Public Health Nutrition: 5(6A), 821–827 DOI: 10.

1079/PHN2002368

Keynote Address

Biomarkers in nutritional epidemiology

Sheila A Bingham*
Medical Research Council, Dunn Human Nutrition Unit, Welcome Trust/MRC Building, Hills Road,
Cambridge CB2 2XY, UK

Abstract
Objective: To illustrate biomarkers of diet that can be used to validate estimates of
dietary intake in the study of gene –environment interactions in complex diseases.
Design: Prospective cohort studies, studies of biomarkers where diet is carefully
controlled.
Setting: Free-living individuals, volunteers in metabolic suites.
Subjects: Male and female human volunteers.
Results: Recent studies using biomarkers have demonstrated substantial differences in
the extent of measurement error from those derived by comparison with other
methods of dietary assessment. The interaction between nutritional and genetic
factors has so far largely gone uninvestigated, but can be studied in epidemiological
trials that include collections of biological material. Large sample sizes are required to
study interactions, and these are made larger in the presence of measurement errors.
Conclusions: Diet is of key importance in affecting the risk of most chronic diseases in
man. Nutritional epidemiology provides the only direct approach to the
quantification of risks. The introduction of biomarkers to calibrate the measurement
error in dietary reports, and as additional measures of exposure, is a significant
development in the effort to improve estimates of the magnitude of the contribution
of diet in affecting individual disease risk within populations. The extent of
measurement error has important implications for correction for regression dilution Keywords
and for sample size. The collection of biological samples to improve and validate Diet
estimates of exposure, enhance the pursuit of scientific hypotheses, and enable Cancer
gene –nutrient interactions to be studied, should become the routine in nutritional Biomarkers
epidemiology. Epidemiology

Epidemiology has traditionally been regarded as a of very large numbers of free-living individuals, one
hypothesis-generating exercise, but modern methods of the most difficult and challenging problems in
aim to test hypotheses and establish relative risks. In the nutrition. Furthermore, the complexity of the subject
absence of satisfactory animal models of human chronic has increased markedly in recent years with the need
degenerative diseases, such as cancer, nutritional epi- to assess risks from dietary items, such as phyto-
demiology provides the only direct approach to the chemicals and contaminants, for which no databases of
assessment of risk from diet in man. food levels are published. The past 15 years have seen
There is a large body of evidence to show that the a marked increase in the recognition of the need to
marked international differences in the occurrence of improve, and quantify the errors involved in, dietary
most of these chronic diseases, such as cancer, are assessment. The success of the four International
mainly due to environmental factors, such as diet. Conferences on Dietary Assessment Methods, at St. Paul,
Many dietary factors are associated with disease Boston, Arnhem and Arizona, and earlier European
prevention or causation; these factors range from meetings, is a testament to the importance with which
traditional nutrients such as antioxidant vitamins, fat nutritional epidemiology is now viewed.
and plant polysaccharides, to foods such as vegetables, Despite strong international associations between diet
meat and fruit, to phytochemicals such as glucosino- and cancer, and high estimates of attributable risk from
lates, phyto-oestrogens and carotenoids, and to diet in Westernised societies, within-population estimates
contaminants such as heterocyclic amines and afla- of relative risks between dietary factors and chronic
toxins. However, causal associations require evidence disease such as cancer are rarely greater than 1.5 for most
of individual risk from exposure to particular items of foods or dietary items1,2. It is difficult to ascribe causality to
diet. This entails accurate measures of the habitual diet these low estimates, in comparison with relative risks of

*Corresponding author: Email sab@mrc-dunn.cam.ac.uk q The Author 2002

822 SA Bingham
the order of 15 for smoking and lung cancer, for example. Measurement error in nutritional epidemiology
Two main factors may account for these low within-
population estimates: genetic susceptibility and measure- There are a variety of methods in use for the measurement
ment error in dietary assessment. of diet in cohort, cross-sectional and intervention studies,
where the aim is to assess contemporaneous diet. Details
Gene –environment interactions in nutritional together with practical advice on equipment, timing and
epidemiology protocols are given elsewhere12 – 15. Methods generally
involve either the collation of observations from a number
Genetic factors are important in influencing individual of separate days’ investigations, as in records, checklists
susceptibility to environmental factors, including the and 24-hour recalls, or attempts to obtain average intake
effects of diet. Studies of common variants of genes alone by asking about the usual frequency of food consumption,
have generally proved disappointing in elucidating strong as in the diet history and food-frequency questionnaire
risks of sporadic cancer, for example in studies of (FFQ). In all methods of dietary assessment, some estimate
polymorphisms of genes controlling xenobiotic metab- of the weight of food consumed is required and, for the
olism3,4. However, there are examples where the inter- determination of nutrient or other food component intake,
actions between dietary factors and single nucleotide either an appropriate description for use with food tables
polymorphisms have markedly increased risk estimates. In or an aliquot for chemical analysis is necessary.
individuals with the fast alcohol-to-aldehyde metabolising Methods of measuring diet are associated with both
form of alcohol dehydrogenase (ADH) – ADH13 – random and systematic error. These errors arise from the
genotype, the odds ratio for oral cancer in moderate use of food tables, assessment of the frequency of food
alcohol drinkers is only 1.2. In heavy drinkers it rises to 40 in consumption, portion size, daily variation and failure to
those with the fast genotype ADH13 but to only 4.4 in those report usual diet, due to either changes in habits whilst
who are ADH23 . These associations provide strong causal taking part in an investigation or misreporting of food
evidence for the carcinogenicity of acetaldehyde5. There choice or amount. The presence of these errors has
are numerous examples emerging in the literature of the generated much controversy and discussion as to the most
interaction between disease risk, common polymorphisms ‘accurate’ method of dietary investigation. Numerous
and dietary factors, including N-acetyl transferase poly- validation studies have been conducted comparing results
morphisms in relation to meat consumption and cancer of one dietary assessment method with another, presumed
risk; apolipoprotein E in relation to saturated fat and more accurate, method on the same individuals. However,
coronary heart disease; methylene tetrahydrofolate unless food consumption can be observed independently,
reductase in relation to alcohol, folate, pyridoxine status the ‘true’ value for assessing validity is unknown because
and adenomatous polyps; and glutathione-S-transferase all traditional methods, even weighed records, rely on
status in relation to DNA adduct levels and diet6 – 9. food consumption as reported by the individuals.
In general, however, the interaction between nutrition, Furthermore, errors associated with the method under
biomarkers and genetic factors has so far only been investigation may be correlated with those of the reference
studied in comparatively small populations, so that the method, so that correction for regression dilution is
reported elevated risks associated with an interaction may substantially underestimated16.
have been based on very small numbers of subjects.
Findings might thus have arisen by chance. Furthermore, Biomarkers of nutrient intake
misclassification error in dietary assessment will increase
sample size requirements markedly; with only a 20% loss Beaton et al.17 stated that: ‘There will always be error in
in sensitivity in dietary assessment, the required sample dietary assessments. The challenge is to understand,
size for interactive investigations may more than double10. estimate, and make use of the error structure during
In addition, the preferred method of investigation is a analysis.’ The ability to do this, however, has only become
prospective approach, rather than a case –control study, to possible with the advent of biological markers in
avoid the introduction of bias in the assessment of the biological specimens such as blood, urine or hair, which
environmental exposure. reflect intake sufficiently closely to act as objective indices
Very large sample sizes are required for prospective of true intake. These biomarkers could also replace
studies of gene– environment interactions in cancer, and estimates of intake based on traditional methods.
they can be studied only in epidemiological cohorts that Biochemical markers of intake are potentially independ-
have measured environmental exposures, collections of ent of the errors associated with dietary survey methods.
biological material and a sufficient number of disease Their usefulness is illustrated in a follow-up study of
endpoints. The European Prospective Investigation of markers of aflatoxin exposure in relation to liver cancer.
Cancer (EPIC) is one such study, with a cohort size of The range of aflatoxin contamination of foods is very
400 000. Over 10 000 cases of cancer are expected within great, so that use of food tables of average levels of
only five years of follow-up11. contamination is unlikely to reflect individual exposure.
Biomarkers in nutritional epidemiology 823
24
Relative risks of cancer from aflatoxin consumption were population samples . In early reports, energy expendi-
only 0.9 and insignificant (confidence interval 0.4–1.9) for ture assessed from this method was unexpectedly low, 1.4
individuals classified to have had high dietary exposure, as times the basal metabolic rate (BMR) on average in a small
assessed by an interview of the frequency of consumption group of sedentary women. In women of normal weight,
of 45 foods. However, aflatoxin exposure biomarkers in the mean energy intake from weighed dietary records
urine samples obtained from individuals in the cohort agreed with mean energy expenditure data. But in obese
were able to detect substantial and significant relative risks women, mean energy intake assessed from 7-day weighed
for liver cancer of the order of 6–10. The estimated relative records was about 2 MJ (465 kcal) lower than mean
risk was 59.4 (16.6–212.0) in individuals positive for urine expenditure, suggesting that overweight women do not
biomarkers of both aflatoxin and hepatitis B18. report their habitual food intake25. In a later study, energy
Biomarkers may also be used to validate the accuracy of expenditure also exceeded energy intake measured from
dietary assessment methods, but prior calibration studies 7-day records in 31 normal individuals, on average by
under controlled conditions, for example in a metabolic 20%. As a ratio to BMR, energy intake was 1:46 ^ 0:31 MJ;
suite, are necessary to ascertain that the predictability of and energy expenditure was 1:82 ^ 0:24 MJ26. These and
the biomarker in humans consuming varying diets is at other studies have been summarised and show that, in
least as good as the dietary intake method that is being general, self-reported energy intake tends to be less than
validated. Few biomarkers of dietary intake have been energy expenditure as measured by the doubly labelled
studied in this way. water method27. Published results from dietary surveys
show that, using the ratio of energy intake to calculated
Fatty acids BMR (see below), 24-hour recalls tend to give low results,
Subcutaneous adipose tissue samples are reputedly easy many below limits compatible with normal energy
to obtain and their fatty acid composition can be related to expenditure with no loss in weight, whereas diet histories
estimates of fatty acid intake. Plakke et al.19 demonstrated give higher and records give intermediate values28.
quite close relations between polyunsaturated fat intake as The doubly labelled water method is too expensive for
assessed from adipose tissue and a 2-day food record in routine use by most investigators. BMR can be calculated
321 individuals ðr ¼ 0:5Þ; although the agreement was less from body weight, using equations29. The BMR then has to
good for monounsaturated fat and saturated fat (r ¼ 0:22 be multiplied by a factor to allow for total energy
and 0.24, respectively). More recently, the utility of expenditure, which has been derived from the difference
pentadecanoic acid (15:0) as a marker for dairy fat has between total energy expenditure and BMR in different
been investigated; this fatty acid is synthesised by the population samples, and ranges from 1.2 to 4.5. In
bacteria in the rumen and is not produced endogenously sedentary men and women, the factor to allow for physical
by mammalian cells. The correlation between total dairy activity is about 1.630. Individuals are said to ‘underreport’
product intake from 4-week food records and 15:0 in when the ratio of energy intake to BMR is less than this.
subcutaneous tissue was 0.63, and between total dairy However, the measurement of both energy intake and the
product intake from an FFQ and 15:0 in subcutaneous calculated estimate of BMR and total energy expenditure
tissue, 0.4020. There were also significant associations are imprecise, so that ‘cut-off’ limits for underreporting are
between the total amount of fat from milk products quite wide; for example, 0.90 if a single 24-hour recall is
assessed by a 7-day record and 15:0 in serum cholesterol used and 1.35 for ‘normal circumstances’31.
esters ðr ¼ 0:46Þ21. Biomarkers of fat intake are discussed
elsewhere in this supplement22. 24-Hour urine nitrogen
24-Hour urine nitrogen is the most well known biological
Doubly labelled water marker, with results from published metabolic studies,
The doubly labelled water method is an important where individual dietary intake is kept constant over
advance in the measurement of energy expenditure prolonged periods of time, showing a fair correlation
since it can be used on free-living individuals with between daily nitrogen intake and daily urine nitrogen
virtually no interference with everyday life, in contrast to excretion. Its use depends on the assumption that subjects
previous procedures. Subjects are given a carefully are in nitrogen balance, there being no accumulation due
weighed oral dose of 2H18 2 O and are then required only to growth or repair of lost muscle tissue, or loss due to
to donate timed urine samples over the next 15 days. starvation, slimming or injury. This was appreciated as
Carbon dioxide production is measured as the difference early as 1924, when it was suggested that actual protein
in the water pool (measured by 2H2) and the bicarbonate intake, as assessed from 24-hour urine excretion, was far
plus water pool (measured as 18O) and energy expendi- lower than the recommended level32. The apparent
ture estimated. The energy expenditure should be equal to accuracy of 24-hour urine nitrogen as a biological marker
energy intake, taking into account changes in body has led to the suggestion that it be used to validate
weight23. This method has been used to validate dietary estimates of protein intake from various dietary survey
assessment methods intended for surveillance of UK methods33. In 1980, Isaksson33 summarised a number of
824 SA Bingham
studies carried out by his group and showed that estimates ratio than the others, and their intakes of energy and all
of protein intake obtained from 24-hour recalls of food energy-yielding nutrients calculated from weighed records
intake were low when compared with urinary nitrogen, were significantly lower than those from individuals
but those estimated from diet histories and records were in who did not underreport. On average, there was an 18 g
good agreement with the urine values. Van Staveren et al.34 difference in reported fat consumption, and a 27 g
also found good agreement between 24-hour urine and difference in reported sugar consumption, between the
diet history estimates of protein intake. However, these values reported by the underreporters according to the
comparisons were investigated only on a group basis urine to dietary N ratio and the rest of the population.
because each individual contributed only a single or two Mean consumption of cakes, breakfast cereals, milk, eggs,
24-hour urine collections. Other early comparisons fats and sugars was also significantly lower in those
between average urine nitrogen and dietary intake have individuals classified as underreporters. However, there
been summarised12. was no difference in reported consumption of meat, fruits,
To investigate the applicability of using 24-hour urine vegetables and potatoes between these underreporters
nitrogen to validate estimates of protein intake on an and the other 80% of the population who gave valid
individual basis, four men and four women were given records, nor in vitamin C or carotene39. Hence, bias in
their usual varying diet over a 28-day period whilst living reports of food intake obtained from methods of dietary
in a metabolic suite35. Duplicates of diets were made up assessment can be assessed from 24-hour urine nitrogen.
each day for each individual, 24-hour urine and faecal Only some, not all, nutrients and foods may be
collections were also made over this period, and diets, underreported, and differences in means are reduced by
urine, faeces and skin losses measured for their nitrogen energy adjustment. Overweight individuals in particular
content. Urine nitrogen underestimated intake at higher are likely to underreport the amount they eat17. There are
levels of protein intake and overestimated at lower levels, also age effects; whereas there was good agreement
but a constant factor for faecal and skin losses can be used between intake from records and expenditure by the
to counteract this, and output from urine can be expressed doubly labelled water technique in 8-year-old girls, there
as a percentage of intake, 0.8135. Although this study was was a marked trend towards worsening agreement
based on results from a comparatively small group, a later with age, so that approximately 20% of energy was
meta-analysis of a large set of data has confirmed that underreported by 12- to 16-year-olds40.
urine nitrogen should be approximately 80% of dietary This problem is not confined to weighed dietary
intake on average36. However, less good correlations records, since it has been documented with all methods
between individual estimates of usual protein intake and of dietary assessment, including diet histories, FFQs and
the 24-hour urine nitrogen output will be obtained if fewer 24-hour recalls25,33,39,41 – 46.
observations on each individual are made, and if the
collections are not verified for their completeness with 24-Hour urine potassium
p-aminobenzoic acid (PABA)37. PABA is now in routine In healthy persons, urine is the major route of excretion of
use and has been used extensively in methodological potassium. Faecal excretion of potassium constitutes from
studies. Individuals who are judged to underreport by the 5 to 13 mmol per day in Western populations, or 11 –15%
24-hour urine nitrogen method also tend to be classified as of the dietary intake47 – 49. The correlation between intake
underreporters by the doubly labelled water method38. and excretion of potassium can be high, even when
To assess the validity of several different methods of dietary intakes are calculated from food tables rather than
dietary assessment, 160 women were asked to complete analysed, provided that sufficient 24-hour urine samples
16 days of weighed food records over one year, as four are obtained. Studies that have obtained at least eight
repeated 4-day records. The volunteers were also asked to 24-hour urine collections, validated for their complete-
provide eight 24-hour urine collections, as four repeated ness, have shown correlations of at least 0.7 between
2-day collections, and completeness of the urine collec- calculated intake and excretion50,51. Potassium has an
tions was assessed using the PABA check method. advantage as a biomarker of diet because a greater variety
Different methods of dietary assessment were completed of foods are good sources of potassium than those
during the year by the volunteers and it was shown that containing protein, for example vegetables and fruits.
correlations were greater between the biomarker 24-hour
urine nitrogen and estimates of nitrogen intake from Use of biomarkers for calibration and to assess
records, than from estimates of intake with other methods measurement error
including the FFQ39.
Using the ratio of urine to dietary nitrogen, it was also To meet the sample size requirements for studies in gene –
possible to distinguish between those individuals who nutrient interactions in chronic disease, pooling of data
provided valid records of food intake and the 20% who from multiple cohorts is becoming common practice. This
underreported their food intake. Individuals who under- also increases the heterogeneity of dietary habits, which is
reported were heavier, with a lower energy intake to BMR useful for overcoming measurement error in individual
Biomarkers in nutritional epidemiology 825
dietary assessments. However, since each participating between the dietary method and biomarker were
centre may have used different methods of dietary completely independent of each other. In both men
assessment, all with different measurement errors, and women, results calculated from the 7-day food diary
calibration is then necessary to correct for any bias in were much closer to estimates of output from urinary
mean intakes associated with these methods52. Within the biomarkers than those calculated from the FFQ. The
main EPIC study for example, a standardised compu- agreement between plasma vitamin C and vitamin C intake
terised 24-hour recall method, EPICSOFT, has been was of the same order no matter which method was
developed and administered to representative sub- used54.
samples within each cohort53. One hundred to 350 The design of this study also allowed error variance
participants within each sub-sample have also provided analysis to be conducted from the repeated dietary intake
24-hour urine samples, verified for completeness using measures and the repeated urine collections. Marked
PABA, in order to assess the validity of the EPICSOFT differences in error variances associated with the different
method. Initial results suggest a high correlation between dietary assessments were shown. The most accurate
mean nitrogen intake and nitrogen excretion levels across method, the 7-day food diary, had substantially less error
the populations studied, allowing confidence to be placed variance than the FFQ. Using the urine biomarkers as
in the validity of the calibration method (Slimani et al., indices of ‘true’ intake, the correction factors for
unpublished results). measurement error of relative risk estimates from the
In large epidemiological studies, it is now common dietary assessment methods could be estimated. Correc-
practice to correct for measurement error in the tion factors for regression dilution from the food diary
assessment of relative risk by regression calibration, and were only 1.8 to 2.0, whereas those for the FFQ were too
the correction factors are derived by comparison of the large to use with confidence (4.8 for potassium and 9.0 for
method in use, such as an FFQ, with a ‘reference’ method, nitrogen). Furthermore, the confidence limits around
such as a record. However, this practice relies on the these estimates for the FFQ became impossibly wide, 1.7
assumptions that errors in the reference instrument are to 16.2 for nitrogen for example55.
uncorrelated with both ‘true’ intake and errors in the EPIC within the UK thus will utilise dietary data from
method in use. These assumptions can be examined if different methods. The FFQ data are associated with a
repeat estimates of intake and repeat biomarker compari- greater degree of measurement error, as described above,
sons are available. Kipnis et al.36 have re-examined the so that correction factors for regression dilution are
work comparing repeat measures of intake of nitrogen substantially greater than those required for the food diary.
using different methods of dietary assessment and with Findings in the UK, where dietary variation between
repeat 24-hour urine excretion values. Using a new individuals is smaller and hence the need to use a more
measurement error model that allows for individual bias, accurate individual method greater, will be derived from
for example in the tendency to underreport food intake, the 7-day diary information on a nested case –control
they showed that neither of these assumptions was true, basis. All subjects within EPIC–Norfolk have dietary
leading to greatly underestimated attenuation factors and information from two 24-hour recalls that can be used in
consequently under-powered studies. The National the event that diary information should not have been
Cancer Institute is now initiating large studies of the forthcoming from some eventual cases. However, if
measurement errors of dietary assessments using doubly between-individual variation is increased, correction
labelled water and 24-hour urine nitrogen and potassium factors become smaller. Hence the FFQ is to be used
from urine samples, verified for completeness using PABA, particularly in pooled analyses of risk from diet in
as underlying validation measures, for example in the relation to cancer incidence within the larger European
Observing Protein and Energy Nutrition (OPEN) study EPIC study, where measurement error is more likely to
(Subar et al., unpublished results). be overcome by large dietary heterogeneity on an
In a recent study to assess the accuracy of methods in international basis.
the EPIC UK cohorts, repeat biomarker estimates were also
obtained from EPIC participants over a 9-month period. Summary
Urinary nitrogen, potassium and sodium were estimated
from two to six complete 24-hour urine collections in 134 Some biomarkers for the validation of methods for
subjects and plasma ascorbic acid from two or three fasting assessing dietary intake have been developed. There is a
blood samples in 118 subjects. PABA was used to verify the need for a greater variety of dietary biomarkers to be
completeness of the 24-hour urine collections. Subjects developed to reflect wider aspects of diet. At present, the
completed two FFQs and two 7-day food diaries, and the doubly labelled water technique and 24-hour urine
second diary and FFQ were sent at varying times over the nitrogen and potassium are in routine use for validation
course of the study. 24-Hour urine samples were not studies. Using these biomarkers, it has been shown that
collected during the time that subjects were recording their there could be substantial attenuation of diet effects and
dietary intake, making it more likely that any errors loss of statistical power in epidemiological studies.
826 SA Bingham
Attenuation and loss of power, together with genetic Epidemiology, Oxford Medical Publications, 2nd ed. Oxford:
Oxford University Press, 1997.
variation in response, could account for the inability of 16 Kipnis V, Carroll RJ, Freedman LS, Li L. Implications of a new
existing studies to show causal links between diet and dietary measurement error model for estimation of relative
chronic disease such as cancer. Some of this measurement risk: application to four calibration studies. Am. J. Epidemiol.
error can be overcome by studying populations whose 1999; 150: 642 –51.
17 Beaton GH, Burema J, Ritenbaugh C. Errors in the
dietary habits are more heterogeneous than those of single
interpretation of dietary assessments. Am. J. Clin. Nutr.
populations. But biomarker studies suggest that improved, 1997; 65: 1100S –7S.
more detailed, methods of dietary assessment will be 18 Qian GS, Ross RK, Yu MC, Yuan JM, Gao YT, Henderson BE,
necessary if causal associations between diet and disease et al. A follow-up study of urinary markers of aflatoxin
exposure and liver cancer risk in Shanghai, People’s
are to be established in future large-scale epidemiological
Republic of China. Cancer Epidemiol. Biomark. Prev. 1994;
studies. 3: 3 –10.
19 Plakke T, Berkel J, Beynen AC, Hermus RJJ, Katan MB.
References Relationship between the fatty acid composition of the diet
and that of the subcutaneous adipose tissue in individual
human subjects. Hum. Nutr. Appl. Nutr. 1983; 37: 365 – 72.
1 World Cancer Research Fund, Potter J, ed. Food, Nutrition 20 Wolk A, Vessby B, Ljung H, Barrefors P. Evaluation of a
and the Prevention of Cancer: A Global Perspective.
biological marker of dairy fat intake. Am. J. Clin. Nutr. 1998;
Washington, DC: American Institute for Cancer Research,
68: 291– 5.
1997.
21 Smedman AEM, Gustafsson IB, Berglund LGTM, Vessby
2 Committee on Medical Aspects of Food and Nutrition Policy.
BOH. Pentadecanoic acid in serum as a marker for intake of
Nutritional Aspects of the Development of Cancer. Report of
milk fat: relations between intake of milk fat and metabolic
the Working Group on Diet and Cancer of the Committee on
risk factors. Am. J. Clin. Nutr. 1999; 69: 22– 9.
Medical Aspects of Food and Nutrition Policy. Report on
22 Arab L, Akbar J. Biomarkers and the measurement of fatty
Health and Social Subjects No. 48. London: Stationery Office
acids. Public Health Nutr 2002; 5: 865 – 71.
[for the Department of Health], 1998.
23 Prentice AM, Coward WA, Davies HL, Murgatroyd PR, Black
3 D’Errico A, Taioli E, Chen X, Vineis P. Genetic metabolic
AE, Goldberg GR, et al. Unexpectedly low levels of energy
polymorphisms and the risk of cancer: a review of the
literature. Biomarkers 1996; 1: 149 – 73. expenditure in healthy women. Lancet 1985; 1: 1419– 22.
4 Cotton SC, Sharp L, Little J, Brockton N. Glutathione 24 Davies PSW, Coward WA, Gregory J, White A, Mills A. Total
S-transferase polymorphisms and colorectal cancer: a HuGE energy expenditure and energy intake in the pre-school
review. Am. J. Epidemiol. 2000; 151: 7 –32. child: a comparison. Br. J. Nutr. 1994; 72: 13– 20.
5 Harty LC, Caporaso NE, Hayes RB, Winn DM, Bravo-Otero E, 25 Prentice AM, Black AE, Coward WA, Davies HL, Goldberg
Blot WJ, et al. Alcohol dehydrogenase 3 genotype and risk of GR, Murgatroyd PR, et al. High levels of energy expenditure
oral cavity and pharyngeal cancers. J. Natl. Cancer Inst. in obese women. Br. Med. J. 1986; 292: 983– 7.
1997; 89: 1698 – 705. 26 Livingstone MBE, Prentice AM, Strain JJ, Coward WA, Black
6 Hein DW, Doll MA, Fretland AJ, Leff MA, Webb SJ, Xiao GH, AE, Barker ME, et al. Accuracy of weighed dietary records in
et al. Molecular genetics and epidemiology of the NAT1 and studies of diet and health. Br. Med. J. 1990; 300: 708–12.
NAT2 acetylation polymorphisms. Cancer Epidemiol. 27 Schoeller DA. How accurate is self-reported dietary energy
Biomark. Prev. 2000; 9: 29 –42. intake? Nutr. Rev. 1990; 48: 373 – 9.
7 Loktionov A, Scollen S, McKeown N, Bingham SA. Gene – 28 Black AE, Goldberg GR, Jebb SA, Livingstone MBE, Cole TJ,
nutrient interactions: dietary behaviour associated with high Prentice AM. Critical evaluation of energy intake data using
coronary heart disease risk particularly affects serum LDL fundamental principles of energy physiology: 2. Evaluating
cholesterol in apolipoprotein E14-carrying free-living the results of published surveys. Eur. J. Clin. Nutr. 1991; 45:
individuals. Br. J. Nutr. 2000; 84: 885 – 90. 583 – 99.
8 Slattery ML, Potter JD, Samowitz W, Schaffer D, Leppert M. 29 Schofield WN, Schofield C, James WPT. Basal metabolic rate.
Methylenetetrahydrofolate reductase, diet, and risk of colon Hum. Nutr. Clin. Nutr. 1985; 39C(Suppl.): 1 –96.
cancer. Cancer Epidemiol. Biomark. Prev. 1999; 8: 513 – 8. 30 Black AE, Coward WA, Cole TJ, Prentice AM. Human energy
9 Palli D, Vineis P, Russo A, Berrino F, Krogh V, Masala G, et al. expenditure in affluent societies: an analysis of 574 doubly-
Diet, metabolic polymorphisms and DNA adducts: the labelled water measurements. Eur. J. Clin. Nutr. 1996; 50:
EPIC –Italy cross-sectional study. Int. J. Cancer 2000; 87: 72 –92.
444 – 51. 31 Goldberg GR, Black AE, Jebb SA, Cole TJ, Murgatroyd PR,
10 Garcia-Closas M, Rothman N, Lubin J. Misclassification in Coward WA, et al. Critical evaluation of energy intake data
case –control studies of gene –environment interactions: using fundamental principles of energy physiology: 1.
assessment of bias and sample size. Cancer Epidemiol. Derivation of cut-off limits to identify under-recording. Eur.
Biomark. Prev. 1999; 8: 1043 – 50. J. Clin. Nutr. 1991; 45: 569 – 81.
11 Riboli E, Kaaks R. The EPIC Project: rationale and study 32 Denis W, Borgstrom P. A study of the effect of temperature
design. European Prospective Investigation into Cancer and on protein intake. J. Biol. Chem. 1924; 61: 109 – 16.
Nutrition. Int. J. Epidemiol. 1997; 26(Suppl. 1): S6– S14. 33 Isaksson B. Urinary nitrogen output as a validity test in
12 Bingham SA. The dietary assessment of individuals: dietary surveys. Am. J. Clin. Nutr. 1980; 33: 4– 5.
methods, accuracy, new techniques and recommendations. 34 Van Staveren WA, de Boer JO, Burema J. Validity and
Nutr. Abstr. Rev. 1987; 57: 705 –42. reproducibility of a dietary history method estimating the
13 Cameron ME, van Staveren WA, eds. Manual on usual food intake during one month. Am. J. Clin. Nutr. 1985;
Methodology for Food Consumption Studies. Oxford Medical 42: 554– 9.
Publications. Oxford: Oxford University Press, 1988. 35 Bingham SA, Cummings JH. Urine nitrogen as an
14 Willett W. Nutritional Epidemiology. Monographs in independent validatory measure of dietary intake: a study
Epidemiology and Biostatistics, 2nd ed. New York: Oxford of nitrogen balance in individuals consuming their normal
University Press, 1988. diet. Am. J. Clin. Nutr. 1985; 42: 1276– 89.
15 Margetts BM, Nelson M. Design Concepts in Nutritional 36 Kipnis V, Midthune D, Freedman LS, Bingham SA, Schatzkin A,
Biomarkers in nutritional epidemiology 827
Subar A, et al. Empirical evidence of correlated biases in 46 Black AE. The logistics of dietary surveys. Hum. Nutr. Appl.
dietary assessment instruments and its implications. Am. Nutr. 1982; 36: 85 –94.
J. Epidemiol. 2001; 153: 394 – 403. 47 Holbrook JT, Patterson KY, Bodner JE, Douglas LW, Veillon
37 Bingham SA, Williams R, Cole TJ, Price CP, Cummings JH. C, Kelsay JL, et al. Sodium and potassium intake and balance
Reference values for analytes of 24-h urine collections in adults consuming self-selected diets. Am. J. Clin. Nutr.
known to be complete. Ann. Clin. Biochem. 1988; 25: 1984; 40: 786– 93.
610 – 9. 48 Bingham SA, Goldberg GR, Coward WA, Prentice AM,
38 Black AE, Bingham SA, Johansson G, Coward WA. Validation Cummings JH. The effect of exercise and improved physical
of dietary intakes of protein and energy against 24 hour fitness on basal metabolic rate. Br. J. Nutr. 1989; 61: 155 – 73.
urinary N and DLW energy expenditure in middle-aged 49 Barlow RJ, Connell MA, Milne FJ. A study of 48-hour faecal
women, retired men and post-obese subjects: comparisons
and urinary electrolyte excretion in normotensive black and
with validation against presumed energy requirements. Eur.
white South African males. J. Hypertens. 1986; 4: 197 – 200.
J. Clin. Nutr. 1997; 51: 405 –13.
39 Bingham SA, Cassidy A, Cole TJ, Welch A, Runswick SA, 50 Bingham S. Validation of dietary assessment through
Black AE, et al. Validation of weighed records and other biomarkers. In: Kok FJ, van’t Veer P, eds. Biomarkers of
methods of dietary assessment using the 24 h urine nitrogen Dietary Exposure, Proceedings of the 3rd Meeting on
technique and other biological markers. Br. J. Nutr. 1995; 73: Nutritional Epidemiology. London: Smith-Gordon, 1991.
531 – 50. 51 Bingham SA, Gill C, Welch A, Cassidy A, Runswick SA, Oakes
40 Bandini LG, Cyr H, Must A, Dietz WH. Validity of reported S, et al. Validation of dietary assessment methods in the UK
energy intake in preadolescent girls. Am. J. Clin. Nutr. 1997; arm of EPIC using weighed records, and 24-hour urinary
65: 1138S– 41S. nitrogen and potassium and serum vitamin C and
41 Steen B, Isaksson B, Svanborg A. Intake of energy and carotenoids as biomarkers. Int. J. Epidemiol. 1997;
nutrients and meal habits in 70-year-old males and females 26(Suppl. 1): S137 – 51.
in Gothenburg, Sweden. A population study. Acta Med. 52 Kaaks R, Riboli E. Validation and calibration of dietary intake
Scand. 1977; 611: 39– 86. measurements in the EPIC project: methodological con-
42 Warnold I, Carlgren G, Krotkiewski M. Energy expenditure siderations. European Prospective Investigation into Cancer
and body composition during weight reduction in hyper- and Nutrition. Int. J. Epidemiol. 1997; 26(Suppl. 1): S15 – 25.
plastic obese women. Am. J. Clin. Nutr. 1978; 31: 750–63. 53 Slimani N, Deharveng G, Charrondière RU, van Kappel AL,
43 Heitmann BL, Lissner L. Dietary underreporting by obese Ocké MC, Welch A, et al. Structure of the standardized
individuals – is it specific or non-specific?. Br. Med. J. 1995; computerized 24-h diet recall interview used as reference
311: 986– 9. method in the 22 centers participating in the EPIC project.
44 Visser M, de Groot LCPGM, Deurenberg P, van Staveren WA.
European Prospective Investigation into Cancer and Nutri-
Validation of dietary history method in a group of elderly
tion. Comput. Methods Programs Biomed. 1999; 58: 251 – 66.
women using measurements of total energy expenditure. Br.
J. Nutr. 1995; 74: 775 – 85. 54 McKeown NM, Welch AM, Runswick SA, Luben R, Mulligan
45 Kroke A, Klipstein-Grobusch K, Voss S, Möseneder J, A, McTaggart A, et al. The use of biological markers to
Thielecke F, Noack R, et al. Validation of a self-administered validate self reported dietary intake in a random sample of
food-frequency questionnaire administered in the European the European Prospective Investigation into Cancer (EPIC)
Prospective Investigation into Cancer and Nutrition (EPIC) UK Norfolk Cohort. Am. J. Clin. Nutr. 2001; 74: 188 –96.
Study: comparison of energy, protein, and macronutrient 55 Day NE, McKeown N, Wong MY, Welch A, Bingham S.
intakes estimated with the doubly labeled water, urinary Epidemiological assessment of diet: a comparison of a 7 day
nitrogen, and repeated 24-h dietary recall methods. Am. diary with a food frequency questionnaire. Int. J. Epidemiol.
J. Clin. Nutr. 1999; 70: 439 –47. 2001; 30: 309– 17.