obesity risk prior to subsequent pregnancies, given that GWG (and
not pre-pregnancy weight status) is the strongest determinant of postpartum weight retention (Linné
et al. 2004; Rooney and Schauberger 2002).

Although maternal glucose tolerance status during pregnancy does not appear to influence child
body size and adiposity in this cohort, there is an indication that exposure to maternal GD increases
child insulin resistance. Moreover, if children exposed to maternal GD also become obese, their risk
of developing insulin resistance is even greater than for their counterparts. These findings have
further implications for female offspring, who could perpetuate insulin resistance and GD cycle over
subsequent generations. Management of GD with diet, exercise and insulin when required has been
shown beneficial for preventing perinatal complications (except for labour induction) in both mothers
and children (Alwan et al. 2009), so there is enough evidence to intervene in women with GD.
However, there is no data on long-term outcomes of such interventions, hence it would be
reasonable to recommend organising long-term follow-up of existing trials on GD management to
evaluate the effects on child outcomes beyond birth, in particular on child glucose-insulin
homeostasis. Based on Generation 1 data, one could not suggest the need for intervention in

195
Discussion and Conclusion

women with milder degrees of gestational glucose intolerance, but more research (using data from
larger cohorts) should be done to explore in more detail the influence that maternal BGGI may have
during pregnancy and after birth, birth for the women and their children.

The inter-conception period, that is the period following childbirth and before the next pregnancy,
could also provide an opportune time window for engaging with women in targeted programs to
enhance postpartum weight loss, minimise weight retention and avoid obesity development, which
would further impact on future pregnancies. Any program designed to improve women’s weight
status after childbirth needs to recognise the numerous challenges that women face during this
period, including time constraints, motivation issues (sometimes related to postpartum depression)
and the need for sustained support (Montgomery et al. 2010).

There is no standard definition of what constitutes excessive postpartum weight retention or what an
optimal rate of weight loss would be (Amorim Adegboye et al. 2008). It has been estimated that by
one year postpartum, women retain on average 0.5-4 kg (Linné and Rössner 2003; Olson et al.
2003). This postpartum weight retention appears to be mostly influenced by excessive GWG, low
SES and parity (Linné and Rössner 2003). It has been indicated that women who do not lose the
weight they had gained during pregnancy by 6 months postpartum are more likely to become obese
10 years later (Rooney and Schauberger 2002), suggesting that encouraging postpartum weight loss
would be beneficial.

The best strategy to achieve postpartum weight loss has not been clearly established, mainly due to
inconsistencies related to the potential effects on breastfeeding (quantity and quality of breast milk)
(Amorim Adegboye et al. 2008). A recent systematic review concluded that attention to diet alone or
associated with exercise enhance postpartum weight loss, while exercise alone does not have a
significant effect on weight loss, but improves woman’s cardio-vascular fitness; both approaches are
safe in breastfeeding women (Amorim Adegboye et al. 2008).

Supporting exclusive breastfeeding for at least 6 months (World Health Organization 2001) has been
shown to have multiple health benefits, including favourable effects on women’s body size and their
children’s future metabolic health (Toschke et al. 2007). While a great majority of mothers initiate
breastfeeding early postpartum, with rates exceeding 80% in countries such as Australia and the
USA (Haas et al. 2006; Scott et al. 2006), continuation of breastfeeding during the following few
months remains below the recommended targets, with less than half of the infants still receiving
breast milk by the age of 6 months (Donath and Amir 2000). In other words, identifying effective

196
Discussion and Conclusion

strategies to increase breastfeeding duration has been and continues to be a challenge.


Breastfeeding duration has been observed to be much shorter in obese compared to normal weight
mothers (Oddy et al. 2006), which suggests that more effective support is needed for obese women.
Breastfeeding support may be provided by a range of people with variable involvement in mothers’
lives, including the fathers (particularly when living together), other family members (with frequent
contact with the mother), doctors, midwives, nurses (which should extend beyond the postpartum
period), lactation consultants, breastfeeding support groups, employers (for when breastfeeding
mothers return to work) and last, but not least important, the community, protecting breastfeeding in
public (legislation being in place in some countries, including Australia, UK, and the USA) (Clifford
and McIntyre 2008).

In addition to health promotion in women at an individual level (at all stages of their reproductive life,
during pre-conception period, pregnancy, or postpartum), more public health attention needs to be
directed towards strategies to optimise weight status in women of childbearing age at the population
level, by favourably influencing the obesogenic environment.

Although obesity has been recognised as a global health priority (World Health Organization 2000),
population strategies developed to date to control the risk factors have been largely ineffective
(Gortmaker et al. 2011; Swinburn et al. 2011), as reflected by the persistently high obesity
prevalence in both developed and developing countries (Finucane et al. 2011), with the well known
socio-economic gradient (Adler and Newman 2002).

There is growing evidence that the current obesity epidemic is due to a chronic energy imbalance
gap, secondary mainly to major shifts in the global food system that have occurred over the last
couple of decades (Gortmaker et al. 2011), with the fastest increase in high income countries
(Finucane et al. 2011). These changes are related to the rising consumption of mass prepared food
(as opposed to individual preparation of food in the past) and the greater availability of highly
processed, energy-dense, nutrient-poor foods, which have become cheap, easily accessible and
aggressively promoted by marketing campaigns (Gortmaker et al. 2011).

Based on the evaluation of cost-effectiveness of obesity interventions in adults in Australia (Vos et al.
2010), the two most cost effective interventions at population level are arguably the introduction of
10% tax on unhealthy foods and beverages, and making nutrition information visible at-a-glance on
food packaging to help consumers select healthier foods (‘front-of-pack traffic light nutrition labelling’)
(Sacks et al. 2011). However, it is not clear that such measures could be implemented soon.

197
Discussion and Conclusion

It is increasingly recognised that weight is not only a matter of individual responsibility for food
choices and exercise habits, but also a matter of the society that shapes the living environment,
influencing personal choice. “It takes more than just willpower to maintain a healthy weight – a
strong support system is necessary” (Strategies to Overcome and Prevent (STOP) Obesity Alliance
2010). The STOP Obesity Alliance launched in 2007 also highlights the importance of emphasising
through consistent messages to the public health-related benefits of not being obese rather than
those related to body image (appearance) (Strategies to Overcome and Prevent (STOP) Obesity
Alliance 2010).

A comprehensive approach to obesity prevention needs to encompass strategies at multiple levels of


the society, combining direct initiatives, which rely on the beneficial influence of certain interventions
on energy balance, with documented cost-effectiveness, and indirect, cross-cutting actions, with less
documented cost-effectiveness, which support the implementation of direct interventions (Institute of
Medicine 2010).

As there is no consensus regarding the most effective strategies to prevent and control the complex
issue of obesity worldwide, the need for sustained, integrated efforts to create successful
programmes to lower obesity rates has to be a priority (Australian Government and National
Preventative Health Taskforce 2009b; Gortmaker et al. 2011). This would be possible within a co-
ordinated ‘systems approach’, involving multiple parties: governments, international agencies, private
sector, civil society, health professionals and individuals (Australian Government and National
Preventative Health Taskforce 2009b; Gortmaker et al. 2011).

Governments have been recognised as “the most important actors in reversing the obesity
epidemic” (Gortmaker et al. 2011), as their role is to protect public interest. As recommended by
WHO (World Health Organization 2009b) and outlined in the recent paper by Gortmaker et al.
(2011), core actions for governments to reduce obesity prevalence are related to leadership, healthy
public policies (e.g., enforcement of subsidies for healthy foods and taxes for junk foods; creation of
safe walking, cycling and recreation environments), funding, supporting research in obesity
prevention, workforce development, networks for co-ordination, and communications (e.g.,
establishing targets for the food industry on nutrient composition of foods, and ensuring nutrition
claims comply with a nutrient profiling system). Therefore, commitment of the health system for
obesity prevention needs to be accompanied by actions across a range of other sectors such as
education, agriculture, transportation, urban planning, or finance (Australian Government and
National Preventative Health Taskforce 2009b; Gortmaker et al. 2011).

198
Discussion and Conclusion

Actions regarding obesity prevention and control have also been proposed for international
agencies (e.g., WHO, Food and Agriculture Organisation, World Food Programme) including the
need for global leadership (Gortmaker et al. 2011; World Health Organization 2009b). The private
sector (encompassing food and built environment industries, and media) play a key role in
influencing individuals’ choice regarding their lifestyle, mainly through lobbying activities which often
undermine obesity prevention policies (Gortmaker et al. 2011). The recommended actions include
the development of less processed food products with healthier nutrient composition, voluntary
restrictions on marketing promotions of unhealthy foods, appropriate nutrition labelling of food
packaging (Gortmaker et al. 2011). Civil society (including consumer associations, charities,
foundations, professional associations, sporting clubs) plays an important advocacy role in relation to
promotion of healthier environments and lowering obesity rates (particularly in democratic countries)
(Gortmaker et al. 2011). Health professionals need to monitor their patients’ weight status and
support them to maintain or achieve healthy weight (Gortmaker et al. 2011). Finally, individuals are
encouraged to opt for a healthy lifestyle, by choosing healthy foods and activities (Gortmaker et al.
2011; Swinburn et al. 2011), but they need support from all the other sectors.

A number of issues have been identified and need consideration in relation to the implementation of
policies on obesity prevention and control, such as feasibility (based on the availability of trained
staff, leadership involved and existing programmes), sustainability (influenced by the level of policy
support and funding), effects on equity (with regard to SES, ethnicity, gender), potential side effects
(on other health conditions, household costs) and acceptability to stakeholders (Swinburn et al.
2005). Although high intensity interventions are usually more effective, they reach a limited segment
of the population due to the higher cost, hence low intensity interventions, often with mixed effects,
are considered more feasible at population level, as a greater number of people may benefit from
them at lower costs (Gortmaker et al. 2011).

In Australia, a National Preventative Health Strategy to halt the rise in obesity prevalence has been
recently put forward as part of the initiative of turning Australia into the healthiest country by 2020
(Australian Government and National Preventative Health Taskforce 2009b). In addition to the
multilevel directions for action outlined above, this strategy recommends a staged implementation of
actions, so that subsequent actions can be informed by evidence of effects from the previous
phases, in a cyclical approach (“do, measure, report – do, measure, report”) (Australian Government
and National Preventative Health Taskforce 2009b). For the time being, it appears that Australian
Government has had limited response to the recommendations advanced by the National

199
Discussion and Conclusion

Preventative Health Taskforce in 2009 (Australian Government and National Preventative Health
Taskforce 2009b) in adopting laws that would promote healthy eating, with the food industry being
pressured to self-regulate its actions and practices (MacKay 2011 - in press).

While no specific strategies have been developed for women to prevent weight gain or promote
weight loss, the need for such actions has been flagged in the Technical report on obesity,
particularly in women with children (Australian Government and National Preventative Health
Taskforce 2009a). To be effective in combating obesity among women in community settings, these
strategies would have to be tailored to address personal, social and environmental constraints that
women in particular may be confronted with when attempting to adopt healthy behaviours. A number
of perceived barriers to healthy eating and exercise habits have been reported by young women,
including lack of motivation, time limitations (mainly due to work responsibilities), lack of social
support (particularly in women with children), and cost constraints (Andajani-Sutjahjo et al. 2004).
Getting married, motherhood and commencement of paid work have been associated with lower
physical activity levels in women (Brown and Trost 2003). Among overweight women, other reasons
for not engaging in physical activity were feeling too fat or too embarrassed to exercise (Ball et al.
2000) and feeling less confident about being able to engage in physical activity (Jewson et al. 2008).
Taken together, these findings suggest that community-based interventions to prevent obesity in
women should tackle these perceived barriers, by enhancing women’s motivation, being family-
friendly, and providing social support for adopting healthy behaviours. For overweight women, it has
been recommended that interventions should also consider developing women’s skills and
confidence to improve self-efficacy with regard to their physical activity behaviour (rather than
targeting attitudes through health messages) (Jewson et al. 2008).

An example of a feasible low-intensity community-based strategy to prevent weight gain is the


Healthy Lifestyle Program. A cluster-randomised controlled intervention was conducted in a
population of middle SES women with children, apparently healthy, who were recruited through their
children’s schools in Victoria, Australia (Lombard et al. 2009). The intervention was based on social
cognitive theory (namely goal setting, self monitoring, social support and relapse prevention training)
and consisted of four face-to-face group sessions on behaviour change, followed by ongoing support
for one year (Lombard et al. 2009). With regard to self efficacy, women appeared to be more
confident in making a change towards a healthier diet than towards increasing levels of physical
activity (Lombard et al. 2009). In this study, several factors were identified by women as facilitating
behaviour change, such as intervention delivery by a health professional in a community setting

200
Discussion and Conclusion

(school), various types of reminders (phone, text, mail), and handouts on physical activity and diet
(Lombard et al. 2009).

The US Task Force to prevent obesity in women, comprising almost 20 health advocacy
organisations, is currently aiming to advance understanding among policy makers with regard to the
specific impact obesity has on women’s health and to launch practical initiatives, that are “culturally,
gender and age-appropriate”, with focus on multiple levels (family, community, wider society)
(Strategies to Overcome and Prevent (STOP) Obesity Alliance Task Force on Women 2010).

In summary, this section of the Discussion has outlined arguments for an intergenerational value of
preconception care to all women of reproductive age, with an emphasis on behavioural change
related to optimal weight status. Reasons for implementing strategies for limiting weight gain and
improving glycaemic control during pregnancy were reiterated, with acknowledgement that despite
the interventions trialled in these areas, they remain a largely unsolved challenge. The last body of
this section described the multifaceted approach to obesity prevention and control at a population
level, highlighting the need for co-ordinated action by wider community, civil society, private sector,
local and national authorities.

5.5 Future directions for research

The work presented in this thesis invites further research principally in two key areas: to disentangle
intrauterine programming from genetic or social postnatal influences, and to identify effective
interventions with the potential to optimise maternal weight status prior to conception, and glycaemic
control during pregnancy, with long-term beneficial effects for the offspring.

First, it is worth mentioning that, given the comprehensive data collected so far and the follow-ups
planned for this cohort of children, it would be possible to undertake further work related to the
current project, but beyond the scope of the present study. This may include extending the analyses
carried out in this project, considering programming of other (related) health outcomes by fetal
overnutrition, and investigating the changes in body size, body composition, fat pattern or insulin
resistance longitudinally, in a life course approach. The relationship between maternal factors
associated with fetal overnutrition and child global obesity or insulin resistance could be further
investigated by considering additional potential mediating influence of other factors (that are on the
pathway between the intrauterine environment and child obesity). These factors may include child’s

201
Discussion and Conclusion

birth weight (previously considered in numerous studies, including those by Blair et al. (2007),
Gillman et al. (2003), Lawlor et al. (2007b), Oken et al. (2007)), postnatal growth trajectory (i.e., child
BMI at younger ages) (Crume et al. 2011b), breastfeeding (Blair et al. 2007; Gillman et al. 2003;
Oken et al. 2007; Oken et al. 2008; Reilly et al. 2005), diet, and physical activity (Reilly et al. 2005).

In addition, sex-specific analyses could be performed in order to identify potential differences in the
associations of interest between girls and boys. This approach would be justified by the emerging
evidence of sexual dimorphism in the programming of body composition (Huang et al. 2011;
Labayen et al. 2006) or insulin action (Sugden and Holness 2002). The underlying mechanisms of
these differences are still largely unknown, but they could be linked to the sex-specific pathways
underlying the development of most organs, to different epigenetic marks leading to different gene
expression, and to the influence of sex hormones (Gabory et al. 2009). However, by undertaking
sex-specific analyses, statistical power would be reduced. (In addition to the power-related issue,
sex did not meet criteria for a potential confounder and was therefore not included in the analyses
presented in the current study.)

Although a research question that was beyond the scope of the current project, investigating whether
fetal overnutrition contributes to the programming of child insulin secretion in addition to that of
insulin resistance could help better understand the area of intrauterine programming of impaired
glucose tolerance and type 2 diabetes prior to puberty. There is very little information in the current
literature regarding this aspect (Bush et al. 2011), so more research in this area would be welcome.

Investigating longitudinal changes in child BMI z-score, %BF, WHtR and HOMA-IR in relation to
intrauterine exposures could contribute to better understanding of the life course epidemiology of
obesity, body composition, fat pattern and insulin resistance. This could be achieved by examining
cohort members at later points in time, including at least one follow-up during adolescence, when
some further intrauterine influences on metabolic outcomes might become apparent.

Unravelling intrauterine from genetic or postnatal environmental influences on childhood


obesity and insulin resistance involves clearly demonstrating that maternal exposures of interest are
a direct cause of these outcomes in children. The relatively limited ability to infer causation is a
drawback that applies to any observational study. Although regarded as the highest quality
observational study design, providing level II evidence for aetiology (surpassed only by a systematic
review of level II studies) (National Health and Medical Research Council 2009), a prospective cohort
study can only provide evidence of an association between exposures and outcomes, which may or

202
Discussion and Conclusion

may not be causal. However, as indicated by Bradford Hill (1965), causality may be inferred if the
association of interest is characterised by a combination of strength, consistency upon repetition
across different studies (with different populations, settings and time points), specificity (but
acknowledging that one condition may have multiple aetiologies and that one exposure could
contribute to several outcomes), time sequence, biological gradient (dose-response), biological
plausibility and coherence of explanation, experimental evidence (which makes epidemiological
evidence more convincing), and analogy with other relationships.

In order to consolidate the evidence, as recommended by Bradford Hill (1965), findings presented in
this thesis would benefit from replication in other cohort studies, which should address the main
limitations identified in Section 5.2. Broadly, these approaches could include undertaking similar
work within larger prospective birth cohorts, with enhanced participation at follow-up (compared to
other international cohorts), to permit detection of potential associations between exposures with
relatively low-prevalence (such as GD or BGGI) and obesity or insulin resistance in the offspring. In
addition, glucose levels at OGCT and OGTT could be considered as continuous rather than
categorical variables to predict metabolic risk in the child, given the documented linear relationship
between glucose levels at OGCT and the risk of adverse perinatal outcomes (HAPO Study
Cooperative Research Group 2008).

Theoretically, a randomised controlled trial (RCT) of optimising weight status in women of


childbearing age, with long-term follow-up of the children would provide a more conclusive answer
regarding the possibility of preventing obesity in the offspring than the current cohort study.
Practically, however, such a trial could face several issues, in addition to the challenges related to
costs, time and the constrained external validity. From an ethical point of view, it would seem
inequitable to deprive women randomised to the controlled group of the well-known benefits for the
women themselves of achieving normal BMI prior to becoming pregnant. There could be logistical
problems related to timing of weight loss and timing of the pregnancy (i.e., some women might
become pregnant a long time after the beginning of the trial, or in some cases, never, which would
impact on the timing of the follow-up and, potentially, on sample size). Similarly, expanding on the
beneficial effects of GD treatment for perinatal outcomes (Alwan et al. 2009), a long-term follow-up of
existing RCTs to evaluate the effects of these interventions on child body size, adiposity and, more
importantly, insulin resistance beyond infancy would seem appealing. Also, it appears relevant and
important to conduct more research (larger cohort studies) to explore in more depth the influence of

203
Discussion and Conclusion

milder degrees of gestational glucose intolerance on child body size, adiposity and IR and identify
suitable management for these women.

For studies assessing early life origins of health and disease, characteristically involving a long time
lag between exposures and outcomes assessment, several approaches have been recently
proposed to improve the ability to infer causation and prove fetal programming (Davey Smith 2008).
They include:

 comparing maternal-offspring and paternal-offspring associations: if a stronger maternal-


offspring association is identified, this indicates a specific effect of intrauterine milieu (e.g., fetal
overnutrition in case of maternal obesity or glucose intolerance in pregnancy) on the offspring;

 examining how a certain exposure (e.g. maternal obesity or diabetes) influences child outcomes
when it occurs prior or after pregnancy;

 conducting discordant sibling exposure studies: when one sibling is born prior to the
development of a certain condition in the mother (e.g., diabetes), thus being unexposed, while
the younger sibling is exposed to that particular condition;

 Mendelian randomisation studies, in which maternal genotype is considered an indicator of


environmentally modifiable influences on the intrauterine environment, have the advantage of
minimising residual confounding in examining the association between a given exposure and
the outcomes of interest (Davey Smith 2008). With respect to this project, in order to test the
hypothesis that maternal obesity, glucose intolerance during pregnancy and excessive GWG,
through the modified intrauterine environment, have a causal effect on child obesity and insulin
resistance, it would have been helpful to assess the influence of maternal genetic variants
related to obesity (i.e., FTO gene (Frayling et al. 2007)), blood glucose levels (i.e., glucokinase
gene (Weedon et al. 2006)) or excessive gestational weight gain conditioned on offspring
genotype (Davey Smith 2008).

The above mentioned specialised approaches could not be applied in this study, as necessary data
were not available. (Indeed, such data are currently scarce worldwide.)

The above issues notwithstanding, the current findings provide further support for the well
recognised need to identify effective interventions to prevent and control obesity and GD in
women of childbearing age. Interventions that address optimising weight status prior to conception
in women and improving glucose tolerance status during pregnancy, particularly in obese women,

204
Discussion and Conclusion

could prove beneficial not only for the women themselves, but also for reducing children’s propensity
to gain weight or to develop metabolic perturbations, such as insulin resistance.

There is debate with respect to what constitutes sufficient evidence to translate research findings into
practice in public health. It has been proposed that the concept of evidence should move from the
‘best possible’ evidence that could be provided by RCTs to the ‘best available’ evidence (Glasgow
and Emmons 2007). In principle, an association that appears to be well documented, strong and
consistent in independent studies (Bradford Hill 1965), even if not entirely specific and without a
clear biological underlying mechanism, may be regarded as good evidence for action in the absence
of an RCT (Baum 2008). An example of such an approach is the “Back to sleep” campaign
(American Academy of Pediatrics and Task Force on Infant Sleep Position and Sudden Infant Death
Syndrome 2000), which, based on previous observational (not experimental) studies, raised public
awareness regarding the risk of sudden infant death syndrome associated with prone sleeping and
recommended back sleep position in infants. Likewise, the well established link between smoking
and lung cancer has led to the development of strategies to promote smoking cessation without the
need of confirming the association in a trial (US Department of Health and Human Services 1990).
Translating research findings into practice is a complex process influenced by the relevance of the
health issue, sustainability of behaviour change (Colditz and Taylor 2010), associated costs (both of
action and inaction) (Glasgow and Emmons 2007), and involves advocacy and lobbying (Baum
2008). Existing epidemiological evidence on the adverse effects of obesity and glucose intolerance
during pregnancy (in the short- and the long-term, both for the women and their children) arguably
deserves to be translated into practical approaches to optimise women’s weight and improve their
glucose control during pregnancy.

5.6 Conclusion

Located within the paradigm of developmental origins of health and disease and building on data
collected from a contemporary prospective birth cohort study in Australia, this thesis has presented
findings for early (in utero) origins of obesity and insulin resistance in pre-pubertal children,
particularly in relation to maternal obesity and glucose intolerance during pregnancy. The current
study has extended the earlier data on the deleterious effects of maternal pre-pregnancy obesity on
child global obesity before puberty by establishing that even more specific measures of adiposity,
such as percentage body fat and central adiposity, are influenced by maternal obesity prior to

205
Discussion and Conclusion

pregnancy, taking into account potential confounders. Another novel contribution of this study is the
finding of an exacerbated risk of accumulating abdominal fat in children whose mothers had greater
body size prior to pregnancy and either gained more weight during pregnancy or developed
borderline gestational glucose intolerance. This research has investigated for the first time the long-
term influences of maternal gestational glucose intolerance across the spectrum, from borderline to
GD, on child obesity and insulin resistance. Although global obesity and adiposity in pre-pubertal
children do not appear to be associated with maternal glucose tolerance status during pregnancy,
insulin resistance seems to originate, at least in part, from intrauterine exposure to maternal GD and
these perturbations in glucose-insulin homeostasis are exacerbated if the child becomes obese.

The findings generated from this thesis contribute to the body of evidence in relation to major public
health implications for the prevention of childhood obesity and related metabolic disorders, which are
increasingly observed in Western societies. While addressing factors that occur during pregnancy,
such as glucose intolerance and weight gain, is, without doubt, important, more public health
strategies need to be developed to promote the benefits of women conceiving at an optimal weight
status, both for the mother and the child in the future.

206
References

Aaltonen R, Heikkinen T, Hakala K, Laine K, Alanen A (2005) Transfer of proinflammatory cytokines across
term placenta. Obstet Gynecol 106(4):802-807.
Adler NE, Newman K (2002) Socioeconomic disparities in health: Pathways and policies. Health Aff 21(2):60-76.
Aggoun Y, Farpour-Lambert NJ, Marchand LM, Golay E, Maggio ABR, Beghetti M (2008) Impaired endothelial
and smooth muscle functions and arterial stiffness appear before puberty in obese children and are
associated with elevated ambulatory blood pressure. Eur Heart J 29(6):792-799.
Al-Mahmood AK, Ismail AA, Rashid FA, Azwany YN, Singh R, Gill G (2007) Effect of therapeutic lifestyle
changes on insulin sensitivity of non-obese hyperlipidemic subjects: Preliminary report. J Atheroscler
Thromb 14(3):122-127.
Al-Shawwa BA, Al-Huniti NH, DeMattia L, Gershan W (2007) Asthma and insulin resistance in morbidly obese
children and adolescents. J Asthma 44(6):469-473.
Alberti G, Zimmet P, Shaw J, Bloomgarden Z, Kaufman F, Silink M (2004) Type 2 diabetes in the young: the
evolving epidemic: The International Diabetes Federation Consensus Workshop. Diabetes Care
27(7):1798-1811.
Alberti KGMM, Zimmet P, Shaw J (2005) The metabolic syndrome - a new worldwide definition. The Lancet
366(9491):1059-1062.
Alemzadeh R, Kichler J, Babar G, Calhoun M (2008) Hypovitaminosis D in obese children and adolescents:
Relationship with adiposity, insulin sensitivity, ethnicity, and season. Metabolism 57(2):183-191.
Alessi MC, Morange P, Juhan-Vague I (2000) Fat cell function and fibrinolysis. Horm Metab Res 32(11-
12):504-508.
Allen VM, Wilson RD, Cheung A (2006) Pregnancy outcomes after assisted reproductive technology. J Obstet
Gynaecol Can 28(3):220-250.
Althuizen E, van Poppel MNM, Seidell JC, van Mechelen W (2009) Correlates of absolute and excessive
weight gain during pregnancy. J Womens Health 18(10):1559-1566.
Alwan N, Tuffnell DJ, West J (2009) Treatments for gestational diabetes. Cochrane Database Syst Rev
8(3):CD003395.
American Academy of Pediatrics, Task Force on Infant Sleep Position and Sudden Infant Death Syndrome
(2000) Changing concepts of sudden infant death syndrome: Implications for infant sleeping
environment and sleep position. Pediatrics 105(3):650-656.
American College of Obstetricians and Gynecologists (2005) ACOG Committee Opinion number 315,
September 2005. Obesity in pregnancy. Obstet Gynecol 106(3):671-675.
American Diabetes Association (2000) Type 2 diabetes in children and adolescents. Diabetes Care 23(3):381-
389.
American Diabetes Association (2002) Gestational diabetes mellitus. Diabetes Care 25(Suppl 1):S94-96.
American Diabetes Association (2006) Diagnosis and classification of diabetes mellitus. Diabetes Care
29(Suppl 1):S43-48.

207
References

American Dietetic Association, American Society of Nutrition, Siega-Riz AM, King JC (2009) Position of the
American Dietetic Association and American Society for Nutrition: Obesity, reproduction, and
pregnancy outcomes. J Am Diet Assoc 109(5):918-927.
Amorim Adegboye AR, Linné Y, Laurenco PMC (2008) Diet or exercise, or both, for weight reduction in
women after childbirth. Cochrane Database Syst Rev(3):CD005627.
Andajani-Sutjahjo S, Ball K, Warren N, Inglis V, Crawford D (2004) Perceived personal, social and
environmental barriers to weight maintenance among young women: A community survey. Int J
Behav Nutr Pysical Act 1(15).
Anderssen N, Wold B, Torsheim T (2006) Are parental health habits transmitted to their children? An eight
year longitudinal study of physical activity in adolescents and their parents. J Adolesc 29(4):513-524.
Anna V, van der Ploeg HP, Cheung NW, Huxley RR, Bauman AE (2008) Sociodemographic correlates of the
increasing trend in prevalence of gestational diabetes mellitus in a large population of women
between 1995 and 2005. Diabetes Care 31(12):2288-2293.
Arslanian SA (2005) Clamp techniques in paediatrics: What have we learned? Horm Res 64(Suppl 3):S16-24.
Asbee SM, Jenkins TR, Butler JR, White J, Elliot M, Rutledge A (2009) Preventing excessive weight gain
during pregnancy through dietary and lifestyle counseling: A randomized controlled trial. Obstet
Gynecol 113(2 Pt 1):305-312.
Ashwell M, Cole TJ, Dixon AK (1985) Obesity: New insight into the anthropometric classification of fat
distribution shown by computed tomography. Br Med J (Clin Res Ed) 290(6483):1692-1694.
Ashwell M, Lejeune S, McPherson K (1996) Ratio of waist circumference to height may be better indicator of
need for weight management BMJ 312(7027):377.
Ashwell M, Hsieh SD (2005) Six reasons why the waist-to-height ratio is a rapid and effective global indicator
for health risks of obesity and how its use could simplify the international public health message on
obesity. Int J Food Sci Nutr 56(5):303-307.
Australian Bureau of Statistics (2001) Socioeconomic Indexes for Areas, Canberra: ABS.
Australian Bureau of Statistics (2006) Births, Australia 2005, Canberra: ABS.
Australian Bureau of Statistics (2009a) Australian Social Trends: Children who are overweight or obese,
Canberra: ABS.
Australian Bureau of Statistics (2009b) National Health Survey: Summary of results, 2007-2008 (reissue),
Canberra: ABS.
Australian Government, Department of Health and Ageing (2009) Reproductive health: Findings from the
Australian Longitudinal Study on Women’s Health, Canberra: Australian Government Publishing
Service.
Australian Government, National Preventative Health Taskforce (2009a) Australia: The healthiest country by
2020 - Technical Report 1 - Obesity in Australia: a need for urgent action, Canberra: Australian
Government Publishing Service.
Australian Government, National Preventative Health Taskforce (2009b) Australia: The healthiest country by
2020 - National Preventative Health Strategy - Overview, Canberra: Australian Government
Publishing Service.
Australian Institute of Health and Welfare (2000) Australia's mothers and babies, 1998 (Nassar N, Sullivan
EA, Lancaster P, Day P, eds), Sydney: AIHW: National Perinatal Statistics Unit.
Australian Institute of Health and Welfare (2006) The national system for monitoring diabetes in Australia
(Dixon T, Webbie K, eds), Canberra: AIHW.
Australian Institute of Health and Welfare (2008) Gestational diabetes mellitus in Australia, 2005-2006
(Templeton M, Pieris-Caldwell I, eds), Canberra: AIHW.

208
References

Australian Institute of Health and Welfare (2009) Australia's mothers and babies, 2007 (Laws P, Sullivan E,
eds), Sydney: AIHW: National Perinatal Statistics Unit.
Australian Institute of Health and Welfare (2010) Australia's mothers and babies, 2008 (Laws PJ, Li Z,
Sullivan EA, eds), Canberra: AIHW: National Perinatal Statistics Unit.
Azcona C, Köek N, Frühbeck G (2006) Fat mass by air-displacement plethysmography and impedance in
obese/non-obese children and adolescents. Int J Pediatr Obes 1(3):176-182.
Ball K, Crawford D, Owen N (2000) Obesity as a barrier to physical activity. Aust N Z J Public Health
24(3):331-333.
Banach A, Peralta-Huertas J, Livingstone K, Petrella N, Klentrou P, Faught B, Wade T, O’Leary D (2010)
Arterial distensibility is reduced in overweight pre- and early pubescent children. Eur J Pediatr
169(6):695-703.
Bandini LG, Vu DM (1997) Body fatness and bioelectrical impedance in non-obese pre-menarcheal girls:
Comparison to anthropometry and evaluation of predictive equations. Eur J Clin Nutr 51(10):673-
677.
Barbour LA, McCurdy CE, Hernandez TL, Kirwan JP, Catalano PM, Friedman JE (2007) Cellular mechanisms
for insulin resistance in normal pregnancy and gestational diabetes. Diabetes Care 30(Suppl
2):S112-S119.
Barker DJP, Osmond C (1986) Infant mortality, childhood nutrition, and ischaemic heart disease in England
and Wales. The Lancet 327(8489):1077-1081.
Barker DJP (1994) Mothers, babies and disease in later life. London: BMJ Publishing Group.
Barker DJP (1995) Intrauterine programming of adult disease. Mol Med Today 1(9):418-423.
Barker DJP (2003) The developmental origins of adult disease. Eur J Epidemiol 18(8):733-736.
Baum F (2008) Epidemiology and public health. In: The new public health (Baum F, ed), 3rd ed., Melbourne,
Victoria: Oxford University Press, 156-167.
Beardsall K, Diderholm BMS, Dunger DB (2008) Insulin and carbohydrate metabolism. Best Pract Res Clin
Endocrinol Metab 22(1):41-55.
Beischer NA, Wein P, Sheedy MT, Steffen B (1996) Identification and treatment of women with
hyperglycaemia diagnosed during pregnancy can significantly reduce perinatal mortality rates. Aust
NZ J Obst Gynaecol 36(3):239-247.
Ben-Shlomo Y, Kuh D (2002) A life course approach to chronic disease epidemiology: Conceptual models,
empirical challenges and interdisciplinary perspectives. Int J Epidemiol 31(2):285-293.
Berenson GS, Wattigney WA, Tracy RE, Newman III WP, Srinivasan SR, Webber LS, Dalferes Jr ER, Strong
JP (1992) Atherosclerosis of the aorta and coronary arteries and cardiovascular risk factors in
persons aged 6 to 30 years and studied at necropsy (The Bogalusa Heart Study). Am J Cardiol
70(9):851-858.
Berkey CS, Rockett HRH, Field AE, Gillman MW, Frazier AL, Camargo CA, Colditz GA (2000) Activity, dietary
intake, and weight changes in a longitudinal study of preadolescent and adolescent boys and girls.
Pediatrics 105(4):e56.
Berry WD, Feldman S (1985) Multiple regression in practice, Vol. 7-50. Newbury Park, CA: Sage University
Paper.
Beyerlein A, Toschke AM, von Kries R (2010) Risk factors for childhood overweight: shift of the mean body
mass index and shift of the upper percentiles: Results from a cross-sectional study. Int J Obes
34(4):642-648.
Biro FM, Khoury P, Morrison JA (2006) Influence of obesity on timing of puberty. Int J Androl 29(1):272-277.

209
References

Blair NJ, Thompson JMD, Black PN, Becroft DMO, Clark PM, Han DY, Robinson E, Waldie KE, Wild CJ,
Mitchell EA (2007) Risk factors for obesity in 7-year-old European children: The Auckland Birthweight
Collaborative Study. Arch Dis Child 92(10):866-871.
Bo S, Menato G, Lezo A, Signorile A, Bardelli C, De Michieli F, Massobrio M, Pagano G (2001) Dietary fat and
gestational hyperglycaemia. Diabetologia 44(8):972-978.
Bo S, Menato G, Bardelli C, Lezo A, Signorile A, Repetti E, Massobrio M, Pagano G (2002) Low
socioeconomic status as a risk factor for gestational diabetes. Diabetes Metab 28(2):139-140.
Bo S, Marchisio B, Volpiano M, Menato G, Pagano G (2003) Maternal low birth weight and gestational
hyperglycemia. Gynecol Endocrinol 17(2):133-136.
Bo S, Menato G, Gallo M-L, Bardelli C, Lezo A, Signorile A, Gambino R, Cassader M, Massobrio M, Pagano
G (2004) Mild gestational hyperglycemia, the metabolic syndrome and adverse neonatal outcomes.
Acta Obstet Gynecol Scand 83(4):335-340.
Boerschmann H, Pfluger M, Henneberger L, Ziegler A-G, Hummel S (2010) Prevalence and predictors of
overweight and insulin resistance in offspring of mothers with gestational diabetes mellitus. Diabetes
Care 33(8):1845-1849.
Boettcher M, Machann J, Stefan N, Thamer C, Häring H-U, Claussen CD, Fritsche A, Schick F (2009)
Intermuscular adipose tissue (IMAT): Association with other adipose tissue compartments and insulin
sensitivity. J Magn Reson Imaging 29(6):1340-1345.
Bonen A, Campagna P, Gilchrist L, Young DC, Beresford P (1992) Substrate and endocrine responses during
exercise at selected stages of pregnancy. J Appl Physiol 73(1):134-142.
Boney CM, Verma A, Tucker R, Vohr BR (2005) Metabolic syndrome in childhood: Association with birth
weight, maternal obesity, and gestational diabetes mellitus. Pediatrics 115(3):e290-296.
Bonomo M, Corica D, Mion E, Goncalves D, Motta G, Merati R, Ragusa A, Morabito A (2005) Evaluating the
therapeutic approach in pregnancies complicated by borderline glucose intolerance: A randomized
clinical trial. Diabet Med 22(11):1536-1541.
Bottalico JN (2007) Recurrent gestational diabetes: risk factors, diagnosis, management, and implications.
Semin Perinatol 31(3):176-184.
Botton J, Heude B, Kettaneh A, Borys J-M, Lommez A, Bresson J-L, Ducimetiere P, Charles M-A (2007)
Cardiovascular risk factor levels and their relationships with overweight and fat distribution in
children: The Fleurbaix Laventie Ville Santé II study. Metabolism 56(5):614-622.
Bouret SG (2009) Early life origins of obesity: Role of hypothalamic programming. J Pediatr Gastroenterol
Nutr 48:S31-S38.
Bradford Hill A (1965) The environment and disease: association or causation? Proc R Soc Med 58:295-300.
Branchtein L, Schmidt MI, Matos MCG, Yamashita T, Pousada JMDC, Duncan BB, the Brazilian Gestational
Diabetes Study Group (2000) Short stature and gestational diabetes in Brazil. Diabetologia
43(7):848-851.
Breslow NE (1996) Generalized linear models: Checking assumptions and strengthening conclusions.
Statistica Applicata 8:23-41.
Brion M-J, Zeegers M, Jaddoe V, Verhulst F, Tiemeier H, Lawlor DA, Smith GD (2011) Intrauterine effects of
maternal prepregnancy overweight on child cognition and behavior in 2 cohorts. Pediatrics
127(1):e202-e211.
Britz B, Siegfried W, Ziegler A, Lamertz C, Herpertz-Dahlmann BM, Remschmidt H, Wittchen HU, Hebebrand
J (2000) Rates of psychiatric disorders in a clinical study group of adolescents with extreme obesity
and in obese adolescents ascertained via a population based study. Int J Obes Relat Metab Disord
24(12):1707-1714.

210
References

Brown WJ, Trost SG (2003) Life transitions and changing physical activity patterns in young women. Am J
Prev Med 25(2):140-143.
Browning JD, Horton JD (2004) Molecular mediators of hepatic steatosis and liver injury. J Clin Invest
114(2):147-152.
Browning LM, Hsieh SD, Ashwell M (2010) A systematic review of waist-to-height ratio as a screening tool for
the prediction of cardiovascular disease and diabetes: 0.5 could be a suitable global boundary value.
Nutr Res Rev 23(2):247-269.
Buchanan TA (2001) Pancreatic B-cell defects in gestational diabetes: Implications for the pathogenesis and
prevention of type 2 diabetes. J Clin Endocrinol Metab 86(3):989-993.
Buchanan TA, Xiang AH (2005) Gestational diabetes mellitus. J Clin Invest 115(3):485-491.
Bush NC, Chandler-Laney PC, Rouse DJ, Granger WM, Oster RA, Gower BA (2011) Higher maternal
gestational glucose concentration is associated with lower offspring insulin sensitivity and altered β-
cell function. J Clin Endocrinol Metab 96(5):E803-E809.
Callaway LK, Prins JB, Chang AM, McIntyre HD (2006) The prevalence and impact of overweight and obesity
in an Australian obstetric population. Med J Aust 184(2):56-59.
Callaway LK, O'Callaghan MJ, McIntyre HD (2009) Barriers to addressing overweight and obesity before
conception. Med J Aust 191(8):425-428.
Cameron AJ, Welborn TA, Zimmet PZ, Dunstan DW, Owen N, Salmon J, Dalton M, Jolley D, Shaw JE (2003)
Overweight and obesity in Australia: The 1999-2000 Australian Diabetes, Obesity and Lifestyle Study
(AusDiab). Med J Aust 178(9):427-432.
Cañete R, Gil-Campos M, Aguilera C, Gil A (2007) Development of insulin resistance and its relation to diet in
the obese child. Eur J Nutr 46(4):181-187.
Carpenter MW, Coustan DR (1982) Criteria for screening for gestational diabetes. Am J Obstet Gynecol
144(7):768-773.
Carr DB, Utzschneider KM, Hull RL (2004) Intraabdominal fat is a major determinant of the National
Cholesterol Education Program Adult Treatment Panel III Criteria for the Metabolic Syndrome.
Diabetes 53:2087-2094.
Catalano P, Hoegh M, Minium J, Huston-Presley L, Bernard S, Kalhan S, Hauguel-De Mouzon S (2006)
Adiponectin in human pregnancy: implications for regulation of glucose and lipid metabolism.
Diabetologia 49(7):1677-1685.
Catalano PM, Roman NM, Tyzbir ED, Merritt AO, Driscoll P, Amini SB (1993a) Weight gain in women with
gestational diabetes. Obstet Gynecol 81(4):523-528.
Catalano PM, Tyzbir ED, Wolfe RR, Calles J, Roman NM, Amini SB, Sims EA (1993b) Carbohydrate
metabolism during pregnancy in control subjects and women with gestational diabetes. Am J Physiol
Endocrinol Metab 264(1):E60-E67.
Catalano PM, Drago NM, Amini SB (1995) Maternal carbohydrate metabolism and its relationship to fetal
growth and body composition. Am J Obstet Gynecol 172(5):1464-1470.
Catalano PM, Roman-Drago NM, Amini SB, Sims EAH (1998) Longitudinal changes in body composition and
energy balance in lean women with normal and abnormal glucose tolerance during pregnancy. Am J
Obstet Gynecol 179(1):156-165.
Catalano PM, Huston L, Amini SB, Kalhan SC (1999) Longitudinal changes in glucose metabolism during
pregnancy in obese women with normal glucose tolerance and gestational diabetes mellitus. Am J
Obstet Gynecol 180(4):903-916.
Catalano PM, Kirwan JP, Haugel-de Mouzon S, King J (2003a) Gestational diabetes and insulin resistance:
Role in short- and long-term implications for mother and fetus. J Nutr 133(5):1674S-1683S.

211
References

Catalano PM, Thomas A, Huston-Presley L, Amini SB (2003b) Increased fetal adiposity: A very sensitive
marker of abnormal in utero development. Am J Obstet Gynecol 189(6):1698-1704.
Catalano PM, Farrell K, Thomas A, Huston-Presley L, Mencin P, de Mouzon SH, Amini SB (2009a) Perinatal
risk factors for childhood obesity and metabolic dysregulation. Am J Clin Nutr 90(5):1303-1313.
Catalano PM, Presley L, Minium J, Hauguel-de Mouzon S (2009b) Fetuses of obese mothers develop insulin
resistance in utero. Diabetes Care 32(6):1076-1080.
Catalano PM (2010) Obesity, insulin resistance, and pregnancy outcome. Reproduction 140(3):365-371.
Cedergren M (2006) Effects of gestational weight gain and body mass index on obstetric outcome in Sweden.
Int J Gynecol Obstet 93(3):269-274.
Centers for Disease Control and Prevention (2010) 2009 Pregnancy nutrition surveillance - Summary of
trends in maternal health indicators by race/ethnicity. Table 20D, Atlanta: CDC.
Centers for Disease Control and Prevention (CDC), Agency for Toxic Substances and Disease Registry
(ATSDR) (2006) Recommendations to improve preconception health and health care—United
States: A report of the CDC/ATSDR Preconception Care Work Group and the Select Panel on
Preconception Care. In: MMWR Recomm Rep. 55 [RR-06], Atlanta: CDC.
Chagnon YC, Rankinen T, Snyder EE, Weisnagel SJ, Perusse L, Bouchard C (2003) The human obesity
gene map: The 2002 update. Obesity 11(3):313-367.
Challier JC, Basu S, Bintein T, Minium J, Hotmire K, Catalano PM, Hauguel-de Mouzon S (2008) Obesity in
pregnancy stimulates macrophage accumulation and inflammation in the placenta. Placenta
29(3):274-281.
Chan A, Scott J, Nguyen A-M, Keane R (2000) Pregnancy outcome in South Australia 1999: Pregnancy
Outcome Unit, Epidemiology Branch, Department of Human Services, Adelaide, SA.
Chandler-Laney PC, Bush NC, Rouse DJ, Mancuso MS, Gower BA (2011) Maternal glucose concentration
during pregnancy predicts fat and lean mass of prepubertal offspring. Diabetes Care 34(3):741-745.
Chandra PC, Schiavello HJ, Ravi B, Weinstein AG, Hook FB (2002) Pregnancy outcomes in urban teenagers.
Int J Gynecol Obstet 79(2):117-122.
Cheung WN, Byth K (2003) Population health significance of gestational diabetes. Diabetes Care 26(7):2005-
2009.
Chevalier S, Marliss EB, Morais JA, Lamarche M, Gougeon R (2005) Whole-body protein anabolic response
is resistant to the action of insulin in obese women. Am J Clin Nutr 82(2):355-365.
Chiarelli F, Marcovecchio ML (2008) Insulin resistance and obesity in childhood. Eur J Endocrinol
159(suppl_1):S67-74.
Cho NH, Silverman BL, Rizzo TA, Metzger BE (2000) Correlations between the intrauterine metabolic
environment and blood pressure in adolescent offspring of diabetic mothers. J Pediatr 136(5):587-
592.
Chu SY, Bachman DJ, Callaghan WM, Whitlock EP, Dietz PM, Berg CJ, O'Keeffe-Rosetti M, Bruce FC,
Hornbrook MC (2008) Association between obesity during pregnancy and increased use of health
care. N Engl J Med 358(14):1444-1453.
Chu SY, Callaghan WM, Bish CL, D'Angelo D (2009) Gestational weight gain by body mass index among US
women delivering live births, 2004-2005: Fueling future obesity. Am J Obstet Gynecol
200(3):271.e271-271.e277.
Clapp JF, Little KD (1995) Effect of recreational exercise on pregnancy weight gain and subcutaneous fat
deposition. Med Sci Sports Exerc 27(2):170-177.
Clapp JF (2002) Maternal carbohydrate intake and pregnancy outcome. Proc Nutr Soc 61(1):45-50.

212
References

Clapp JF (2006a) Influence of endurance exercise and diet on human placental development and fetal growth.
Placenta 27(6-7):527-534.
Clapp JF (2006b) Effects of diet and exercise on insulin resistance during pregnancy. Metab Syndr Relat
Disord 4(2):84-90.
Clapp JFr (2000) Exercise during pregnancy. A clinical update. Clin Sports Med 19(2):273-286.
Clausen T, Øyen N, Henriksen T (2006) Pregnancy complications by overweight and residential area. A
prospective study of an urban Norwegian cohort. Acta Obstet Gynecol Scand 85(5):526-533.
Cleary J, Daniells S, Okely AD, Batterham M, Nicholls J (2008) Predictive validity of four bioelectrical
impedance equations in determining percent fat mass in overweight and obese children. J Am Diet
Assoc 108(1):136-139.
Clifford J, McIntyre E (2008) Who supports breastfeeding? Breastfeed Rev 16(2):9-19.
Colditz GA, Taylor PR (2010) Prevention trials: Their place in how we understand the value of prevention
strategies. Ann Rev Public Health 31(1):105-120.
Cole TJ (1988) Fitting smoothed centile curves to reference data. J Royal Stat Soc Series A (Statistics in
Society) 151(3):385-418.
Cole TJ, Green PJ (1992) Smoothing reference centile curves: The lms method and penalized likelihood. Stat
Med 11(10):1305-1319.
Cole TJ, Freeman JV, Preece MA (1995) Body mass index reference curves for the UK, 1990. Arch Dis Child
73(1):25-29.
Cole TJ (2000) Sympercents: symmetric percentage differences on the 100 log e scale simplify the
presentation of log transformed data. Statistics in Medicine 19(22):3109-3125.
Cole TJ, Bellizzi MC, Flegal KM, Dietz WH (2000) Establishing a standard definition for child overweight and
obesity worldwide: International survey. BMJ 320(7244):1240-1245.
Cole TJ (2005) RE: "Why evidence for the fetal origins of adult disease might be a statistical artifact: the
'reversal paradox' for the relation between birth weight and blood pressure in later life". Am J
Epidemiol 162(4):394-395.
Colomiere M, Permezel M, Lappas M (2010) Diabetes and obesity during pregnancy alter insulin signalling
and glucose transporter expression in maternal skeletal muscle and subcutaneous adipose tissue. J
Mol Endocrinol 44(4):213-223.
Committee on the Impact of Pregnancy Weight on Maternal and Child Health (2007) Influence of pregnancy
weight on maternal and child health: Workshop report, Washington, DC: National Research Council
Institute of Medicine.
Conwell LS, Trost SG, Brown WJ, Batch JA (2004) Indexes of insulin resistance and secretion in obese
children and adolescents. Diabetes Care 27(2):314-319.
Cook S, Weitzman M, Auinger P, Nguyen M, Dietz WH (2003) Prevalence of a metabolic syndrome
phenotype in adolescents. Findings from the Third National Health and Nutrition Examination Survey.
Arch Pediatr Adolesc Med 157(8):821-827.
Cooper LG, Leland NL, Alexander G (1995) Effect of maternal age on birth outcomes among adolescents.
Soc Biol 42(1-2):22-35.
Cooper Z, Fairburn CG (2001) A new cognitive behavioural approach to the treatment of obesity. Behav Res
Therapy 39(5):499-511.
Corrado F, D'Anna R, Cannata ML, Caputo F, Rizzo P, Cannizzaro D, Di Benedetto A (2006) Prevalence of
risk factors in the screening of carbohydrate intolerance in pregnancy. Nutr Metab Cardiovasc Dis
16(1):79-80.

213
References

Cox BD, Whichelow MJ, Ashwell M, Prevost AT, Lejeune SRE (1997) Association of anthropometric indices
with elevated blood pressure in British adults. Int J Obes 21(8):674-680.
Crowther CA, Hiller JE, Moss JR, McPhee AJ, Jeffries WS, Robinson JS (2005) Effect of treatment of
gestational diabetes mellitus on pregnancy outcomes. N Engl J Med 352(24):2477-2486.
Crume T, Ogden L, West N, Vehik K, Scherzinger A, Daniels S, McDuffie R, Bischoff K, Hamman R, Norris J,
Dabelea D (2011a) Association of exposure to diabetes in utero with adiposity and fat distribution in
a multiethnic population of youth: The Exploring Perinatal Outcomes among Children (EPOCH)
Study. Diabetologia 54(1):87-92.
Crume TL, Ogden L, Daniels S, Hamman RF, Norris JM, Dabelea D (2011b) The impact of in utero exposure
to diabetes on childhood body mass index growth trajectories: The EPOCH study. J Pediatr
158(6):941-946.
Cruz ML, Bergman RN, Goran MI (2002) Unique effect of visceral fat on insulin sensitivity in obese Hispanic
children with a family history of type 2 diabetes. Diabetes Care 25(9):1631-1636.
Cruz ML, Weigensberg MJ, Huang TTK, Ball G, Shaibi GQ, Goran MI (2004) The metabolic syndrome in
overweight Hispanic youth and the role of insulin sensitivity. J Clin Endocrinol Metab 89(1):108-113.
Cutfield WS, Jefferies CA, Jackson WE, Robinson EM, Hofman PL (2003) Evaluation of HOMA and QUICKI
as measures of insulin sensitivity in prepubertal children. Pediatr Diabetes 4(3):119-125.
Dabelea D, Hanson RL, Lindsay RS, Pettitt DJ, Imperatore G, Gabir MM, Roumain J, Bennett PH, Knowler
WC (2000) Intrauterine exposure to diabetes conveys risks for type 2 diabetes and obesity: A study
of discordant sibships. Diabetes 49(12):2208-2211.
Daniels SR (2009) Complications of obesity in children and adolescents. Int J Obes 33(S1):S60-S65.
Davey RX, Hamblin P (2001) Selective versus universal screening for gestational diabetes mellitus: An
evaluation of predictive risk factors. Med J Aust 174(3):118-121.
Davey Smith G (2008) Assessing intrauterine influences on offspring health outcomes: Can epidemiological
studies yield robust findings? Basic Clin Pharmacol Toxicol 102(2):245-256.
Dawes MG, Grudzinskas JG (1991) Patterns of maternal weight gain in pregnancy. Br J Obstet Gynaecol
98(2):195-201.
Denison FC, Chiswick C (2011) Improving pregnancy outcome in obese women. Proc Nutr Soc 70(4):457-
464.
Department of Health Public Health Research Consortium, Law C, Power C, Graham H, Merrick D (2007)
Obesity and health inequalities. Obesity Reviews 8(Suppl 1):S19-22.
Despres J-P, Lemieux I (2006) Abdominal obesity and metabolic syndrome. Nature 444(7121):881-887.
Deurenberg P, Pieters JJL, Hautvast JGAJ (1990) The assessment of the body fat percentage by skinfold
thickness measurements in childhood and young adolescence. Br J Nutr 63(2):293-303.
Diaz EO, Villar J, Immink M, Gonzales T (1989) Bioimpedance or anthropometry? Eur J Clin Nutr 43(2):129-
137.
Dietz WH (1998) Health consequences of obesity in youth: Childhood predictors of adult disease. Pediatrics
101(3):518-525.
Dishy V, Gupta S, Landau R, Xie HG, Kim RB, Smiley RM, Byrne DW, Wood AJ, Stein CM (2003) G-protein
beta(3) subunit 825 C/T polymorphism is associated with weight gain during pregnancy.
Pharmacogenetics 13(4):241-242.
Dodd J, Turnbull D, McPhee A, Wittert G, Crowther C, Robinson J (2011) Limiting weight gain in overweight
and obese women during pregnancy to improve health outcomes: The LIMIT randomised controlled
trial. BMC Pregnancy and Childbirth 11(1).

214
References

Dodd JM, Crowther CA, Antoniou G, Baghurst P, Robinson JS (2007) Screening for gestational diabetes: The
effect of varying blood glucose definitions in the prediction of adverse maternal and infant health
outcomes. Aust N Z J Obstet Gynaecol 47(4):307-312.
Dodd JM, Grivell RM, Crowther CA, Robinson JS (2010) Antenatal interventions for overweight or obese
pregnant women: a systematic review of randomised trials. BJOG 117(11):1316-1326.
Dodd JM, Robinson JS (2011) Gestational weight loss in overweight and obese women is associated with an
increased risk of small for gestational age infants. Evid Based Med 16(4):125-126.
Dode MA, dos Santos IS (2009) Non classical risk factors for gestational diabetes mellitus: a systematic
review of the literature. Cadernos de Saúde Pública 25(Suppl 3):S341-S359.
Doherty DA, Magann EF, Francis J, Morrison JC, Newnham JP (2006) Pre-pregnancy body mass index and
pregnancy outcomes. Int J Gynecol Obstet 95(3):242-247.
Dollman J, Olds T (2006) Secular changes in fatness and fat distribution in Australian children matched for
body size. Int J Pediatr Obesity 1(2):109-113.
Donath S, Amir LH (2000) Rates of breastfeeding in Australia by State and socio-economic status: Evidence
from the 1995 National Health Survey. J Paediatr Child Health 36(2):164-168.
Donders ART, van der Heijden GJMG, Stijnen T, Moons KGM (2006) Review: A gentle introduction to
imputation of missing values. J Clin Epidemiol 59(10):1087-1091.
DuBose KD, Stewart EE, Charbonneau SR, Mayo MS, Donnelly JE (2006) Prevalence of the metabolic
syndrome in elementary school children. Acta Pædiatr 95(8):1005-1011.
Dudhbhai M, Lim L, Bombard A, Juliard K, Meenakshi B, Trachelenberg Y, Weiner Z (2006) Characteristics of
patients with abnormal glucose challenge test and normal oral glucose tolerance test results:
Comparison with normal and gestational diabetic patients. Am J Obstet Gynecol 194(5):e42-e45.
Dulloo AG, Jacquet J, Seydoux J, Montani JP (2006) The thrifty 'catch-up fat' phenotype: its impact on insulin
sensitivity during growth trajectories to obesity and metabolic syndrome. Int J Obes 30:S23-S35.
Duncan GE, Li SM, Zhou X-H (2004) Prevalence and trends of a metabolic syndrome phenotype among U.S.
adolescents, 1999-2000. Diabetes Care 27(10):2438-2443.
Duranthon V, Watson AJ, Lonergan P (2008) Preimplantation embryo programming: Transcription,
epigenetics, and culture environment. Reproduction 135(2):141-150.
Dwyer T, Blizzard L, Venn A, Stankovich JM, Ponsonby AL, Morley R (2002) Syndrome X in 8-y-old Australian
children: stronger associations with current body fatness than with infant size or growth. Int J Obes
Relat Metab Disord 26(10):1301-1309.
Egeland GM, Meltzer SJ (2010) Following in mother's footsteps? Mother - daughter risks for insulin resistance
and cardiovascular disease 15 years after gestational diabetes. Diabetic Med 27(3):257-265.
Ehrenberg HM, Huston-Presley L, Catalano PM (2003) The influence of obesity and gestational diabetes
mellitus on accretion and the distribution of adipose tissue in pregnancy. Am J Obstet Gynecol
189(4):944-948.
Ehrenberg HM, Mercer BM, Catalano PM (2004) The influence of obesity and diabetes on the prevalence of
macrosomia. Am J Obstet Gynecol 191(3):964-968.
Eisenmann JC, Heelan KA, Welk GJ (2004) Assessing body composition among 3- to 8-year-old children:
anthropometry, BIA, and DXA. Obesity 12(10):1633-1640.
Elgar FJ, Roberts C, Tudor-Smith C, Moore L (2005) Validity of self-reported height and weight and predictors
of bias in adolescents. J Adolesc Health 37(5):371-375.
Entringer S, Wüst S, Kumsta R, Layes IM, Nelson EL, Hellhammer DH, Wadhwa PD (2008) Prenatal
psychosocial stress exposure is associated with insulin resistance in young adults. Am J Obstet
Gynecol 199(5):498.e491-497.

215
References

Expert Panel on Detection Evaluation and Treatment of High Blood Cholesterol in Adults (2001) Executive
summary of the Third Report of the National Cholesterol Education Program (Adult Treatment Panel
III). JAMA 285(19):2486-2497.
Farooqi IS (2006) Genetic aspects of severe childhood obesity. Pediatr Endocrinol Rev 3 (Suppl 4):S528-536.
Ferguson MA, Gutin B, Owens S, Litaker M, Tracy RP, Allison J (1998) Fat distribution and hemostatic
measures in obese children. Am J Clin Nutr 67(6):1136-1140.
Ferrara A (2007) Increasing prevalence of gestational diabetes mellitus. Diabetes Care 30(Suppl 2):S141-146.
Fewtrell MS, Doherty C, Cole TJ, Stafford M, Hales CN, Lucas A (2000) Effects of size at birth, gestational
age and early growth in preterm infants on glucose and insulin concentrations at 9–12 years.
Diabetologia 43(6):714-717.
Field AE, Laird N, Steinberg E, Fallon E, Semega-Janneh M, Yanovski JA (2003) Which metric of relative
weight best captures body fatness in children? Obesity 11(11):1345-1352.
Finer LB, Henshaw SK (2006) Disparities in rates of unintended pregnancy in the United States, 1994 and
2001. Perspect Sex Reprod Health 38(2):90-96.
Finkelstein EA, Trogdon JG (2008) Public health interventions for addressing childhood overweight: analysis
of the business case. Am J Public Health 98(3):411-415.
Finucane MM, Stevens GA, Cowan MJ, Danaei G, Lin JK, Paciorek CJ, Singh GM, Gutierrez HR, Lu Y,
Bahalim AN, Farzadfar F, Riley LM, Ezzati M (2011) National, regional, and global trends in body-
mass index since 1980: Systematic analysis of health examination surveys and epidemiological
studies with 960 country-years and 9.1 million participants. The Lancet 377(9765):557-567.
Fisch B, Harel L, Kaplan B, Pinkas H, Amit S, Ovadia J, Tadir Y, Merlob P (1997) Neonatal assessment of
babies conceived by in vitro feritilization. J Perinatol 17(6):473-476.
Fisk CM, Crozier SR, Inskip HM, Godfrey KM, Cooper C, Robinson SnM (2011) Influences on the quality of
young children's diets: The importance of maternal food choices. Br J Nutr 105(2):287-296.
Fogelholm M, Nuutinen O, Pasanen M, Myöhänen E, Säätelä T (1999) Parent-child relationship of physical
activity patterns and obesity. Int J Obes Relat Metab Disord 23(12):1262-1268.
Forbes GB (1978) Body composition in adolescence. In: Human growth (Falkner F, Tanner JM, eds), London:
Bailliere Tindall, 239-272.
Ford ES, Ajani UA, Mokdad AH (2005) The metabolic syndrome and concentrations of C-reactive protein
among U.S. youth. Diabetes Care 28(4):878-881.
Fox-Wasylyshyn SM, El-Masri MM (2005) Handling missing data in self-report measures. Res Nurs Health
28(6):488-495.
Foxcroft K, Rowlands I, Byrne N, McIntyre H, Callaway L, the Bg (2011) Exercise in obese pregnant women:
The role of social factors, lifestyle and pregnancy symptoms. BMC Pregnancy Childbirth 11(1):4-10.
Franks PW, Looker HC, Kobes S, Touger L, Tataranni PA, Hanson RL, Knowler WC (2006) Gestational
glucose tolerance and risk of type 2 diabetes in young Pima Indian offspring. Diabetes 55(2):460-465.
Franks S (2008) Polycystic ovary syndrome in adolescents. Int J Obes 32(7):1035-1041.
Fraser A, Tilling K, Macdonald-Wallis C, Sattar N, Brion M-J, Benfield L, Ness A, Deanfield J, Hingorani A,
Nelson SM, Davey Smith G, Lawlor DA (2010) Association of maternal weight gain in pregnancy with
offspring obesity and metabolic and vascular traits in childhood. Circulation 121(23):2557-2564.
Frayling TM, Hattersley AT (2001) The role of genetic susceptibility in the association of low birth weight with
type 2 diabetes. Br Med Bull 60(1):89-101.
Frayling TM, Timpson NJ, Weedon MN, Zeggini E, Freathy RM, Lindgren CM, Perry JRB, Elliott KS, Lango H,
Rayner NW, Shields B, Harries LW, Barrett JC, Ellard S, Groves CJ, Knight B, Patch A-M, Ness AR,
Ebrahim S, Lawlor DA, Ring SM, Ben-Shlomo Y, Jarvelin M-R, Sovio U, Bennett AJ, Melzer D,

216
References

Ferrucci L, Loos RJF, Barroso Is, Wareham NJ, Karpe F, Owen KR, Cardon LR, Walker M, Hitman
GA, Palmer CNA, Doney ASF, Morris AD, Davey Smith G, The Wellcome Trust Case Control
Consortium, Hattersley AT, McCarthy MI (2007) A common variant in the FTO gene is associated
with body mass index and predisposes to childhood and adult obesity. Science 316(5826):889-894.
Frayn KN (2000) Visceral fat and insulin resistance - causative or correlative? Br J Nutr 83(Suppl 1):S71-S77.
Freedman DS, Srinivasan SR, Burke GL, Shear CL, Smoak CG, Harsha DW, Webber LS, Berenson GS
(1987) Relation of body fat distribution to hyperinsulinemia in children and adolescents: The
Bogalusa Heart Study. Am J Clin Nutr 46(3):403-410.
Freedman DS, Khan LK, Serdula MK, Dietz WH, Srinivasan SR, Berenson GS (2005a) The relation of
childhood BMI to adult adiposity: The Bogalusa Heart Study. Pediatrics 115(1):22-27.
Freedman DS, Ogden CL, Berenson GS, Horlick M (2005b) Body mass index and body fatness in childhood.
Curr Opin Clin Nutr Metab Care 8(6):618-623.
Freedman DS, Kahn HS, Mei Z, Grummer-Strawn LM, Dietz WH, Srinivasan SR, Berenson GS (2007a)
Relation of body mass index and waist-to-height ratio to cardiovascular disease risk factors in
children and adolescents: The Bogalusa Heart Study. Am J Clin Nutr 86(1):33-40.
Freedman DS, Wang J, Ogden CL, Thornton JC, Mei Z, Pierson RN, Dietz WH, Horlick M (2007b) The
prediction of body fatness by BMI and skinfold thicknesses among children and adolescents. Ann
Hum Biol 34(2):183-194.
Freinkel N (1980) Banting Lecture 1980. Of pregnancy and progeny. Diabetes 29(12):1023-1035.
Friedman JE, Ishizuka T, Shao J, Huston L, Highman T, Catalano P (1999) Impaired glucose transport and
insulin receptor tyrosine phosphorylation in skeletal muscle from obese women with gestational
diabetes. Diabetes 48(9):1807-1814.
Gabbe GC (2003) Management of diabetes mellitus complicating pregnancy. Obstet Gynecol 102(5 Pt 1):857-
868.
Gabory A, Attig L, Junien C (2009) Sexual dimorphism in environmental epigenetic programming. Mol Cell
Endocrinol 304(1-2):8-18.
Galtier-Dereure F, Boegner C, Bringer J (2000) Obesity and pregnancy: complications and cost. Am J Clin
Nutr 71(5):1242S-1248S.
Garcia-Bailo B, Toguri C, Eny KM, El-Sohemy A (2008) Genetic variation in taste and its influence on food
selection. OMICS 13(1):69-80.
García-Estévez DA, Araújo-Vilar D, Fiestras-Janeiro G, Saavedra-González Ã, Cabezas-Cerrato J (2003)
Comparison of several insulin sensitivity indices derived from basal plasma insulin and glucose
levels with minimal model indices. Horm Metab Res 35(1):13-17.
Garcia M, Fernandez E, Schiaffino A, Borrell C, Marti M, Borras JM (2005) Attrition in a population-based
cohort eight years after baseline interview: The Cornella Health Interview Survey Follow-up
(CHIS.FU) Study. Ann Epidemiol 15(2):98-104.
Garnett SP, Baur LA, Cowell CT (2008) Waist-to-height ratio: A simple option for determining excess central
adiposity in young people. Int J Obes 32(6):1028-1030.
Gesink Law DC, Maclehose RF, Longnecker MP (2007) Obesity and time to pregnancy. Hum Reprod
22(2):414-420.
Gillman MW, Rifas-Shiman S, Berkey CS, Field AE, Colditz GA (2003) Maternal gestational diabetes, birth
weight, and adolescent obesity. Pediatrics 111(3):e221-226.
Gillman MW (2004) A life course approach to obesity. In: A life course approach to chronic disease
epidemiology (Kuh D, Ben-Shlomo Y, eds), 2nd ed., New York: Oxford University Press, 189-217.
Gillman MW (2005) Developmental origins of health and disease. N Engl J Med 353(17):1848-1850.

217
References

Gillman MW, Oakey H, Baghurst PA, Volkmer RE, Robinson JS, Crowther CA (2010) Effect of treatment of
gestational diabetes mellitus on obesity in the next generation. Diabetes Care 33(5):964-968.
Glasgow RE, Emmons KM (2007) How can we increase translation of research into practice? Types of
evidence needed. Ann Rev Public Health 28(1):413-433.
Gluckman PD, Hanson MA (2007) Developmental plasticity and human disease: Research directions. J Intern
Med 261(5):461-471.
Gluckman PD, Hanson MA, Beedle AS (2007) Early life events and their consequences for later disease: A
life history and evolutionary perspective. Am J Hum Biol 19(1):1-19.
Godfrey KM, Barker DJP, Robinson S, Osmond C (1997) Maternal birthweight and diet in pregnancy in
relation to the infant's thinness at birth. Br J Obstet Gynaecol 104(6):663-667.
Goldfield GS, Cloutier P, Mallory R, Prud'homme D, Parker T, Doucet E (2006) Validity of foot-to-foot
bioelectrical impedance analysis in overweight and obese children and parents. J Sports Med Phys
Fitness 46(3):447-453.
Goran MI (1998) Measurement issues related to studies of childhood obesity: Assessment of body
composition, body fat distribution, physical activity, and food intake. Pediatrics 101(3):505-518.
Goran MI, Gower BA (1999) Relation between visceral fat and disease risk in children and adolescents. Am J
Clin Nutr 70(1):149-156.
Goran MI, Ball GDC, Cruz ML (2003) Obesity and risk of type 2 diabetes and cardiovascular disease in
children and adolescents. J Clin Endocrinol Metab 88(4):1417-1427.
Gorber SC, Tremblay M, Moher D, Gorber B (2007) A comparison of direct vs. self-report measures for
assessing height, weight and body mass index: A systematic review. Obes Rev 8(4):307-326.
Gortmaker SL, Swinburn BA, Levy D, Carter R, Mabry PL, Finegood DT, Huang T, Marsh T, Moodie ML
(2011) Changing the future of obesity: science, policy, and action. The Lancet 378(9793):838-847.
Government of South Australia (2007) Diabetes mellitus and abnormal glucose tolerance. In: Perinatal
Practice Guidelines, Adelaide, SA: Department of Health.
Gower BA, Nagy TR, Goran MI (1999) Visceral fat, insulin sensitivity, and lipids in prepubertal children.
Diabetes 48(8):1515-1521.
Greenland S (1983) Tests for interaction in epidemiologic studies: A review and a study of power. Stat Med
2(2):243-251.
Griffiths LJ, Parsons TJ, Hill AJ (2010) Self-esteem and quality of life in obese children and adolescents: A
systematic review. Int J Pediatr Obes 5(4):282-304.
Guelinckx I, Devlieger R, Beckers K, Vansant G (2008) Maternal obesity: pregnancy complications,
gestational weight gain and nutrition. Obes Rev 9(2):140-150.
Guillaume M (1999) Defining obesity in childhood: Current practice. Am J Clin Nutr 70(1):126S-130S.
Gungor N, Saad R, Janosky J, Arslanian S (2004) Validation of surrogate estimates of insulin sensitivity and
insulin secretion in children and adolescents. J Pediatr 144(1):47-55.
Gunnell DJ, Frankel SJ, Nanchahal K, Peters TJ, Davey Smith G (1998) Childhood obesity and adult
cardiovascular mortality: A 57-y follow-up study based on the Boyd Orr cohort. Am J Clin Nutr
67(6):1111-1118.
Gustafson SL, Rhodes RE (2006) Parental correlates of physical activity in children and early adolescents.
Sports Med 36(1):79-97.
Haas DM, Howard CS, Christopher M, Rowan K, Broga MC, Corey T (2006) Assessment of breastfeeding
practices and reasons for success in a military community hospital. J Hum Lact 22(4):439-445.

218
References

Hackmon R, James R, Green COR, Ferber A, Barnhard Y, Divon M (2007) The impact of maternal age, body
mass index and maternal weight gain on the glucose challenge test in pregnancy. J Matern Fetal
Neonatal Med 20(3):253-257.
Hales CN, Barker DJP (2001) The thrifty phenotype hypothesis: Type 2 diabetes. British Medical Bulletin
60(1):5-20.
Hall JE (1997) Mechanisms of abnormal renal sodium handling in obesity hypertension. Am J Hypertens
10(5):49S-55S.
Han JC, Lawlor DA, Kimm SYS (2010) Childhood obesity. The Lancet 375(9727):1737-1748.
Hanna FWF, Peters JR (2002) Screening for gestational diabetes; past, present and future. Diabetic Med
19(5):351-358.
HAPO Study Cooperative Research Group (2002) The Hyperglycemia and Adverse Pregnancy Outcome
(HAPO) Study. Int J Gynecol Obstet 78(1):69-77.
HAPO Study Cooperative Research Group (2008) Hyperglycemia and Adverse Pregnancy Outcomes. N Engl
J Med 358(19):1991-2002.
Hara M, Saitou E, Iwata F, Okada T, Harada K (2002) Waist-to-height ratio is the best predictor of
cardiovascular disease risk factors in Japanese schoolchildren. J Atheroscler Thromb 9(3):127-132.
Harris HE, Ellison GT (1998) Practical approaches for estimating prepregnant body weight. J Nurse Midwifery
43(2):97-101.
Harris MM, Stevens J, Thomas N, Schreiner P, Folsom AR (2000) Associations of fat distribution and obesity
with hypertension in a bi-ethnic population: The ARIC Study. Obesity 8(7):516-524.
Hattersley AT, Tooke JE (1999) The fetal insulin hypothesis: An alternative explanation of the association of
low birthweight with diabetes and vascular disease. The Lancet 353(9166):1789-1792.
Haworth CMA, Carnell S, Meaburn EL, Davis OSP, Plomin R, Wardle J (2008) Increasing heritability of BMI
and stronger associations with the FTO gene over childhood. Obesity 16(12):2663-2668.
Hearnshaw C, Matyka K (2010) Managing childhood obesity: When lifestyle change is not enough. Diabetes
Obes Metab 12(11):947-957.
Hedderson MM, Williams MA, Holt VL, Weiss NS, Ferrara A (2008) Body mass index and weight gain prior to
pregnancy and risk of gestational diabetes mellitus. Am J Obstet Gynecol 198(4):409.e401-
409.e407.
Hennekens CH, Buring JE (1987) Epidemiology in medicine. 1st ed.: Sherry L. Mayrent.
Herring SJ, Oken E, Rifas-Shiman SL, Rich-Edwards JW, Stuebe AM, Kleinman KP, Gillman MW (2009)
Weight gain in pregnancy and risk of maternal hyperglycemia. Am J Obstet Gynecol 201(1):61.e61-
61.e67.
Heslehurst N, Rankin J, Wilkinson JR, Summerbell CD (2010) A nationally representative study of maternal
obesity in England, UK: trends in incidence and demographic inequalities in 619323 births, 1989-
2007. Int J Obes 34(3):420-428.
Higgins M, Kannel W, Garrison R, Pinsky J, Stokes Jr (1988) Hazards of obesity - The Framingham
experience. Acta Med Scand Suppl.(723):23-26.
Hill DJ, Hogg J (1989) Growth factors and the regulation of pre- and postnatal growth. Bailliere's Clin
Endocrinol Metab 3(3):579-625.
Hillier TA, Pedula KL, Schmidt MM, Mullen JA, Charles M-A, Pettitt DJ (2007) Childhood obesity and
metabolic imprinting: The ongoing effects of maternal hyperglycemia. Diabetes Care 30(9):2287-
2292.

219
References

Himmelmann A, Himmelmann K, Svensson A, Hansson L (1997) Glucose and insulin levels in young subjects
with different maternal histories of hypertension: The Hypertension in Pregnancy Offspring Study. J
Intern Med 241(1):19-22.
Hinney A, Vogel C, Hebebrand J (2010) From monogenic to polygenic obesity: recent advances. Eur Child
Adolesc Psychiatry 19(3):297-310.
Hoffman L, Nolan C, Wilson JD, Oats JJN, Simmons D (1998) Gestational diabetes mellitus - Management
guidelines. Med J Aust 169:93-97.
Hosking J, Metcalf BS, Jeffery AN, Voss LD, Wilkin TJ (2006) Validation of foot-to-foot bioelectrical
impedance analysis with dual-energy X-ray absorbtiometry in the assessment of body composition in
young children: The EarlyBird cohort. Br J Nutr 96(6):1163-1168.
Hsia Y, Neubert AC, Rani F, Viner RM, Hindmarsh PC, Wong ICK (2009) An increase in the prevalence of
type 1 and 2 diabetes in children and adolescents: Results from prescription data from a UK general
practice database. Br J Clin Pharmacol 67(2):242-249.
Hsieh SD, Muto T (2005) The superiority of waist-to-height ratio as an anthropometric index to evaluate
clustering of coronary risk factors among non-obese men and women. Prev Med 40(2):216-220.
Huang R-C, de Klerk N, Mori TA, Newnham JP, Stanley FJ, Landau LI, Oddy WH, Hands B, Beilin LJ (2011)
Differential relationships between anthropometry measures and cardiovascular risk factors in boys
and girls. Int J Pediatr Obes 6(2-2):e271-e282.
Hughes AR, Sherriff A, Lawlor DA, Ness AR, Reilly JJ (2011) Incidence of obesity during childhood and
adolescence in a large contemporary cohort. Prev Med 52(5):300-304.
Hunt KJ, Schuller KL (2007) The increasing prevalence of diabetes in pregnancy. Obstet Gynecol Clin North
Am 34(2):173-199.
Hytten F (1991) Weight gain in pregnancy In: Clinical physiology in obstetrics (Hytten F, Chamberlain G, eds),
2nd ed., Oxford: Blackwell Scientific Publications, 173-203.
Innes KE, Byers TE, Marshall JA, Baron A, Orleans M, Hamman RF (2002) Association of a woman's own
birth weight with subsequent risk for gestational diabetes. JAMA 287(19):2534-2541.
Institute of Medicine (1990) Weight gain. In: Nutrition during pregnancy, Washington, DC: National Academy
of Sciences.
Institute of Medicine (2010) Bridging the evidence gap in obesity prevention: A framework to inform decision
making, Washington, DC: National Academy Press.
International Association of Diabetes and Pregnancy Study Groups Consensus Panel, Metzger BE, Gabbe
SG, Persson B, Buchanan TA, Catalano PA, Damm P, Dyer AR, Leiva A, Hod M, Kitzmiler JL, Lowe
LP, McIntyre HD, Oats JJ, Omori Y, Schmidt MI (2010) International Association of Diabetes and
Pregnancy Study Groups recommendations on the diagnosis and classification of hyperglycemia in
pregnancy. Diabetes Care 33(3):676-682.
Ishak M, Petocz P (2003) Gestational diabetes among Aboriginal Australians: Prevalence, time trend, and
comparisons with non-Aboriginal Australians. Ethn Dis 13(1):55-60.
Jang HC, Min HK, Lee HK, Cho NH, Metzger BE (1998) Short stature in Korean women: A contribution to the
multifactorial predisposition to gestational diabetes mellitus. Diabetologia 41(7):778-783.
Jeffery AN, Metcalf BS, Hosking J, Murphy MJ, Voss LD, Wilkin TJ (2006) Little evidence for early
programming of weight and insulin resistance for contemporary children: EarlyBird Diabetes Study
Report 19. Pediatrics 118(3):1118-1123.
Jensen DM, Damm P, Sørensen B, Mølsted-Pedersen L, Westergaard JG, Ovesen P, Beck-Nielsen H (2003)
Pregnancy outcome and prepregnancy body mass index in 2459 glucose-tolerant Danish women.
Am J Obstet Gynecol 189(1):239-244.

220
References

Jewson E, Spittle M, Casey M (2008) A preliminary analysis of barriers, intentions, and attitudes towards
moderate physical activity in women who are overweight. J Sci Med Sport 11(6):558-561.
Joffe GM, Esterlitz JR, Levine RJ, Clemens JD, Ewell MG, Sibai BM, Catalano PM (1998) The relationship
between abnormal glucose tolerance and hypertensive disorders of pregnancy in healthy nulliparous
women. Am J Obstet Gynecol 179(4):1032-1037.
Jones JI, Clemmons DR (1995) Insulin-like growth factors and their binding proteins: Biological actions.
Endocr Rev 16(1):3-34.
Jovanovic L (2000) A tincture of time does not turn the tide. Diabetes Care 23(9):1219-1220.
Ju H, Rumbold A, Willson K, Crowther C (2008) Borderline gestational diabetes mellitus and pregnancy
outcomes. BMC Pregnancy Childbirth 8(1):31-37.
Kaechele V, Wabitsch M, Thiere D, Kessler AL, Haenle MM, Mayer H, Kratzer W (2006) Prevalence of
gallbladder stone disease in obese children and adolescents: Influence of the degree of obesity, sex,
and pubertal development. J Pediatr Gastroenterol Nutr 42(1):66-70.
Kahn HS, Imperatore G, Cheng YJ (2005) A population-based comparison of BMI percentiles and waist-to-
height ratio for identifying cardiovascular risk in youth. J Pediatr 146(4):482-488.
Katsuki A, Sumida Y, Gabazza EC, Murashima S, Furuta M, Araki-Sasaki R, Hori Y, Yano Y, Adachi Y (2001)
Homeostasis model assessment is a reliable indicator of insulin resistance during follow-up of
patients with type 2 diabetes. Diabetes Care 24(2):362-365.
Kaufman FR (2002) Type 2 diabetes mellitus in children and youth: A new epidemic. J Pediatr Endocrinol
Metab 15(Suppl 2):S737-744.
Keely EJ, Malcolm JC, Hadjiyannakis S, Gaboury I, Lough G, Lawson ML (2008) Prevalence of metabolic
markers of insulin resistance in offspring of gestational diabetes pregnancies. Pediatric Diabetes
9(1):53-59.
Keesey RE, Hirvonen MD (1997) Body weight set-points: determination and adjustment. J Nutr 127(9):1875-
1883.
Kennelly MM, Geary M, McCaffrey N, McLoughlin P, Staines A, McKenna P (2002) Exercise-related changes
in umbilical and uterine artery waveforms as assessed by Doppler ultrasound scans. Am J Obstet
Gynecol 187(3):661-666.
Kershaw EE, Flier JS (2004) Adipose tissue as an endocrine organ. J Clin Endocrinol Metab 89(6):2548-2556.
Keskin M, Kurtoglu S, Kendirci M, Atabek ME, Yazici C (2005) Homeostasis model assessment is more
reliable than the fasting glucose/insulin ratio and quantitative insulin sensitivity check index for
assessing insulin resistance among obese children and adolescents. Pediatrics 115(4):e500-503.
Kettaneh A, Heude B, Lommez A, Borys JM, Ducimetiere P, charles MA (2005) Reliability of bioimpedance
analysis compared with other adiposity measurements in children: The FLVS II Study. Diabetes
Metab 31(6):534-541.
Kieffer EC, Carman WJ, Gillespie BW, Nolan GH, Worley SE, Guzman JR (2001) Obesity and gestational
diabetes among African-American women and Latinas in Detroit: Implications for disparities in
women's health. J Am Med Womens Assoc 56(4):181-187.
Kieffer EC, Tabaei BP, Carman WJ, Nolan GH, Guzman JR, Herman WH (2006) The influence of maternal
weight and glucose tolerance on infant birthweight in Latino mother-infant pairs. Am J Public Health
96(12):2201-2208.
Kim C, Newton KM, Knopp RH (2002) Gestational diabetes and the incidence of type 2 diabetes: A systematic
review. Diabetes Care 25(10):1862-1868.
Kim C, Liu T, Valdez R, Beckles GL (2009) Does frank diabetes in first-degree relatives of a pregnant woman
affect the likelihood of her developing gestational diabetes mellitus or nongestational diabetes? Am J
Obstet Gynecol 201(6):576.e571-576.e576.

221
References

Kim S, Humphrey MD (1999) Decrease in incidence of gestational diabetes mellitus in Far North Queensland
between 1992 and 1996. Aust N Z J Obstet Gynaecol 39(1):40-43.
Kim SY, Dietz PM, England L, Morrow B, Callaghan WM (2007) Trends in pre-pregnancy obesity in nine
states, 1993-2003. Obesity 15(4):986-993.
King JC (2006) Maternal obesity, metabolism, and pregnancy outcomes. Annu Rev Nutr 26(1):271-291.
Kirwan JP, Hauguel-De Mouzon S, Lepercq J, Challier J-C, Huston-Presley L, Friedman JE, Kalhan SC,
Catalano PM (2002) TNF-α is a predictor of insulin resistance in human pregnancy. Diabetes
51(7):2207-2213.
Koebnick C, Getahun D, Smith N, Porter AH, Der-Sarkissian JK, Jacobsen SJ (2010) Extreme childhood
obesity is associated with increased risk for gastroesophageal reflux disease in a large population-
based study. Int J Pediatr Obes 6(2-2):257-263.
Koukkou E, Young P, Lowy C (1997) The effect of maternal glycemic control on fetal growth in diabetic
pregnancies. Am J Perinatol 14(9):547-552.
Kral JG, Biron S, Simard S, Hould F-S, Lebel S, Marceau S, Marceau P (2006) Large maternal weight loss
from obesity surgery prevents transmission of obesity to children who were followed for 2 to 18
years. Pediatrics 118(6):e1644-1649.
Kramer MS, McDonald SW (2006) Aerobic exercise for women during pregnancy. Cochrane Database Syst
Rev 19(3):CD000180.
Krishnaveni GV, Veena SR, Hill JC, Kehoe S, Karat SC, Fall CHD (2010) Intrauterine exposure to maternal
diabetes is associated with higher adiposity and insulin resistance and custering of cardiovascular
risk markers in Indian children. Diabetes Care 33(2):402-404.
Kristman V, Manno M, Côté P (2004) Loss to follow-up in cohort studies: How much is too much? European
Journal of Epidemiology 19(8):751-760.
Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, Flegal KM, Guo SS, Wei R, Mei Z, Curtin LR, Roche AF,
Johnson CL (2000) CDC growth charts: United States. Adv Data 2000 Jun 8 (314):1-27 8(314):1-27.
Kuh D, Ben-Shlomo Y (2004) A life course approach to chronic disease epidemiology 2nd ed. New York:
Oxford University Press, 474pp.
Kyle UG, Bosaeus I, De Lorenzo AD, Deurenberg P, Elia M, Gómez JM, Heitmann BL, Kent-Smith L, Melchior
J-C, Pirlich M, Scharfetter H, Schols AMWJ, Pichard C (2004a) Bioelectrical impedance analysis -
Part I: Review of principles and methods. Clin Nutr 23(5):1226-1243.
Kyle UG, Bosaeus I, De Lorenzo AD, Deurenberg P, Elia M, Manuel Gómez J, Lilienthal Heitmann B, Kent-
Smith L, Melchior J-C, Pirlich M, Scharfetter H, M.W.J Schols A, Pichard C (2004b) Bioelectrical
impedance analysis--Part II: Utilization in clinical practice. Clin Nutr 23(6):1430-1453.
Labayen I, Moreno LA, Blay MG, Blay VA, Mesana MI, González-Gross M, Bueno G, Sarría A, Bueno M
(2006) Early programming of body composition and fat distribution in adolescents. J Nutr 136(1):147-
152.
Landon MB, Osei K, Platt M, O'Dorisio T, Samuels P, Gabbe SG (1994) The differential effects of body fat
distribution on insulin and glucose metabolism during pregnancy. Am J Obstet Gynecol 171(4):875-
884.
Landon MB, Spong CY, Thom E, Carpenter MW, Ramin SM, Casey B, Wapner RJ, Varner MW, Rouse DJ,
Thorp JM, Jr., Sciscione A, Catalano P, Harper M, Saade G, Lain KY, Sorokin Y, Peaceman AM,
Tolosa JE, Anderson GB, the Eunice Kennedy Shriver National Institute of Child Health, Human
Development Maternal-Fetal Medicine Units Network (2009) A multicenter, randomized trial of
treatment for mild gestational diabetes. N Engl J Med 361(14):1339-1348.
Landsberg L (1999) Role of the sympathetic adrenal system in the pathogenesis of the insulin resistance
syndrome. Ann N Y Acad Sci 892(1):84-90.

222
References

Landsberg L (2008) Body fat distribution and cardiovascular risk: A tale of 2 sites. Arch Intern Med
168(15):1607-1608.
Langer O, Yogev Y, Most O, Xenakis EMJ (2005) Gestational diabetes: The consequences of not treating. Am
J Obstet Gynecol 192(4):989-997.
Lapolla A, Dalfra MG, Bonomo M, Castiglioni MT, Di Cianni G, Masin M, Mion E, Paleari R, Schievano C,
Songini M, Tocco G, Volpe L, Mosca A (2007) Can plasma glucose and HbA1c predict fetal growth in
mothers with different glucose tolerance levels? Diabetes Res Clin Pract 77(3):465-470.
Larsen TB, Sørensen HT, Gislum M, Johnsen SP (2007) Maternal smoking, obesity, and risk of venous
thromboembolism during pregnancy and the puerperium: A population-based nested case-control
study. Thromb Res 120(4):505-509.
Law CM, Gordon GS, Shiell AW, Barker DJP, Hales CN (1995) Thinness at birth and glucose tolerance in
seven-year-old children. Diabet Med 12(1):24-29.
Lawlor DA, Riddoch CJ, Page AS, Anderssen SA, Froberg K, Harro M, Stansbie D, Davey Smith G (2005)
The association of birthweight and contemporary size with insulin resistance among children from
Estonia and Denmark: findings from the European Youth Heart Study. Diabet Med 22(7):921-930.
Lawlor DA, Leon DA, Rasmussen F (2007a) Growth trajectory matters: Interpreting the associations among
birth weight, concurrent body size, and systolic blood pressure in a cohort study of 378,707 Swedish
men. Am J Epidemiol 165(12):1405-1412.
Lawlor DA, Smith GD, O'Callaghan M, Alati R, Mamun AA, Williams GM, Najman JM (2007b) Epidemiologic
evidence for the fetal overnutrition hypothesis: Findings from the Mater-University study of pregnancy
and its outcomes. Am J Epidemiol 165(4):418-424.
Lawlor DA, Fraser A, Lindsay R, Ness A, Dabelea D, Catalano P, Davey Smith G, Sattar N, Nelson S (2010)
Association of existing diabetes, gestational diabetes and glycosuria in pregnancy with macrosomia
and offspring body mass index, waist and fat mass in later childhood: Findings from a prospective
pregnancy cohort. Diabetologia 53(1):89-97.
Le J, Zhang D, Menees S, Chen J, Raghuveer G (2010) "Vascular age" is advanced in children with
atherosclerosis-promoting risk factors Circ Cardiovasc Imaging 3(1):8-14.
Lederman S, Paxton A, Heymsfield SB, Wang J, Thornton J, Pierson RNJ (1997) Body fat and water changes
during pregnancy in women with different body weight and weight gain. Obstet Gynecol 90(4, Part
1):483-488.
Lee J (2006) Insulin resistance in children and adolescents. Rev Endocr Metab Disord 7(3):141-147.
Lee JM, Okumura MJ, Davis MM, Herman WH, Gurney JG (2006) Prevalence and determinants of insulin
resistance among U.S. adolescents: A population-based study. Diabetes Care 29(11):2427-2432.
Lee SY, Gallagher D (2008) Assessment methods in human body composition. Curr Opin Clin Nutr Metab
Care 11(5):566-572.
Li C, Kaur H, Choi WS, Huang TT, Lee RE, Ahluwalia JS (2005) Additive interactions of maternal
prepregnancy BMI and breast-feeding on childhood overweight. Obes Res 13(2):362-371.
Li M, McDermott RA (2010) Using anthropometric indices to predict cardio-metabolic risk factors in Australian
indigenous populations. Diabetes Res Clin Pract 87(3):401-406.
Li S, Chen W, Srinivasan SR, Boerwinkle E, Berenson GS (2003) The peroxisome proliferator-activated
receptor-γ2 gene polymorphism (Pro12Ala) beneficially influences insulin resistance and its tracking
from childhood to adulthood. Diabetes 52(5):1265-1269.
Linné Y, Rössner S (2003) Interrelationships between weight development and weight retention in subsequent
pregnancies: The SPAWN study. Acta Obstet Gynecol Scand 82(4):318-325.
Linné Y, Dye L, Barkeling B, Rössner S (2004) Long-term weight development in women: A 15-year follow-up
of the effects of pregnancy. Obes Res 12(7):1166-1178.

223
References

Lobstein T, Baur L, Uauy R (2004) Obesity in children and young people: A crisis in public health. Obes Rev
5(s1):4-85.
Lobstein TJ, James WPT, Cole TJ (2003) Increasing levels of excess weight among children in England. Int J
Obes Relat Metab Disord 27(9):1136-1138.
Lombard C, Deeks A, Jolley D, Teede H (2009) Preventing weight gain: The baseline weight related
behaviors and delivery of a randomized controlled intervention in community based women. BMC
Public Health 9(1):2.
Loos RJF, Phillips DIW, Fagard R, Beunen G, Derom C, Mathieu C, Verhaeghe J, Vlietinck R (2002) The
influence of maternal BMI and age in twin pregnancies on insulin resistance in the offspring.
Diabetes Care 25(12):2191-2196.
Loveman E, Frampton GK, Shepherd J, Picot J, Cooper K, Bryant J, Welch K, Clegg A (2011) The clinical
effectiveness and cost-effectiveness of long-term weight management schemes for adults: A
systematic review. Health Technol Assess 15(2):1-182.
Lucas A (1998) Programming by early nutrition: an experimental approach. J Nutr 128(2):401S-406S.
Lucas A, Fewtrell MS, Cole TJ (1999) Fetal origins of adult disease - the hypothesis revisited. BMJ
319(7204):245-249.
Lustig RH (2003) Autonomic dysfunction of the β-cell and the pathogenesis of obesity. Rev Endocr Metab
Disord 4(1):23-32.
MacKay S (2011 - in press) Legislative solutions to unhealthy eating and obesity in Australia. Public Health
doi: 10.1016/j.puhe.2011.06.004.
MacKinnon DP, Krull JL, Lockwood CM (2000) Equivalence of the mediation, confounding and suppression
effect. Prev Sci 1(4):173-181.
Madeira IR, Carvalho CNM, Gazolla FM, Matos HJd, Borges MA, Bordallo MAN (2008) [Cut-off point for
homeostatic model assessment for insulin resistance (HOMA-IR) index established from receiver
operating characteristic (ROC) curve in the detection of metabolic syndrome in overweight pre-
pubertal children] [Article in Portuguese, abstract in English]. Arquivos Brasileiros de Endocrinologia
& Metabologia 52:1466-1473.
Maffeis C, Banzato C, Brambilla P, Cerutti F, Corciulo N, Cuccarolo G, Di Pietro M, Franzese A, Gennari M,
Balsamo A, Grugni G, Iughetti L, Del Giudice EM, Petri A, Trada M, Yiannakou P (2010) Insulin
resistance is a risk factor for high blood pressure regardless of body size and fat distribution in obese
children. Nutr Metab Cardiovasc Dis 20(4):266-273.
Magarey AM, Daniels LA, Boulton TJ (2001) Prevalence of overweight and obesity in Australian children and
adolescents: Reassessment of 1985 and 1995 data against new standard international definitions.
Med J Aust 174(11):561-564.
Malcolm JC, Lawson ML, Gaboury I, Lough G, Keely E (2006) Glucose tolerance of offspring of mother with
gestational diabetes mellitus in a low-risk population. Diabetic Med 23(5):565-570.
Mallory GBJ, Fiser DH, Jackson R (1989) Sleep-associated breathing disorders in morbidly obese children
and adolescents. J Pediatr 115(6):892-897.
Mamun AA, O'Callaghan M, Callaway L, Williams G, Najman J, Lawlor DA (2009) Associations of gestational
weight gain with offspring body mass index and blood pressure at 21 years of age: Evidence from a
birth cohort study. Circulation 119(13):1720-1727.
Manios Y, Moschonis G, Kourlaba G, Bouloubasi Z, Grammatikaki E, Spyridaki A, Hatzis C, Kafatos A,
Fragiadakis GA (2008) Prevalence and independent predictors of insulin resistance in children from
Crete, Greece: The Children Study. Diabetic Med 25(1):65-72.
Manley SE, Stratton IM, Clark PM, Luzio SD (2007) Comparison of 11 human insulin assays: Implications for
cinical investigation and research. Clin Chem 53(5):922-932.

224
References

Maple-Brown LJ, Sinha AK, Davis EA (2010) Type 2 diabetes in Indigenous Australian children and
adolescents. J Paediatr Child Health 46(9):487-490.
Marfell-Jones M, Olds T (2007) Kinanthropometry X: Proceedings of the 10th International Society for the
Advancement of Kinanthropometry Conference, held in conjunction with the 13th Commonwealth
International Sport Conference (Marfell-Jones M, Olds T, eds), Melbourne, Australia
Marie Stopes International (2006) What women want when faced with an unplanned pregnancy
http://www.mariestopes.org.au/research (accessed 23 August 2011)
Martin FIR, Ratnaike S, Wootton A, Condos P, Suter PEN (1995) The 75 g oral glucose tolerance in
pregnancy. Diabetes Res Clin Pract 27(2):147-151.
Matsuda M, DeFronzo RA (1999) Insulin sensitivity indices obtained from oral glucose tolerance testing:
comparison with the euglycemic insulin clamp. Diabetes Care 22(9):1462-1470.
Matsuzawa Y (2005) White adipose tissue and cardiovascular disease. Best Pract Res Clin Endocrinol Metab
19(4):637-647.
Matthaei S, Stumvoll M, Kellerer M, Haring H-U (2000) Pathophysiology and pharmacological treatment of
insulin resistance. Endocr Rev 21(6):585-618.
Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC (1985) Homeostasis model
assessment: insulin resistance and β-cell function from fasting plasma glucose and insulin
concentrations in man. Diabetologia 28(7):412-419.
Mattsson S, Thomas BJ (2006) Development of methods for body composition studies. Phys Med Biol
51(13):R203-R228.
Mazza D, Chapman A (2010) Improving the uptake of preconception care and periconceptional folate
supplementation: What do women think? BMC Public Health 10(1):786-791.
Mazza D, Shand LK, Warren N, Keleher H, Browning CJ, Bruce EJ (2011) General practice and preventive
health care: A view through the eyes of community members. Med J Aust 195(4):180-183.
McAuley KA, Mann JI, Chase JG, Lotz TF, Shaw GM (2007) Point: HOMA - satisfactory for the time being;
HOMA: The best bet for the simple determination of insulin sensitivity, until something better comes
along. Diabetes Care 30(9):2411-2413.
McCarthy HD, Jarett KV, Crawley HF (2001) The development of waist circumference percentiles in British
children aged 5.0-16.9 y. Eur J Clin Nutr 55(10):902-907.
McCarthy HD, Ellis SM, Cole TJ (2003) Central overweight and obesity in British youth aged 11-16 years:
Cross sectional surveys of waist circumference. BMJ 326(7390):624-627.
McCarthy HD, Ashwell M (2006) A study of central fatness using waist-to-height ratios in UK children and
adolescents over two decades supports the simple message - "Keep your waist circumference to
less than half your height". Int J Obes 30(6):988-992.
McCarthy HD, Cole TJ, Fry T, Jebb SA, Prentice AM (2006) Body fat reference curves for children. Int J Obes
30(4):598-602.
McDonald SD, Han Z, Mulla S, Beyene J (2010) Overweight and obesity in mothers and risk of preterm birth
and low birth weight infants: Systematic review and meta-analyses. BMJ 341:c3428.
McIntyre HD, Cheung WN, Oats JJN, Simmons D (2005) Gestational diabetes mellitus: From consensus to
action on screening and treatment. Med J Aust 183(6):288-289.
McLean M, Chipps D, Cheung NW (2006) Mother to child transmission of diabetes mellitus: Does gestational
diabetes program Type 2 diabetes in the next generation? Diabet Med 23(11):1213-1215.
McLeod ER, Campbell KJ, Hesketh KD (2011) Nutrition knowledge: A mediator between socioeconomic
position and diet quality in Australian first-time mothers. J Am Diet Assoc 111(5):696-704.

225
References

McMillen IC, Robinson JS (2005) Developmental origins of the metabolic syndrome: Prediction, plasticity, and
programming. Physiol Rev 85(2):571-633.
Mei Z, Grummer-Strawn LM, Pietrobelli A, Goulding A, Goran MI, Dietz WH (2002) Validity of body mass
index compared with other body-composition screening indexes for the assessment of body fatness
in children and adolescents. Am J Clin Nutr 75(6):978-985.
Metwally M, Ong KJ, Ledger WL, Li TC (2008) Does high body mass index increase the risk of miscarriage
after spontaneous and assisted conception? A meta-analysis of the evidence. Fertil Steril 90(3):714-
726.
Metzger BE, Coustan DR, the Organizing Committee (1998) Summary and recommendations of the Fourth
International Workshop-Conference on Gestational Diabetes Mellitus. Diabetes Care 21(Suppl
2):B161-167.
Metzger BE, Buchanan TA, Coustan DR, de Leiva A, Dunger DB, Hadden DR, Hod M, Kitzmiller JL, Kjos SL,
Oats JN, Pettitt DJ, Sacks DA, Zoupas C (2007) Summary and recommendations of the Fifth
International Workshop - Conference on Gestational Diabetes Mellitus. Diabetes Care 30(Suppl
2):S251-260.
Mietus-Snyder ML, Lustig RH (2008) Childhood obesity: A drift in the "limbic triangle". Annu Rev Med
59(1):147-162.
Mingrone G, Manco M, Valera Mora ME, Guidone C, Iaconelli A, Gniuli D, Leccesi L, Chiellini C, Ghirlanda G
(2008) Influence of maternal obesity on insulin sensitivity and secretion in offspring. Diabetes Care
31(9):1872-1876.
Mokha JS, Srinivasan SR, Dasmahapatra P, Fernandez C, Chen W, Xu J, Berenson GS (2010) Utility of
waist-to-height ratio in assessing the status of central obesity and related cardiometabolic risk profile
among normal weight and overweight/obese children: The Bogalusa Heart Study. BMC Pediatr
11(10):73-79.
Montan S (2007) Increased risk in the elderly parturient. Curr Opin Obstet Gynecol 19(2):110-112.
Montgomery KS, Bushee TD, Phillips JD, Kirkpatrick T, Catledge C, Braveboy K, O'Rourke C, Patel N,
Prophet M, Cooper A, Mosley L, Parker C, Douglas G (2010) Women's challenges with postpartum
weight loss. Matern Child Health J 15(8):1176-1184.
Monzillo LU, Hamdy O (2003) Evaluation of insulin sensitivity in clinical practice and in research settings. Nutr
Rev 61(12):397-412.
Moore DS, McCabe GP (2003) Introduction to the practice of statistics. 4th ed. New York: W. H. Freeman and
Company.
Moore VM, Davies MJ, Willson KJ, Worsley A, Robinson JS (2004) Dietary composition of pregnant women is
related to size of the baby at birth. J Nutr 134(7):1820-1826.
Moos M-K (2010) From concept to practice: Reflections on the preconception health agenda. J Womens
Health 19(3):561-567.
Moran JL, Solomon PJ, Peisach AR, Martin J (2007) New models for old questions: Generalized linear
models for cost prediction. J Eval Clin Pract 13(3):381-389.
Moreira P, Padez C, Mourao-Carvalhal I, Rosado V (2007) Maternal weigh gain during pregnancy and
overweight in Portuguese children. Int J Obes 31(4):608-614.
Morey WH (2003) Measuring insulin resistance: A task worth doing. But how? Pediatric Diabetes 4(3):115-
118.
Morrison JA, Glueck CJ, Horn PS, Schreiber GB, Wang P (2009) Homeostasis model assessment of insulin
resistance*body mass index interactions at ages 9 to 10 years predict metabolic syndrome risk factor
aggregate score at ages 18 to 19 years: A 10-year prospective study of black and white girls.
Metabolism 58(3):290-295.

226
References

Moses RG, Griffiths RD, McPherson S (1994) The incidence of gestational diabetes mellitus in the Illawarra
area of New South Wales. Aust N Z J Obstet Gynaecol 34(4):425-427.
Mühlhäusler B, Smith SR (2009) Early-life origins of metabolic dysfunction: Role of the adipocyte. Trends
Endocrinol & Metab 20(2):51-57.
Mühlhäusler BS, Adam CL, Findlay PA, Duffield JA, McMillen IC (2006) Increased maternal nutrition alters
development of the appetite-regulating network in the brain. FASEB J 20(8):1257-1259.
Mühlhäusler BS (2007) Programming of the appetite-regulating neural network: A link between maternal
overnutrition and the programming of obesity? J Neuroendocrinol 19(1):67-72.
Muniyappa R, Lee S, Chen H, Quon MJ (2008) Current approaches for assessing insulin sensitivity and
resistance in vivo: Advantages, limitations, and appropriate usage. Am J Physiol Endocrinol Metab
294(1):E15-26.
Nambiar S, Truby H, Abbott RA, Davies PSW (2009) Validating the waist-height ratio and developing centiles
for use amongst children and adolescents. Acta Paediatr 98(1):148-152.
Nanda K (2004) Non-alcoholic steatohepatitis in children. Pediatr Transplant 8(6):613-618.
National Center for Health Statistics (2007) Births: Final data for 2005: National Vital Statistics Reports
National Center for Health Statistics (2009) Births: Final data for 2006: National Vital Statistics Reports
National Center for Health Statistics (2010) Births: Final data for 2008: National Vital Statistics Reports
National Diabetes Data Group (1979) Classification and diagnosis of diabetes mellitus and other categories of
glucose intolerance. Diabetes 28(12):1039-1057.
National Health and Medical Research Council (2009) NHMRC levels of evidence and grades for
recommendations for developers of guidelines, Canberra: NHMRC.
National Institute for Health and Clinical Excellence (2008) Diabetes in pregnancy: Management of diabetes
and its complications from pre-conception to the postnatal period: CG63
Nead KG, Halterman JS, Kaczorowski JM, Auinger P, Weitzman M (2004) Overweight children and
adolescents: A risk group for iron deficiency. Pediatrics 114(1):104-108.
Nelson SM, Matthews P, Poston L (2010) Maternal metabolism and obesity: modifiable determinants of
pregnancy outcome. Hum Reprod Update 16(3):255-275.
Neovius M, Linné Y, Barkeling B, Rössner S (2004) Discrepancies between classification systems of
childhood obesity. Obes Rev 5(2):105-114.
Newsome CA, Shiell AW, Fall CHD, Phillips DIW, Shier R, Law CM (2003) Is birth weight related to later
glucose and insulin metabolism? - A systematic review. Diabetic Med 20(5):339-348.
Nicholson WK, Fleisher LA, Fox HE, Powe NR (2005) Screening for gestational diabetes mellitus: A decision
and cost-effectiveness analysis of four screening strategies. Diabetes Care 28(6):1482-1484.
Nkansah-Amankra S, Luchok K, Hussey J, Watkins K, Liu X (2010) Effects of maternal stress on low birth
weight and preterm birth outcomes across neighborhoods of South Carolina, 2000–2003. Matern
Child Health J 14(2):215-226.
Norman A-C, Drinkard B, McDuffie JR, Ghorbani S, Yanoff LB, Yanovski JA (2005) Influence of excess
adiposity on exercise fitness and performance in overweight children and adolescents. Pediatrics
115(6):e690-696.
Norton K, Olds T (1996) Anthropometrica: a textbook of body measurement for sports and health courses.
Sydney: University of NSW Press.
O'Sullivan JB, Mahan CM (1964) Criteria for the oral glucose tolerance test in pregnancy. Diabetes 13:278-
285.

227
References

Oddy WH, Li J, Landsborough L, Kendall GE, Henderson S, Downie J (2006) The association of maternal
overweight and obesity with breastfeeding duration. J Pediatr 149(2):185-191.
Odibo AO, Nelson D, Stamilio DM, Sehdev HM, Macones GA (2006) Advanced maternal age is an
independent risk factor for intrauterine growth restriction. Am J Perinatol 23(5):325-328.
Ogden CL, Flegal KM, Carroll MD, Johnson CL (2002) Prevalence and trends in overweight among US
children and adolescents, 1999-2000. JAMA 288(14):1728-1732.
Ogden CL, Carroll MD, Flegal KM (2008) High body mass index for age among US children and adolescents,
2003-2006. JAMA 299(20):2401-2405.
Ogonowski J, Miazgowski T (2010) Are short women at risk for gestational diabetes mellitus? Eur J
Endocrinol 162(3):491-497.
Oken E, Gillman MW (2003) Fetal origins of obesity. Obesity 11(4):496-506.
Oken E, Huh SY, Taveras EM, Rich-Edwards JW, Gillman MW (2005) Associations of maternal prenatal
smoking with child adiposity and blood pressure. Obes Res 13(11):2021-2028.
Oken E, Ning Y, Rifas-Shiman SL, Radesky JS, Rich-Edwards JW, Gillman MW (2006) Associations of
physical activity and inactivity before and during pregnancy with glucose tolerance. Obstet Gynecol
108(5):1200-1207.
Oken E, Taveras EM, Kleinman KP, Rich-Edwards JW, Gillman MW (2007) Gestational weight gain and child
adiposity at age 3 years. Am J Obstet Gynecol 196(4):322.e321-322.e328.
Oken E, Rifas-Shiman SL, Field AE, Frazier AL, Gillman MW (2008) Maternal gestational weight gain and
offspring weight in adolescence. Obstet Gynecol 112(5):999-1006.
Oken E, Kleinman KP, Belfort MB, Hammitt JK, Gillman MW (2009) Associations of gestational weight gain
with short- and longer-term maternal and child health outcomes. Am J Epidemiol 170(2):173-180.
Okereke NC, Huston-Presley L, Amini SB, Kalhan S, Catalano PM (2004) Longitudinal changes in energy
expenditure and body composition in obese women with normal and impaired glucose tolerance. Am
J Physiol Endocrinol Metab 287(3):E472-E479.
Okosun IS, Liao Y, Rotimi CN, Dever GEA, Cooper RS (2000) Impact of birth weight on ethnic variations in
subcutaneous and central adiposity in American children aged 5-11 years. A study from the Third
National Health and Nutrition Examination Survey. Int J Obes Relat Metab Disord 24(4):479-484.
Olds TS (2009) One million skinfolds: Secular trends in the fatness of young people 1951-2004. Eur J Clin
Nutr 63(8):934-946.
Olds TS, Tomkinson GR, Ferrar KE, Maher CA (2010) Trends in the prevalence of childhood overweight and
obesity in Australia between 1985 and 2008. Int J Obes 34(1):57-66.
Olson CM, Strawderman MS, Hinton PS, Pearson TA (2003) Gestational weight gain and postpartum
behaviors associated with weight change from early pregnancy to 1y postpartum. Int J Obes Relat
Metab Disord 27(1):117-127.
Ong KK, Dunger DB (2004) Birth weight, infant growth and insulin resistance. Eur J Endocrinol 151(Suppl
3):U131-139.
Ong KK, Petry CJ, Emmett PM, Sandhu MS, Kiess W, Hales CN, Ness AR, Dunger DB (2004) Insulin
sensitivity and secretion in normal children related to size at birth, postnatal growth, and plasma
insulin-like growth factor-I levels. Diabetologia 47(6):1064-1070.
Ong KKL, Ahmed ML, Sherriff A, Woods KA, Watts A, Golding J, The ASTa, Dunger DB (1999) Cord blood
leptin is associated with size at birth and predicts infancy weight gain in humans. J Clin Endocrinol
Metab 84(3):1145-1148.
Oude LH, Baur L, Jansen H, Shrewsbury VA, O'Malley C, Stolk RP, Summerbell CD (2009) Interventions for
treating obesity in children. Cochrane Database Syst Rev(1):CD001872.

228
References

Parsons TJ, Power C, Logan S, Summerbell CD (1999) Childhood predictors of adult obesity: A systematic
review. Int J Obes Relat Metab Disord 23 (Suppl 8):S1-107.
Pecoraro P, Guida B, Caroli M, Trio R, Falconi C, Principato S, Pietrobelli A (2003) Body mass index and
skinfold thickness versus bioimpedance analysis: fat mass prediction in children. Acta Diabetol
40(Suppl 1):s278-281.
Peña AS, Wiltshire E, MacKenzie K, Gent R, Piotto L, Hirte C, Couper J (2006) Vascular endothelial and
smooth muscle function relates to body mass index and glucose in obese and nonobese children. J
Clin Endocrinol Metab 91(11):4467-4471.
Peneau S, Salanave B, Maillard-Teyssier L, Rolland-Cachera MF, Vergnaud AC, Mejean C, Czernichow S,
Vol S, Tichet J, Castetbon K, Hercberg S (2009) Prevalence of overweight in 6- to 15-year-old
children in central/western France from 1996 to 2006: trends toward stabilization. Int J Obes
33(4):401-407.
Perneger TV (1998) What's wrong with Bonferroni adjustments. BMJ 316:1236-1238.
Perusse L, Tremblay A, Leblanc C, Bouchard C (1989) Genetic and environmental influences on level of
habitual physical activity and exercise participation. Am J Epidemiol 129(5):1012-1022.
Pettitt DJ, Baird HR, Aleck KA, Bennett PH, Knowler WC (1983) Excessive obesity in offspring of Pima Indian
women with diabetes during pregnancy. N Engl J Med 3(308):242-245.
Pettitt DJ, Bennett PH, Knowler WC, Baird HR, Aleck KA (1985) Gestational diabetes mellitus and impaired
glucose tolerance during pregnancy: Long-term effects on obesity and glucose tolerance in the
offspring. Diabetes Care 34:119-122.
Pettitt DJ, Knowler WC, Bennett PH, Aleck KA, Baird HR (1987) Obesity in offspring of diabetic Pima Indian
women despite normal birth weight. Diabetes Care 10(1):76-80.
Pettitt DJ, Bennett PH, Saad MF, Charles MA, Nelson RG, Knowler WC (1991) Abnormal glucose tolerance
during pregnancy in Pima Indian women. Diabetes Care 40(Suppl 2):S126-130.
Phelan S (2010) Pregnancy: a "teachable moment" for weight control and obesity prevention. Am J Obstet
Gynecol 202(2):135.e131-135.e138.
Phillips D, Barker D, Hales C, Hirst S, Osmond C (1994) Thinness at birth and insulin resistance in adult life.
Diabetologia 37(2):150-154.
Pi-Sunyer X (2009) The medical risks of obesity. Postgrad Med 121(6):21-33.
Pinhas-Hamiel O, Dolan LM, Daniels SR, Standiford D, Khoury PR, Zeitler P (1996) Increased incidence of
non-insulin-dependent diabetes mellitus among adolescents. J Pediatr 128(5):608-615.
Pirkola J, Pouta A, Bloigu A, Hartikainen A-L, Laitinen J, Jarvelin M-R, Vaarasmaki M (2010) Risks of
overweight and abdominal obesity at age 16 years associated with prenatal exposures to maternal
prepregnancy overweight and gestational diabetes mellitus. Diabetes Care 33(5):1115-1121.
Plagemann A, Heidrich I, Götz F, Rohde W, G. D (1992) Lifelong enhanced diabetes susceptibility and obesity
after temporary intrahypothalamic hyperinsulinism during brain organization. Exp Clin Endocrinol
99(2):91-95.
Plagemann A, Harder T, Melchior K, Rake A, Rohde W, G. D (1999) Elevation of hypothalamic neuropeptide
Y-neurons in adult offspring of diabetic mother rats. Neuroreport 10(15):3211-3216.
Platt MJ, Stanisstreet M, Casson IF, Howard CV, Walkinshaw S, Pennycook S, McKendrick O (2002) St
Vincent's declaration 10 years on: Outcomes of diabetic pregnancies. Diabetic Med 19(3):216-220.
Power C, Jefferis BJMH (2002) Fetal environment and subsequent obesity: A study of maternal smoking. Int J
Epidemiol 31(2):413-419.
Pryjmachuk S, Richards DA (2007) Look before you leap and don't put all your eggs in one basket. J Res
Nursing 12(1):43-54.

229
References

Quinn SM, Baur LA, Garnett SP, Cowell CT (2009) Treatment of clinical insulin resistance in children: A
systematic review. Obes Rev 11(10):722-730.
Radesky JS, Oken E, Rifas-Shiman SL, Kleinman KP, Rich-Edwards JW, Gillman MW (2008) Diet during
early pregnancy and development of gestational diabetes. Pediatr Perinat Epidemiol 22(1):47-59.
Radley D, Cooke CB, Fuller NJ, Oldroyd B, Truscott JG, Coward WA, Wright A, Gately PJ (2009) Validity of
foot-to-foot bio-electrical impedance analysis body composition estimates in overweight and obese
children. Int J Body Compos Res 7(1):15-20.
Ragland DR (1992) Dichotomizing continuous outcome variables: dependence of the magnitude of
association and statistical power on the cutpoint. Epidemiology 3(5):434-440.
Ramsay JE, Ferrell WR, Crawford L, Wallace AM, Greer IA, Sattar N (2002) Maternal obesity is associated
with dysregulation of metabolic, vascular, and inflammatory pathways. J Clin Endocrinol Metab
87(9):4231-4237.
Rasmussen KM, Yaktine AL (2009) Weight gain during pregnancy: Reexamining the guidelines, Washington,
DC: Institute of Medicine.
Ravelli ACJ, van der Meulen JHP, Michels RPJ, Osmond C, Barker DJP, Hales CN, Bleker OP (1998)
Glucose tolerance in adults after prenatal exposure to famine. The Lancet 351(9097):173-177.
Ravelli GP, Stein ZA, Susser MW (1976) Obesity in young men after famine exposure in utero and early
infancy. N Engl J Med 295(7):349-353.
Reaven GM (1988) Banting lecture 1988. Role of insulin resistance in human disease. Diabetes 37(12):1595-
1607.
Reilly JJ, Wilson J, McColl JH, Carmichael M, Durnin JV (1996) Ability of biolectric impedance to predict fat-
free mass in prepubertal children. Pediatr Res 39(1):176-179.
Reilly JJ, Savage SAH, Ruxton CHS, Kirk TR (1999) Assessment of obesity in a community sample of
prepubertal children. Int J Obes Relat Metab Disord 23(2):217-219.
Reilly JJ, Armstrong J, Dorosty AR, Emmett PM, Ness A, Rogers I, Steer C, Sherriff A, Avon Longitudinal
Study of Parents and Children Study Team (2005) Early life risk factors for obesity in childhood:
cohort study. BMJ 330(7504):1357-1363.
Retnakaran R, Hanley AJG, Raif N, Connelly PW, Sermer M, Zinman B (2003) C-reactive protein and
gestational diabetes: The central role of maternal obesity. J Clin Endocrinol Metab 88(8):3507-3512.
Retnakaran R, Hanley AJG, Raif N, Connelly PW, Sermer M, Zinman B (2004) Reduced adiponectin
concentration in women with gestational diabetes. Diabetes Care 27(3):799-800.
Retnakaran R, Qi Y, Sermer M, Connelly PW, Zinman B, Hanley AJG (2008) Isolated hyperglycemia at 1 hour
on oral glucose tolerance test in pregnancy resembles gestational diabetes mellitus in predicting
postpartum metabolic dysfunction. Diabetes Care 31(7):1275-1281.
Retnakaran R, Shah BR (2009) Mild glucose intolerance in pregnancy and risk of cardiovascular disease: a
population-based cohort study. CMAJ: Canadian Medical Association Journal 181(6/7):371-376.
Rhodes ET, Pawlak DB, Takoudes TC, Ebbeling CB, Feldman HA, Lovesky MM, Cooke EA, Leidig MM,
Ludwig DS (2010) Effects of a low-glycemic load diet in overweight and obese pregnant women: A
pilot randomized controlled trial. Am J Clin Nutr 92(6):1306-1315.
Ricart W, López J, Mozas J, Pericot A, Sancho M, González N, Balsells M, Luna R, Cortázar A, Navarro P,
Ramírez O, Flández B, Pallardo L, Hernández-Mijas A, Ampudia J, Fernández-Real J, Corcoy R
(2005) Body mass index has a greater impact on pregnancy outcomes than gestational
hyperglycaemia. Diabetologia 48(9):1736-1742.
Rich-Edwards JW, James-Todd T, Mohllajee A, Kleinman K, Burke A, Gillman MW, Wright RJ (2010) Lifetime
maternal experiences of abuse and risk of pre-natal depression in two demographically distinct
populations in Boston. Int J Epidemiol 40(2):375-384.

230
References

Ridley K, Olds TS, Alison H (2006) The Multimedia Activity Recall for Children and Adolescents (MARCA):
development and evaluation. Int J Behav Nutr Pysical Act 3(10).
Rizzo TA, Metzger BE, Dooley SL, Cho NH (1997) Early malnutrition and child neurobehavioral development:
Insights from the study of children of diabetic mothers. Child Dev 68(1):26-38.
Robbins DC, Andersen L, Bowsher R, Chance R, Dinesen B, Frank B, Gingerich R, Goldstein D, Widemeyer
H-M, Haffner S, Hales CN, Jarett L, Polonsky K, Porte D, Skyler J, Webb G, Gallagher K (1996)
Report of the American Diabetes Association's Task Force on standardization of the insulin assay.
Diabetes 45(2):242-256.
Robinson S, Marriott L, Poole J, Crozier S, Borland S, Lawrence W, Law C, Godfrey K, Cooper C, Inskip H
(2007) Dietary patterns in infancy: the importance of maternal and family influences on feeding
practice. Br J Nutr 98(05):1029-1037.
Robson S, Chan A, Keane RJ, Luke CG (2001) Subsequent birth outcomes after an unexplained stillbirth:
Preliminary population-based retrospective cohort study. Aust N Z J Obstet Gynaecol 41(1):29-35.
Rodriguez G, Moreno LA, Blay MG, Blay VA, Fleta J, Sarria A, Bueno M (2005) Body fat measurement in
adolescents: Comparison of skinfold thickness equations with dual-energy X-ray absorptiometry. Eur
J Clin Nutr 59(10):1158-1166.
Rodríguez G, Moreno LA, Sarría A, Fleta J, Bueno M (2000) Assessment of nutritional status and body
composition in children using physical anthropometry and bioelectrical impedance: influence of
diurnal variations. J Pediatr Gastroenterol Nutr 30(3):305-309.
Rodriguez MA, Winkleby MA, Ahn D, Sundquist J, Kraemer HC (2002) Identification of population subgroups
of children and adolescents with high asthma prevalence: Findings from the Third National Health
and Nutrition Examination Survey. Arch Pediatr Adolesc Med 156(3):269-275.
Rogers I (2003) The influence of birthweight and intrauterine environment on adiposity and fat distribution in
later life. Int J Obes Relat Metab Disord 27(7):755-777.
Rooney BL, Schauberger CW (2002) Excess pregnancy weight gain and long-term obesity: one decade later.
Obstet Gynecol 100(2):245-252.
Rosenfield RL, Lipton RB, Drum ML (2009) Thelarche, pubarche, and menarche attainment in children with
normal and elevated body mass index. Pediatrics 123(1):84-88.
Rothman KJ (1990) No adjustments are needed for multiple comparisons. Epidemiology 1(1):43-46.
Rowland ML (1990) Self-reported weight and height. Am J Clin Nutr 52(6):1125-1133.
Royal Australian College of General Practitioners (2009) RACGP Redbook: Guidelines for preventive
activities in general practice 7th Ed ed. Melbourne.
Rush EC, Puniani K, Valencia ME, Davies PSW, Plank LD (2003) Estimation of body fatness from body mass
index and bioelectrical impedance: Comparison of New Zealand European, Maori and Pacific Island
children. Eur J Clin Nutr 57(11):1394-1401.
Sacheck J (2008) Pediatric obesity: an inflammatory condition? J Parent Enteral Nutr 32(6):633-637.
Sacks G, Veerman JL, Moodie M, Swinburn B (2011) `Traffic-light' nutrition labelling and `junk-food' tax: A
modelled comparison of cost-effectiveness for obesity prevention. Int J Obes 35(7):1001-1009.
Saldana TM, Siega-Riz AM, Adair LS (2004) Effect of macronutrient intake on the development of glucose
intolerance during pregnancy. Am J Clin Nutr 79(3):479-486.
Saldana TM, Siega-Riz AM, Adair LS, Suchindran C (2006) The relationship between pregnancy weight gain
and glucose tolerance status among black and white women in central North Carolina. Am J Obstet
Gynecol 195(6):1629-1635.
Savage DB, Petersen KF, Shulman GI (2005) Mechanisms of insulin resistance in humans and possible links
with inflammation. Hypertension 45(5):828-833.

231
References

Savva SC, Tornaritis M, Savva ME, Kourides Y, Panagi A, Silikiotou N, Georgiou C, Kafatos A (2000) Waist
circumference and waist-to-height ratio are better predictors of cardiovascular disease risk factors in
children than body mass index. Int J Obes Relat Metab Disord 24(11):1453-1458.
Scaglioni S, Salvioni M, Galimberti C (2008) Influence of parental attitudes in the development of children
eating behaviour. Br J Nutr 99(Suppl 1):S22-25.
Schack-Nielsen L, Mortensen EL, Michaelsen KF, Sorensen TIA (2005) High maternal pregnancy weight gain
is associated with an increased risk of obesity in childhood and adulthood independent of maternal
BMI [abstract] In: 3rd International Congress on Developmental Origins of Health & Disease,
Toronto, Canada: Pediatr Res, p 1020.
Schack-Nielsen L, Michaelsen KF, Gamborg M, Mortensen EL, Sorensen TIA (2010) Gestational weight gain
in relation to offspring body mass index and obesity from infancy through adulthood. Int J Obes
34(1):67-74.
Schaefer-Graf UM, Graf K, Kulbacka I, Kjos SL, Dudenhausen J, Vetter K, Herrera E (2008) Maternal lipids as
strong determinants of fetal environment and growth in pregnancies with gestational diabetes
mellitus. Diabetes Care 31(9):1858-1863.
Schaefer F, Georgi M, Zieger A, Scharer K (1994) Usefulness of bioelectric impedance and skinfold
measurements in predicting fat-free mass derived from total body potassium in children. Pediatr Res
35(5):617-624.
Scholl TO, Chen X (2002) Insulin and the “thrifty” woman: The influence of insulin during pregnancy on
gestational weight gain and postpartum weight retention. Matern Child Health J 6(4):255-261.
Schulz KF, Grimes DA (2005) Multiplicity in randomised trials I: endpoints and treatments. Lancet 365:1591-
1595.
Schwimmer JB, Burwinkle TM, Varni JW (2003) Health-related quality of life of severely obese children and
adolescents. JAMA 289(14):1813-1819.
Scott JA, Binns CW, Oddy WH, Graham KI (2006) Predictors of breastfeeding duration: Evidence from a
cohort study. Pediatrics 117(4):e646-e655.
Sebire NJ, Jolly M, Harris JP, Wadsworth J, Joffe M, Beard RW, Regan L, Robinson S (2001) Maternal
obesity and pregnancy outcome: A study of 287,213 pregnancies in London. Int J Obes Relat Metab
Disord 25(8):1175-1182.
Segal KR, Burastero S, Chun A, Coronel P, Pierson RN, Wang J (1991) Estimation of extracellular and total
body water by multiple-frequency bioelectrical-impedance measurement. Am J Clin Nutr 54(1):26-29.
Seghieri G, Anichini R, De Bellis A, Alviggi L, Franconi F, Breschi MC (2002) Relationship between
gestational diabetes mellitus and low maternal birth weight. Diabetes Care 25(10):1761-1765.
Seghieri G, Bellis A, Anichini R, Alviggi L, Franconi F, Breschi MC (2005) Does parity increase insulin
resistance during pregnancy? Diabetic Med 22(11):1574-1580.
Sellers EAC, Singh GR, Sayers SM (2008) Large waist but low body mass index: The metabolic syndrome in
Australian Aboriginal children. J Pediatr 153(2):222-227.
Selvin S (1987) Two issues concerning the analysis of grouped data. Eur J Epidemiol 3(3):284-287.
Sermer M, Naylor CD, Gare DJ, Kenshole AB, Ritchie JWK, Farine D, Cohen HR, McArthur K, Holzapfel S,
Biringer A, Chen E (1995) Impact of increasing carbohydrate intolerance on maternal-fetal outcomes
in 3637 women without gestational diabetes: The Toronto Tri-hospital Gestational Diabetes Project.
Am J Obstet Gynecol 173(1):146-156.
Sewell MF, Huston-Presley L, Super DM, Catalano P (2006) Increased neonatal fat mass, not lean body
mass, is associated with maternal obesity. Am J Obstet Gynecol 195(4):1100-1103.
Shaat N, Groop L (2007) Genetics of gestational diabetes mellitus. Curr Med Chem 14(5):569-583.

232
References

Sharma AJ, Cogswell ME, Grummer-Strawn LM (2005) The association between pregnancy weight gain and
childhood overweight is modified by mother's pre-pregnancy BMI [abstract]. In: 3rd International
Congress on Developmental Origins of Health & Disease, Toronto, Canada: Pediatric Research, p 1038.
Shen W, Wang Z, Punyanita M, Lei J, Sinav A, Kral JG, Imielinska C, Ross R, Heymsfield SB (2003) Adipose
tissue quantification by imaging methods: A proposed classification. Obesity 11(1):5-16.
Shieh G (2009) Detecting interaction effects in moderated multiple regression with continuous variables:
Power and sample size considerations. Organiz Res Methods 12:510-528.
Shiell AW, Campbell DM, Hall MH, Barker DJP (2000) Diet in late pregnancy and glucose-insulin metabolism
of the offspring 40 years later. Br J Obstet Gynaecol 107(7):890-895.
Shinwell ES, Blickstein I (2007) The risks for very low birth weight infants from multiple pregnancies. Clin
Perinatol 34(4):587-597.
Sibai BM (2008) Intergenerational factors: A missing link for preeclampsia, fetal growth restriction, and
cardiovascular disease? Hypertension 51(4):993-994.
Silverman BL, Rizzo T, Green OC, Cho NH, Winter RJ, Ogata ES, Richards GE, Metzger BE (1991) Long-
term prospective evaluation of offspring of diabetic mothers Diabetes 40(Suppl 2):121-125.
Silverman BL, Metzger BE, Cho NH, Loeb CA (1995) Impaired glucose tolerance in adolescent offspring of
diabetic mothers. Relationship to fetal hyperinsulinism. Diabetes Care 18(5):611-617.
Silverman BL, Rizzo TA, Cho NH, Metzger BE (1998) Long-term effects of the intrauterine environment. The
Northwestern University Diabetes in Pregnancy Center. Diabetes Care 21(Suppl 2):B142-149.
Silverwood RJ, De Stavola BL, Cole TJ, Leon DA (2009) BMI peak in infancy as a predictor for later BMI in
the Uppsala Family Study. Int J Obes 33(8):929-937.
Simonen RL, Perusse L, Rankinen T, Rice T, Rao DC, Bouchard C (2002) Familial aggregation of physical
activity levels in the Québec Family Study. Med Sci Sports Exerc 34(7):1137-1142.
Sinaiko AR, Steinberger J, Moran A, Hong C-P, Prineas RJ, Jacobs DR (2006) Influence of insulin resistance
and body mass index at age 13 on systolic blood pressure, triglycerides, and high-density lipoprotein
cholesterol at age 19. Hypertension 48(4):730-736.
Singh AS, Mulder C, Twisk JWR, Van Mechelen W, Chinapaw MJM (2008) Tracking of childhood overweight
into adulthood: A systematic review of the literature. Obes Rev 9(5):474-488.
Singhal A, Fewtrell M, Cole TJ, Lucas A (2003) Low nutrient intake and early growth for later insulin
resistance in adolescents born preterm. The Lancet 361(9363):1089-1097.
Sinha R, Fisch G, Teague B, Tamborlane WV, Banyas B, Allen K, Savoye M, Rieger V, Taksali S, Barbetta G,
Sherwin RS, Caprio S (2002) Prevalence of impaired glucose tolerance among children and
adolescents with marked obesity. N Engl J Med 346(11):802-810.
Slaughter MH, Lohmen TG, Boileau RA, Horswill CA, Stillman RJ, Van Loan MD, Bemben DA (1988) Skinfold
equations for estimation of body fatness in children and youth. Hum Biol 60(5):709-723.
Solomon CG, Willett WC, Carey VJ, Rich-Edwards J, Hunter DJ, Colditz GA, Stampfer MJ, Speizer FE,
Spiegelman D, Manson JE (1997) A prospective study of pregravid determinants of gestational
diabetes mellitus. JAMA 278(13):1078-1083.
Soonthornpun S, Setasuban W, Thamprasit A, Chayanunnukul W, Rattarasarn C, Geater A (2003) Novel
insulin sensitivity index derived from oral glucose tolerance test. J Clin Endocrinol Metab 88(3):1019-
1023.
Sopher AB, Thornton JC, Wang J, Pierson RN, Jr., Heymsfield SB, Horlick M (2004) Measurement of
percentage of body fat in 411 children and adolescents: A comparison of dual-energy X-ray
absorptiometry with a four-compartment model. Pediatrics 113(5):1285-1290.
Sorof J, Daniels S (2002) Obesity hypertension in children: A problem of epidemic proportions. Hypertension
40(4):441-447.

233
References

Spencer EA, Appleby PN, Davey GK, Key TJ (2002) Validity of self-reported height and weight in 4808 EPIC-
Oxford participants. Public Health Nutr 5(4):561-565.
Srivastava T (2006) Nondiabetic consequences of obesity on kidney. Pediatr Nephrol 21(4):463-470.
Staten MA, Stern MP, Miller WG, Steffes MW, Campbell SE, for the Insulin Standardization Workgroup (2010)
Insulin assay standardization. Diabetes Care 33(1):205-206.
Stefan N, Kantartzis K, Machann J, Schick F, Thamer C, Rittig K, Balletshofer B, Machicao F, Fritsche A,
Haring H-U (2008) Identification and characterization of metabolically benign obesity in humans.
Arch Intern Med 168(15):1609-1616.
Stein TP, Scholl TO, Schluter MD, Schroeder CM (1998) Plasma leptin influences gestational weight gain and
postpartum weight retention. Am J Clin Nutr 68(6):1236-1240.
Steinberger J, Moran A, Hong C-P, Jacobs Jr DR, Sinaiko AR (2001) Adiposity in childhood predicts obesity
and insulin resistance in young adulthood. J Pediatr 138(4):469-473.
Stenhouse E, Wright DE, Hattersley AT, Millward BA (2006) Maternal glucose levels influence birthweight and
'catch-up' and 'catch-down' growth in a large contemporary cohort. Diabetic Med 23(11):1207-1212.
Stommel M, Schoenborn CA (2010) Accuracy and usefulness of BMI measures based on self-reported weight
and height: Findings from the NHANES & NHIS 2001-2006. BMC Public Health 19(9):421-430.
Stone CA, McLachlan KA, Halliday JL, Wein P, Tippett C (2002) Gestational diabetes in Victoria in 1996:
incidence, risk factors and outcomes. Med J Aust 177(9):486-491.
Stothard KJ, Tennant PWG, Bell R, Rankin J (2009) Maternal overweight and obesity and the risk of
congenital anomalies: A systematic review and meta-analysis. JAMA 301(6):636-650.
Stotland NE, Caughey AB (2010) What should patients be told about obesity-related risks? J Perinatol
30(7):439-440.
Strategies to Overcome and Prevent (STOP) Obesity Alliance (2010) Pounds and policy: Building a healthier
nation with effective communication and legislation.
Strategies to Overcome and Prevent (STOP) Obesity Alliance Task Force on Women (2010) Statement of
agreement.
Stuebe A, McElrath T, Thadhani R, Ecker J (2010) Second trimester insulin resistance, early pregnancy body
mass index and gestational weight gain. Matern Child Health J 14(2):254-260.
Stuebe AM, Forman MR, Michels KB (2009) Maternal-recalled gestational weight gain, pre-pregnancy body
mass index, and obesity in the daughter. Int J Obes 33(7):743-752.
Sugden MC, Holness MJ (2002) Gender-specific programming of insulin secretion and action. J Endocrinol
175(3):757-767.
Swinburn B, Gill T, Kumanyika S (2005) Obesity prevention: A proposed framework for translating evidence
into action. Obes Rev 6(1):23-33.
Swinburn BA, Nyomba BL, Saad MF, Zurlo F, Raz I, Knowler WC, Lillioja S, Bogardus C, Ravussin E (1991)
Insulin resistance associated with lower rates of weight gain in Pima Indians. J Clin Invest 88(1):168-
173.
Swinburn BA, Sacks G, Hall KD, McPherson K, Finegood DT, Moodie ML, Gortmaker SL (2011) The global
obesity pandemic: shaped by global drivers and local environments. The Lancet 378(9793):804-814.
Tailor AM, Peeters PHM, Norat T, Vineis P, Romaguera D (2010) An update on the prevalence of the
metabolic syndrome in children and adolescents. Int J Pediatr Obes 5(3):202-213.
Taylor GM, Owen P, Mires GJ (1998) Foetal growth velocities in twin pregnancies. Twin Res 1(1):9-14.
Taylor RW, Jones IE, Williams SM, Goulding A (2002) Body fat percentages measured by dual-energy X-ray
absorptiometry corresponding to recently recommended body mass index cutoffs for overweight and
obesity in children and adolescents aged 3–18 y. Am J Clin Nutr 76(6):1416-1421.

234
References

Thornton YS, Smarkola C, Kopacz SM, Ishoof SB (2009) Perinatal outcomes in nutritionally monitored obese
pregnant women: a randomized clinical trial. J Natl Med Assoc 101(6):569-577.
Tieu J, Crowther CA, Middleton P (2008) Dietary advice in pregnancy for preventing gestational diabetes
mellitus. Cochrane Database Syst Rev(2):Art. No: CD006674.
Tok EC, Ertunc D, Bilgin O, Erdal EM, Kaplanoglu M, Dilek S (2006) PPAR-[gamma]2 Pro12Ala
polymorphism is associated with weight gain in women with gestational diabetes mellitus. Eur J
Obstet Gynecol Reprod Biol 129(1):25-30.
Torloni MR, Betrán AP, Horta BL, Nakamura MU, Atallah AN, Moron AF, Valente O (2009) Prepregnancy BMI
and the risk of gestational diabetes: A systematic review of the literature with meta-analysis. Obes
Rev 10(2):194-203.
Toschke AM, Martin RM, von Kries R, Wells J, Davey Smith G, Ness AR (2007) Infant feeding method and
obesity: body mass index and dual-energy X-ray absorptiometry measurements at 9-10 y of age from
the Avon Longitudinal Study of Parents and Children (ALSPAC). Am J Clin Nutr 85(6):1578-1585.
Trasande L (2011) Quantifying the economic consequences of childhood obesity and potential benefits of
interventions. Expert Rev Pharmacoecon Outcomes Res 11(1):47-50.
Tu Y-K, West R, Ellison GTH, Gilthorpe MS (2005) Why evidence for the fetal origins of adult disease might
be a statistical artifact: The "reversal paradox" for the relation between birth weight and blood
pressure in later life. Am J Epidemiol 161(1):27-32.
Twisk J, de Vente W (2002) Attrition in longitudinal studies: How to deal with missing data. J Clin Epidemiol
55(4):329-337.
Tyrrell VJ, Richards G, Hofman P, Gillies GF, Robinson E, Cutfiels WS (2001) Foot-to-foot bioelectrical
impedance analysis: A valuable tool for the measurement of body composition in children. Int J Obes
Relat Metab Disord 25(2):273-278.
UCLA: Academic Technology Services, Statistical Consulting Group How do I interpret a regression model
when some variables are log transformed?
US Department of Health and Human Services (1990) The health benefits of smoking cessation: A report of
the Surgeon General, Atlanta: Centers for Disease Control and Prevention.
Vague J (1956) The degree of masculine differentiation of obesities: a factor determining predisposition to
diabetes, atherosclerosis, gout, and uric calculous disease Am J Clin Nutr 4(1):20-34.
Valerio G, Licenziati MR, Iannuzzi A, Franzese A, Siani P, Riccardi G, Rubba P (2006) Insulin resistance and
impaired glucose tolerance in obese children and adolescents from Southern Italy. Nutr Metab
Cardiovasc Dis 16(4):279-284.
Vambergue A, Nuttens MC, Verier-Mine O, Dognin C, Cappoen JP, Fontaine P (2000) Is mild gestational
hyperglycaemia associated with maternal and neonatal complications? The DIAGEST Study.
Diabetic Med 17(3):203-208.
Van Loan MD, Kopp LE, King JC, Wong WW, Mayclin PL (1995) Fluid changes during pregnancy: use of
bioimpedance spectroscopy. J Appl Physiol 78:1037-1042.
Van Assche FA, Holemans K, Aerts L (2001) Long-term consequences for offspring of diabetes during
pregnancy. Br Med Bull 60(1):173-182.
Van den Broeck J, Argeseanu Cunningham S, Eeckels R, Herbst K (2005) Data cleaning: detecting,
diagnosing, and editing data abnormalities. PLoS Med 2(10):966-970.
Van Dorsten B, Lindley EM (2008) Cognitive and behavioral approaches in the treatment of obesity.
Endocrinol Metab Clin North Am 37(4):905-922.
Veening MA, van Weissenbruch MM, Delemarre-van de Waal HA (2002) Glucose tolerance, insulin
sensitivity, and insulin secretion in children born small for gestational age. J Clin Endocrinol Metab
87(10):4657-4661.

235
References

Vidmar S, Carlin JB, Hesketh K, Cole TJ (2004) Standardizing anthropometric measures in children and
adolescents with new functions for egen. Stata J 4(1):50-55.
Villar J, Carroli G, Wojdyla D, Abalos E, Giordano D, Ba'aqeel H, Farnot U, Bergsjø P, Bakketeig L,
Lumbiganon P, Campodónico L, Al-Mazrou Y, Lindheimer M, Kramer M (2006) Preeclampsia,
gestational hypertension and intrauterine growth restriction, related or independent conditions? Am J
Obstet Gynecol 194(4):921-931.
Viswanathan M, Siega-Riz AM, Moos M-K, Deierlein A, Mumford S, Knaack J, Thieda P, Lux LJ, Lohr KN
(2008) Outcomes of maternal weight gain, Evidence Report/Technology Assessment. In: Prepared
by RTI International-University of North Carolina Evidence-based Practice Cener under Contract No.
290-02-0016, Rockville, MD: Agency for Healthcare Research and Quality.
Vohr BR, McGarvey ST, Coll CG (1995) Effects of maternal gestational diabetes and adiposity on neonatal
adiposity and blood pressure. Diabetes Care 18(4):467-475.
Vohr BR, McGarvey ST (1997) Growth patterns of large-for-gestational-age and appropriate-for-gestational-
age infants of gestational diabetic mothers and control mothers at age 1 year. Diabetes Care
20(7):1066-1072.
Vohr BR, McGarvey ST, Tucker R (1999) Effects of maternal gestational diabetes on offspring adiposity at 4-7
years of age. Diabetes Care 22(8):1284-1291.
Vos T, Carter R, Barendregt J, Mihalopoulos C, Veerman L, Magnus A, Cobiac L, Bertram M, Wallace A,
ACE-Prevention Team (2010) Assessing cost-effectiveness in prevention: University of Queensland
and Deakin University, Australia.
Wallace TM, Levy JC, Matthews DR (2004) Use and abuse of HOMA modeling. Diabetes Care 27(6):1487-
1495.
Wang Y, Lobstein T (2006) Worldwide trends in childhood overweight and obesity. Int J Pediatr Obes 1(1):11-
25.
Wanzhen G, Paterson J, Carter S, Percival T (2006) Risk factors for preterm and small-for-gestational-age
babies: A cohort from the Pacific Islands Families Study. J Paediatr Child Health 42(12):785-792.
Watson J, Collins C, Sibbritt D, Dibley M, Garg M (2006) Validation of a child and adolescent food-frequency
questionnaire in Australia. In: The 10th International Congress on Obesity (ICO). September 3-8,
2006, Sydney, Australia: Obes Rev, p 267.
Weedon MN, Clark VJ, Qian Y, Ben-Shlomo Y, Timpson N, Ebrahim S, Lawlor DA, Pembrey ME, Ring S,
Wilkin TJ, Voss LD, Jeffery AN, Metcalf B, Ferrucci L, Corsi AM, Murray A, Melzer D, Knight B,
Shields B, Smith GD, Hattersley AT, Di Rienzo A, Frayling TM (2006) A common haplotype of the
glucokinase gene alters fasting glucose and birth weight: Association in six studies and population-
genetics analyses. Am J Hum Genet 79(6):991-1001.
Weijers RNM, Bekedam DJ, Smulders YM (2002) Determinants of mild gestational hyperglycemia and
gestational diabetes mellitus in a large Dutch multiethnic cohort. Diabetes Care 25(1):72-77.
Weiss R, Dziura J, Burgert TS, Tamborlane WV, Taksali SE, Yeckel CW, Allen K, Lopes M, Savoye M,
Morrison J, Sherwin RS, Caprio S (2004) Obesity and the metabolic syndrome in children and
adolescents. N Engl J Med 350(23):2362-2374.
Weiss R, Taksali SE, Dufour S, Yeckel CW, Papademetris X, Cline G, Tamborlane WV, Dziura J, Shulman
GI, Caprio S (2005) The "obese insulin-sensitive" adolescent: importance of adiponectin and lipid
partitioning. J Clin Endocrinol Metab 90(6):3731-3737.
Weiss R, Kaufman FR (2008) Metabolic complications of childhood obesity: Identifying and mitigating the risk.
Diabetes Care 31 (Suppl 2):S310-316.
Wells JCK, Chomtho S, Fewtrell MS (2007) Programming of body composition by early growth and nutrition.
Proc Nutr Soc 66(3):423-434.

236
References

Wells JCK, Haroun D, Williams JE, Wilson C, Darch T, Viner RM, Eaton S, Fewtrell MS (2010) Evaluation of
DXA against the four-component model of body composition in obese children and adolescents aged
5-21 years. Int J Obes 34(4):649-655.
Wendland E, Pinto M, Duncan B, Belizan J, Schmidt M (2008) Cigarette smoking and risk of gestational
diabetes: A systematic review of observational studies. BMC Pregnancy Childbirth 8(1):53-60.
Wenig C (2010) The impact of BMI on direct costs in children and adolescents: empirical findings for the
German Healthcare System based on the KiGGS-study. Eur J Health Econ.
Whitaker RC, Wright JA, Pepe MS, Seidel KD, Dietz WH (1997) Predicting obesity in young adulthood from
childhood and parental obesity. N Engl J Med 337(13):869-873.
Whitaker RC, Pepe MS, Seidel KD, Wright JA, Knopp RH (1998) Gestational diabetes and the risk of offspring
obesity. Pediatrics 101(2):e9.
Whitaker RC (2004) Predicting preschooler obesity at birth: The role of maternal obesity in early pregnancy.
Pediatrics 114(1):e29-36.
Whitlock EP, O'Connor EA, Williams SB, Beil TL, Lutz KW (2010) Effectiveness of weight management
interventions in children: A targeted systematic review for the USPSTF. Pediatrics 125(2):e396-e418.
Widschut HIJ (2006) Prepregnancy antecedents of a high risk pregnancy. In: High risk pregnancy:
Management options (James DK, Steer PJ, Weiner CP, Gonik B, eds), 3rd ed., Philadelphia:
Elsevier Saunders, 3-42.
Wolff S, Legarth J, Vangsgaard K, Toubro S, Astrup A (2008) A randomized trial of the effects of dietary
counseling on gestational weight gain and glucose metabolism in obese pregnant women. Int J Obes
32(3):495-501.
World Health Organization (1985) Diabetes mellitus. Report of a WHO Expert Committee. In: WHO Technical
Report Series 727, Geneva: WHO.
World Health Organization (1995) Physical status: The use and interpretation of anthropometry. Report of a
WHO Expert Committee. In: WHO Technical Report Series 854, Geneva: WHO.
World Health Organization (2000) Obesity: Preventing and managing the global epidemic. Report of a WHO
Expert Committee. In: WHO Technical Report Series 894, Geneva: WHO.
World Health Organization (2001) Report of the expert consultation on the optimal duration of exclusive
breastfeeding, Geneva: WHO.
World Health Organization (2009a) Global database on body mass index - An interactive surveillance tool for
monitoring nutrition transition, Geneva: WHO.
World Health Organization (2009b) 2008-2013 Action plan for the global strategy for the prevention and
control of noncommunicable diseases, Geneva: WHO.
World Medical Association Declaration of Helsinki (1997) Recommendations guiding physicians in biomedical
research involving human subjects. JAMA 277(11):925-926.
Wright CM, Sherriff A, Ward SCG, McColl JH, Reilly JJ, Ness AR (2007) Development of bioelectrical
impedance-derived indices of fat and fat-free mass for assessment of nutritional status in childhood.
Eur J Clin Nutr 62(2):210-217.
Wright CS, Rifas-Shiman SL, Rich-Edwards JW, Taveras EM, Gillman MW, Oken E (2009) Intrauterine
exposure to gestational diabetes, child adiposity, and blood pressure. Am J Hypertens 22(2):215-220.
Wrotniak BH, Shults J, Butts S, Stettler N (2008) Gestational weight gain and risk of overweight in the
offspring at age 7 y in a multicenter, multiethnic cohort study. Am J Clin Nutr 87(6):1818-1824.
Yajnik C, Deshpande S, Jackson A, Refsum H, Rao S, Fisher D, Bhat D, Naik S, Coyaji K, Joglekar C, Joshi
N, Lubree H, Deshpande V, Rege S, Fall C (2008) Vitamin B12 and folate concentrations during
pregnancy and insulin resistance in the offspring: The Pune Maternal Nutrition Study. Diabetologia
51(1):29-38.

237
References

Yang X, Hsu-Hage B, Zhang H, Zhang C, Zhang Y, Zhang C (2002) Women with impaired glucose tolerance
during pregnancy have significantly poor pregnancy outcomes. Diabetes Care 25(9):1619-1624.
Yogev Y, Xenakis EMJ, Langer O (2004) The association between preeclampsia and the severity of
gestational diabetes: The impact of glycemic control. American Journal of Obstetrics and Gynecology
191(5):1655-1660.
Yogev Y, Catalano PM (2009) Pregnancy and obesity. Obstet Gynecol Clin North Am 36(2):285-300.
Yoshinaga M, Yuasa Y, Hatano H, Kono Y, Nomura Y, Oku S, Nakamura M, Kanekura S, Otsubo K, Akiba S,
Miyata K (1995) Effect of total adipose weight and systemic hypertension on left ventricular mass in
children. Am J Cardiol 76(11):785-787.
Yu SM, Nagey DA (1992) Validity of self-reported pregravid weight. Ann Epidemiol 2(5):715-721.
Yudkin JS, Eringa E, Stehouwer CDA (2005) "Vasocrine" signalling from perivascular fat: A mechanism
linking insulin resistance to vascular disease. The Lancet 365(9473):1817-1820.
Yue DK, Molyneaux LM, Ross GP, Constantino MI, Child AG, Turtle JR (1996) Why does ethnicity affect
prevalence of gestational diabetes? The underwater volcano theory. Diabetic Med 13(8):748-752.
Zhang C, Schulze M, Solomon C, Hu F (2006a) A prospective study of dietary patterns, meat intake and the
risk of gestational diabetes mellitus. Diabetologia 49(11):2604-2613.
Zhang C, Solomon CG, Manson JE, Hu FB (2006b) A prospective study of pregravid physical activity and
sedentary behaviors in relation to the risk for gestational diabetes mellitus. Arch Intern Med
166(5):543-548.
Zimmet P, Magliano D, Matsuzawa Y, Alberti G, Shaw J (2005) The metabolic syndrome: a global public
health problem and a new definition. J Atheroscler Thromb 12(6):295-300.
Zimmet P, Alberti G, Kaufman F, Tajima N, Silink M, Arslanian S, Wong G, Bennett P, Shaw J, Caprio S
(2007) The metabolic syndrome in children and adolescents. The Lancet 369(9579):2059-2061.

238

You might also like