American Journal of Epidemiology
ª The Author 2007. Published by the Johns Hopkins Bloomberg School of Public Health.
All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org.
Vol. 166, No. 6
DOI: 10.1093/aje/kwm133
Advance Access publication July 13, 2007
Original Contribution
Is High Consumption of Fatty Fish during Pregnancy a Risk Factor for Fetal
Growth Retardation? A Study of 44,824 Danish Pregnant Women
Th. I. Halldorsson1, H. M. Meltzer2, I. Thorsdottir3,4, V. Knudsen1, and S. F. Olsen1,5
1
Maternal Nutrition Group, Department of Epidemiology Research, Statens Serum Institut, Copenhagen, Denmark.
Division of Environmental Medicine, Norwegian Institute of Public Health, Oslo, Norway.
3
Unit for Nutrition Research, Landspı́tali University Hospital and University of Iceland, Reykjavı́k, Iceland.
4
Department of Food Science and Nutrition, Faculty of Science, University of Iceland, Reykjavı́k, Iceland.
5
Department of Epidemiology, Institute of Public Health, University of Aarhus, Aarhus, Denmark.
2
The authors examined the relation between fish consumption during pregnancy and fetal growth among 44,824
women from the Danish National Birth Cohort (1996–2002). They evaluated the associations between consumption of total fish, fatty fish, and lean fish in midpregnancy and birth weight, birth length, and head circumference
among singleton full-term infants. Fish consumption was ascertained by food frequency questionnaire. The birth of
infants classified below the 10th percentile for gestational age and gender was significantly increased among
women who consumed more than 60 g of fish per day, as compared with women who consumed 5 g or less per
day. Adjusted odds ratios were 1.24 (95% confidence interval (CI): 1.03, 1.49) for birth weight and 1.21 (95% CI:
1.01, 1.43) for head circumference. The adjusted odds ratio was borderline significant for birth length (odds ratio ¼
1.20, 95% CI: 1.00, 1.45). These increases in risk were followed by small decreases in average values for these
growth measures. Furthermore, the inverse association for total fish consumption could be explained by consumption of fatty fish, while no association was found for lean fish. These results indicate that consumption of fatty fish,
a known route of exposure to persistent organic pollutants, could be associated with reduced fetal growth.
birth weight; Denmark; diet; fetal development; fishes; pregnancy; seafood
Abbreviations: CI, confidence interval; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid.
Fish contains many nutritionally important components,
including marine n-3 polyunsaturated fatty acids, protein,
vitamin D, and minerals such as selenium and iodine. Previous observational studies (1–4) have shown a positive association between fish consumption and birth weight and
prolonged gestation. While the observed increase in birth
weight was related to both increased length of gestation and
increased fetal growth, much attention has been given to the
role of marine n-3 fatty acids, particularly eicosapentaenoic
acid (EPA) and docosahexaenoic acid (DHA), which have
shown positive associations with prolonged gestation in
several randomized controlled trials (5–8). Other trials,
however, have found no association (9–11). In a large randomized controlled trial focusing on fish oil supplements
and length of gestation conducted within the Danish National
Birth Cohort (12), no association was found, and it appears
that the importance of these fatty acids in length of gestation
is far from clear. In a recent observational study from
the United States (13), fish consumption was even found to
be inversely associated with fetal growth, while no association was observed with length of gestation. Another recent
observational study from Iceland (14) found a positive association between fish consumption and fetal growth but an
inverse association with high intake of fish liver oil.
Correspondence to Thorhallur Ingi Halldorsson, Department of Epidemiology Research, Statens Serum Institut, 5 Artillerivej, 2300
Copenhagen S, Denmark (e-mail: lur@ssi.dk).
687
Am J Epidemiol 2007;166:687–696
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
Received for publication October 19, 2006; accepted for publication March 19, 2007.
688 Halldorsson et al.
MATERIALS AND METHODS
Population and study design
This study was based on data from the Danish National
Birth Cohort, whose structure has been described elsewhere
(20). In brief, 101,046 pregnant women from throughout
Denmark were recruited between 1996 and 2002. All pregnant women living in Denmark who were fluent in Danish
were eligible for recruitment. Recruitment took place
around weeks 6–10 of gestation during the first antenatal
visit to the general practitioner.
It was estimated that during the study period, approximately 30 percent of all deliveries in Denmark were covered
by the cohort (20). Nohr et al. (21) concluded that this
participation rate should not result in biased estimates of
association between lifestyle factors and health outcomes,
such as smoking and risk of small-for-gestational-age birth.
Data collection in this study consisted of a recruitment form
handed out at the first antenatal visit, a self-administered food
frequency questionnaire (22) sent by mail around week 25 of
gestation, and two computer-assisted telephone interviews
of 10–15 minutes’ duration each, administered around weeks
12 and 30 of gestation. Information on birth outcomes was
obtained through linkage to the Medical Birth Registry, which
includes all deliveries occurring in the country.
Dietary assessment
Information on fish consumption was collected through
the food frequency questionnaire (22). The questionnaire
was a modified form of the questionnaire used by the Danish
Cancer Registry (23) and has been validated for use in pregnancy (24). The questionnaire solicited information on frequency and type of fish consumed, either as a meal or with
bread.
Definition of the fish variable
When evaluating the association between fish consumption and fetal growth, we used total fish intake quantified in
grams per day. Consumption of fish as a meal and consumption of fish with bread were combined using assumptions on
standard portion sizes (25). Average fish consumption in the
cohort was 27 g/day (standard deviation, 23), and we divided consumption into five categories: 0–5, >5–20, >20–
40, >40–60, and >60 g/day. By comparison, an average
consumption of 47 g/day has been reported among Icelandic
pregnant women (14), and among US pregnant women an
average of 6.4 servings per month has been reported (26),
which corresponds to approximately 32 g/day (assuming
that one serving equals 150 g).
When examining type of fish, we used frequency of meals
and divided fish consumption into fatty fish and lean fish.
Salmon, herring, mackerel, trout, and Greenland halibut
(Reinhardtius hippoglossoides) were classified as fatty fish,
while cod, pollack, plaice, flounder, garfish, and similar
species were classified as lean fish. According to this definition, the fat contribution ranged from 0.6 g (cod) to 2.7 g
(garfish) per 100 g of fresh weight for lean fish and from
6.7 g (trout) to 24.4 g (autumn herring) per 100 g of fresh
weight for fatty fish, based on the Danish food composition
tables (27). For both types of fish, consumption was divided
into four categories: 0, 1, 2–3, and 4 meals per month.
The outcome variables
Birth weight, birth length, and head circumference were
measured by the midwife who attended the birth. The date
of birth was extracted from the Danish Civil Registration
System. Gestational age was assessed from the last menstrual period, on the basis of information recorded on the
recruitment form (gestation week 6–10) and during the first
telephone interview (gestation week 12). If this estimate was
uncertain because of irregular or abnormally long (>32
days) or short (<24 days) menstrual cycles, gestational
age was calculated from the expected date of delivery,
which is most often based on ultrasound scanning, provided
by the woman during the second telephone interview (gestation week 30). If this information was missing or led to
unrealistic gestational age estimates (>308 days), we used
length of gestation assessed at delivery by the midwife and
reported to the Medical Birth Registry. For our gestational
age estimates, 47 percent, 52 percent, and 1 percent were
based on information on the last menstrual period, information obtained from the second telephone interview, and the
Medical Birth Registry, respectively.
To identify children born small for gestational age, we
used the growth reference curves provided by the British
Child Growth Foundation (28, 29). The growth reference
curves were used to calculate z scores, standardized for
gestational age and gender, for birth weight, birth length,
and head circumference. For a given growth measure, z
scores were calculated for all singleton births in the cohort.
On the basis of that distribution, we defined infants whose
z scores were below the 10th percentile as small for gestational age.
Am J Epidemiol 2007;166:687–696
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
Fish consumption is also a well-known route of exposure
to pollutants such as methyl mercury (15) and persistent
organic pollutants; exposure to the latter comes mainly from
consumption of fatty fish (16, 17). However, only a few
studies have shown a direct association between these contaminants and reduced fetal growth (18). An ecologic study
from Sweden (17) found that women from the eastern coast,
who eat fish from the contaminated Baltic Sea, had a higher
risk of giving birth to low birth weight (<2,500 g) infants
than did women on the western coast, where the locally
caught fish are less contaminated (19). A study from the
Faroe Islands (15) also showed an inverse association between serum levels of EPA and fetal growth, where exposure
to methyl mercury and polychlorinated biphenyls was high
through consumption of whale meat and blubber.
Our objectives in the present study were to examine the
association between fish consumption and fetal growth
among singleton full-term infants and to determine the importance of type of fish in this association by distinguishing
between fatty fish and lean fish. We used birth weight,
length, and head circumference adjusted for gestational
age as measures of fetal growth.
Fatty Fish Consumption and Fetal Growth Retardation
Mother-child pairs available for analysis
Statistical methods
We used linear regression to analyze continuous birth
outcomes and logistic regression for dichotomous birth outcomes. For pairwise comparison of group means, we employed Student’s t test, and for pairwise comparison of
dichotomous outcomes, we used chi-squared tests. We identified and included as covariates a set of nine nondietary
factors that are well-recognized determinants of fetal
growth: gestational age (in days), infant gender (binary variable), maternal parity (binary variable: nulliparous vs. parous), maternal age (<20, 20–40, or >40 years), maternal
height (<160, 160–169, 170–179, or >179 cm), maternal
prepregnancy body mass index (weight (kg)/height (m)2;
<18.5, 18.5–24.9, 25–29.9, 30–34.9, 35–40, or >40), maternal smoking (never smoking, occasional smoking, daily
smoking of <15 cigarettes, or daily smoking of 15 cigarettes), paternal height (<170, 170–179, 180–189, or >189
cm), and familial socioeconomic status (six occupational
categories). Familial socioeconomic status was based on
the occupation of both parents if they were living together;
otherwise, it was based on the mother’s occupation only.
Because of the correlation between intake of food and intake
of energy, we additionally included mother’s total energy
intake (in quintiles) as a covariate, since it is generally important to distinguish between the separate effects of food
and energy intake (30). For statistical analyses, we used
SAS software, version 9.1 (SAS Institute Inc., Cary, North
Carolina).
RESULTS
For the 44,824 women available for analyses, fish consumption contributed on average 1.7 percent of their total
energy intake, 5.7 percent of their total protein intake, and
2.7 percent of their total fat intake. Since most of the women
Am J Epidemiol 2007;166:687–696
reported occasional consumption of fish with bread, only 2.6
percent had a fish intake of zero. However, 8 percent of the
women had total consumption of 5 g/day or less, which
corresponds to less than one meal per month.
The characteristics of birth outcomes and covariates with
respect to total fish consumption are shown in table 1. Across
categories of increased fish consumption, there was a considerable decrease in the percentage of smokers, and women of
high socioeconomic status tended to eat more fish than
women of lower status. Women consuming fish regularly also
tended to be slightly older, having a lower prepregnancy body
mass index and a higher energy intake and greater parity.
Except for maternal age and prepregnancy body mass index,
the changes across categories for these characteristics were
found to be positively associated with fetal growth.
For birth characteristics, slight increases in birth weight
and length were observed with increased fish consumption,
while no association was observed with head circumference
or gestational age among these full-term births.
Total fish consumption
When examining the three measures of fetal growth as
continuous outcomes (table 2, upper half), we obtained in
the unadjusted analyses a positive association (p for trend <
0.05) with respect to birth weight but no association with
respect to birth length or head circumference. For birth
weight and length, the estimates were significant for moderate consumption, but significance was lost for the highest
consumption group. After covariate adjustment, the association was reversed, and fish consumption showed a weak but
inverse association with fetal growth for birth length and
head circumference and a borderline significant association
for birth weight. In the highest consumption group, the estimates were 25.2 g (95 percent confidence interval (CI):
47.4, 3.0) for decrease in birth weight, 0.08 cm (95
percent CI: 0.18, 0.02) for decrease in birth length, and
0.11 cm (95 percent CI: 0.18, 0.03) for decrease in
head circumference.
Maternal smoking, height, parity, and energy intake were
the covariates most responsible for reversing the association
observed in the unadjusted analyses in table 2. As an example, the unadjusted estimate for birth weight (upper half of
table) for the highest fish category was 14.7 g (95 percent
CI: 10.3, 39.7). Adjusting only for smoking resulted in
an estimate of 1.4 g (95 percent CI: 26.1, 23.2), and adding parity to the regression model produced an estimate of
17.8 g (95 percent CI: 42.1, 6.4). When maternal
height and energy intake were also included, the estimate
became 32.7 g (95 percent CI: 57.1, 8.4). Further
adjustment for prepregnancy body mass index increased
the estimate to 24.6 g (95 percent CI: 48.8, 0.4). Inclusion of the five remaining covariates then resulted in the
fully adjusted estimate of 25.2 g (95 percent CI: 47.4,
3.0).
It has previously been reported (2) that smoking might be
an effect modifier with respect to fish consumption and fetal
growth. We investigated the possibility of effect modification with respect to smoking, parity, and prepregnancy body
mass index. To simplify the analyses, we categorized the
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
A total of 101,042 women were registered in the Danish
National Birth Cohort during early pregnancy. Of those
women, 92,892 and 87,902 participated in interviews conducted around gestational weeks 12 and 30, respectively,
and provided information on maternal characteristics. The
number of women filling out the food frequency questionnaire was 70,183. Combining information from these three
sources reduced the number of available women to 66,120.
Restriction to singleton term births resulted in a study sample of 57,946 women, and of those, covariate information
was available for 50,618. Given the hypothesis that use of
fish oil supplements might increase birth weight and the
analytical difficulty of distinguishing between the separate
effects of food- and supplement-derived nutrients, we also
excluded women taking fish oil supplements; this reduced
the data set to 47,635 women. Due primarily to missing data
on birth outcomes and to observations falling out of the
realistic ranges for birth weight (1,000–6,000 g), birth
length (35–65 cm), and head circumference (25–45 cm)
for singleton term infants, the final data set contained
44,824 women.
689
690 Halldorsson et al.
TABLE 1. Parental and infant characteristics according to maternal fish consumption during pregnancy (n ¼ 44,824), Danish National
Birth Cohort, 1996–2002
Fish consumption (g/day)
5 (8.0%)
>5–20 (35.9%)
>20–40 (36.4%)
>40–60 (14.0%)
>60 (5.6%)
p value
Continuous variables (mean and standard deviation)
Gestational age (days)*
280.2 (8.1)
280.2 (8.0)
280.3 (7.9)
280.1 (7.9)
280.1 (7.9)
0.43y
Birth weight (g)
3,597 (498)
3,610 (492)
3,627 (483)
3,625 (500)
3,612 (493)
<0.001y
Birth length (cm)
52.3 (2.3)
52.4 (2.2)
52.5 (2.2)
52.4 (2.2)
52.4 (2.2)
0.02y
Head circumference (cm)
35.3 (1.6)
35.3 (1.6)
35.4 (1.6)
35.3 (1.6)
35.3 (1.6)
0.96y
Mother’s age (years)
27.9 (4.2)
28.7 (4.1)
29.5 (4.1)
29.9 (4.2)
29.6 (4.5)
<0.001y
Mother’s height (cm)
168.2 (6.2)
168.7 (6.1)
169.0 (6.0)
169.1 (6.1)
169.0 (6.1)
<0.001y
Prepregnancy body mass indexz
24.2 (4.7)
23.8 (4.3)
23.3 (3.9)
23.1 (4.0)
23.1 (4.0)
<0.001y
Mother’s total energy intake (MJ/day)
9.6 (2.8)
9.9 (2.6)
10.6 (2.6)
11.5 (2.7)
12.5 (3.3)
<0.001y
Father’s height (cm)
181.5 (7.1)
181.8 (7.0)
182.2 (6.9)
182.2 (6.9)
182.0 (6.9)
<0.001y
Discrete characteristics (%)
50.2
51.0
51.7
50.1
50.6
0.18§
Parous mother (%)
48.9
50.7
54.8
56.2
57.6
<0.001§
Maternal smoking (%)
30.6
24.4
22.1
21.1
25.3
<0.001§
High
5.2
31.5
40.3
17.0
6.1
Intermediate
6.7
35.5
38.3
14.4
5.2
Familial socioeconomic status (row %){
<0.001§
Skilled worker(s)
10.4
39.7
33.0
11.9
5.1
Unskilled worker(s)
11.5
38.5
31.4
12.4
6.3
Student(s)
7.2
34.1
37.5
14.1
7.0
Not working
11.4
34.1
34.1
11.8
8.7
* Sample was restricted to term births.
y Two-sided p value for association as determined by Spearman’s correlation coefficient.
z Weight (kg)/height (m)2.
§ Two-sided p value from chi-squared test for measure of association.
{ Percentages for each level of familial socioeconomic status. Familial socioeconomic status was based on the occupation of both parents if
they were living together; otherwise, it was based on the mother’s occupation only. Of the 44,824 observations, 24.2% of families were classified
as high status, 31.1% as intermediate status, 26.8% as skilled worker(s), 11.9% as unskilled worker(s), 3.8% as student(s), and 2.2% as not
working.
variables in the following way: smoking as yes versus no,
parity as nulliparous versus parous, and prepregnancy body
mass index as <18.5, 18.5–24.9, or 25. We used only the
lowest and highest categories of fish consumption (5 g/day
and >60 g/day). We tested for effect modification by including
the variable for total fish, the covariate of interest, and a term
for interaction between the two variables, both with and without adjustment for the nine remaining covariates. In all cases,
the interaction term was nonsignificant (data not shown).
For the dichotomous growth measures (table 2, lower
half), the odds ratio estimates for birth weight and length
in the unadjusted analyses showed a significantly decreased
risk of small-for-gestational-age birth for moderate fish consumption, but the estimates for the highest consumption
group were not significant. For the adjusted analyses, the
odds ratio estimates were centered around 1, except for the
highest consumption group, where the odds ratio estimates
showed a significantly increased risk of being small for
gestational age for birth weight (odds ratio ¼ 1.24, 95 percent CI: 1.03, 1.49) and head circumference (odds ratio ¼
1.21, 95 percent CI: 1.01, 1.43) and a borderline-significant
risk for birth length (odds ratio ¼ 1.20, 95 percent CI: 1.00,
1.45). In both unadjusted and adjusted analyses, tests for
linear trend were not significant.
Fatty fish versus lean fish
The distributions of fatty and lean fish consumption are
shown in table 3. The majority of the women (n ¼ 24,205)
did not consume any fatty fish, while consumption of lean
fish was more general. High consumption (4 meals/month)
of fatty fish and lean fish was observed for 2,977 and 10,720
women, respectively. Of those, 1,789 had high consumption
of both types. However, the correlation between fatty fish
and lean fish was only moderate (Spearman’s r ¼ 0.33, p <
0.001), which indicates that analyzing the two variables
separately is justifiable.
For fatty fish, using continuous growth measures (table 4,
upper half), we found a significant inverse association for
birth weight and head circumference in the unadjusted analyses, while all three growth measures showed an inverse
Am J Epidemiol 2007;166:687–696
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
Male infant gender (%)
Fatty Fish Consumption and Fetal Growth Retardation
691
TABLE 2. Associations between total fish consumption and measures of fetal growth (n ¼ 44,824) before and after covariate
adjustment, Danish National Birth Cohort, 1996–2002
Total fish
intake (g/day)
% of
subjects
Birth weight
Birth length
Head circumference
Continuous growth measures
Increase in birth
weight (g)
95% CI*
Increase in birth
length (cm)
95% CI
Increase in head
circumference (cm)
95% CI
Unadjusted
8.0
Referent
>5–20
35.9
12.9
4.8, 30.7
0.08
0.00, 0.17
0.02
0.04, 0.07
>20–40
36.4
30.2
12.5, 47.9
0.13
0.05, 0.22
0.06
0.00, 0.12
>40–60
14.0
28.4
8.3, 48.6
0.10
0.00, 0.19
0.00
0.06, 0.07
5.6
14.7
10.3, 39.7
0.07
0.05, 0.19
5
>60
p for trendy
Referent
0.02
Referent
0.04
0.28
0.12, 0.04
0.34
Adjustedz
Referent
35.9
12.6
28.1, 2.9
0.03
0.10, 0.04
0.04
0.09, 0.02
>20–40
36.4
11.6
27.3, 4.0
0.05
0.12, 0.03
0.02
0.07, 0.03
>40–60
14.0
14.0
31.9, 3.9
0.08
0.16, 0.01
0.07
0.13, 0.01
5.6
25.2
47.4, 3.0
0.08
0.18, 0.02
0.11
0.18, 0.03
>60
p for trend
Referent
0.09
Referent
0.04
0.005
Dichotomized growth measures
SGA* for birth weight
SGA for birth length
OR*
OR
95% CI
95% CI
SGA for head circumference
OR
95% CI
Unadjusted
5
8.0
1
1
1
>5–20
35.9
0.94
0.82, 1.06
0.89
0.78, 1.01
0.97
0.86, 1.10
>20–40
36.4
0.82
0.72, 0.93
0.79
0.69, 0.90
0.89
0.79, 1.01
>40–60
14.0
0.86
0.74, 1.00
0.86
0.74, 0.99
0.90
0.78, 1.04
5.6
0.99
0.83, 1.18
1.03
0.86, 1.23
1.08
0.92, 1.28
>60
p for trend
0.26
0.96
0.85
Adjusted§
8.0
1.00
>5–20
35.9
1.08
0.95, 1.23
1.01
0.88, 1.15
1.04
0.92, 1.18
>20–40
36.4
1.02
0.90, 1.17
0.95
0.83, 1.09
0.99
0.87, 1.12
>40–60
14.0
1.10
0.95, 1.29
1.05
0.89, 1.23
1.02
0.88, 1.17
5.6
1.24
1.03, 1.49
1.20
1.00, 1.45
1.21
1.01, 1.43
5
>60
p for trend
1.00
0.08
1.00
0.07
0.20
* CI, confidence interval; SGA, small for gestational age; OR, odds ratio.
y Two-sided p value.
z Adjusted for gestational age, infant gender, parity, maternal age, maternal height, prepregnancy body mass index, energy intake, smoking,
familial socioeconomic status, and paternal height.
§ Adjusted for the same covariates as in the upper half of the table, apart from gestational age and gender, which were adjusted for in the
z scores.
association in the adjusted analyses. The estimates for the
highest consumption group in the adjusted analyses were
significant for all three growth measures and were similar
in magnitude to the estimates for total fish consumption.
With respect to the dichotomous growth measures (table 4,
lower half), in the adjusted analyses we found a significantly increased risk of small size for gestational age for
Am J Epidemiol 2007;166:687–696
birth weight and birth length among women who consumed
fatty fish four times per month or more as compared with
those who did not consume fatty fish, while the increase for
head circumference was nonsignificant. Corresponding odds
ratios were 1.18 (95 percent CI: 1.03, 1.35) for birth weight,
1.22 (95 percent CI: 1.05, 1.40) for birth length, and 1.10
(95 percent CI: 0.97, 1.25) for head circumference. A test
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
8.0
>5–20
5
692 Halldorsson et al.
TABLE 3. Cross-classification of fish meals according to type of fish consumed (n ¼ 44,824), Danish National Birth Cohort,
1996–2002
Lean* fish
consumption
(no. of meals
per month)
Fattyy fish consumption (no. of meals per month)
0
1
2–3
No.
Row total
%z
No.
0
7,809
17.4
1,234
2.8
820
1.8
1
5,052
11.3
1,745
3.9
1,194
2.7
2–3
7,698
17.2
4,094
9.1
3,270
7.3
760
1.7
15,822
35.3
4
3,646
8.1
2,333
5.2
2,952
6.6
1,789
4.0
10,720
23.9
24,205
54.0
9,406
21.0
8,236
18.4
2,977
6.6
44,824
Column total
%z
4
No.
%z
No.
%z
No.
%
182
0.4
10,045
22.4
246
0.5
8,237
18.4
100
* Cod, pollack, plaice, flounder, garfish, and similar species were classified as lean fish.
y Salmon, herring, mackerel, trout, and Greenland halibut (Reinhardtius hippoglossoides) were classified as fatty fish.
z Percentage of total study population.
DISCUSSION
In a large cohort of pregnant women, high intake of fatty
fish was found to be inversely related to birth weight, birth
length, and head circumference. The average change in
these measures of fetal growth was small, but it was associated with an increased risk of children being born small
for gestational age with respect to birth weight, birth length,
and head circumference. No association was observed for
lean fish, and the inverse association observed for total fish
consumption can be explained by intake of fatty fish.
It is difficult to compare these results directly with those
of other studies, since information on type of fish consumed
is often sparse and the exposure variable is assessed in different ways. At least two other studies focusing on intake of
marine n-3 fatty acids, which can be regarded as a marker
for fish consumption (31), have found inverse associations
with fetal growth. Grandjean et al. (15) found an inverse
association between serum levels of EPA and birth weight
in a Faroese fishing community, where fish consumption
was high. However, in that study, a marine diet consisted
not only of fish but also of whale meat and blubber—
substances that are potentially high in persistent organic
pollutants and mercury, which may influence fetal growth
(32–34). A study by Oken et al. (13) focused on quantified
EPA and DHA estimated by food frequency questionnaire.
They found the sum of EPA and DHA to be associated with
reduced fetal growth. In attempting to compare our results
with those of that study, we could not in our data distinguish
between the association of EPA plus DHA with fetal growth,
on the one hand, and the association of fatty fish with fetal
growth on the other, because of high collinearity.
Thorsdottir et al. (14) found a positive association of fish
consumption with birth length and head circumference
among Icelandic women. In Iceland, as in the Faroe Islands,
fish consumption is mostly based on locally caught fish,
which is predominantly lean fish. With respect to lean
fish, consumption in the Icelandic study was considerably
higher than in our cohort. Thorsdottir et al. (14) also found
a decreased birth length and head circumference among
women consuming a high amount of cod liver oil (>8.7 g/
day), a product which is supposed to have been purified with
respect to persistent organic pollutants. However, despite
purification, fish oil may contain substantial amounts of
these pollutants (35, 36). We excluded women taking fish
oil supplements (5 percent), but for total fish consumption,
the women in the highest category had an average intake of
7.6 g of fish fat per day. We therefore have evidence that
high amounts of marine fats, consumed through either intake of fish oil (14), consumption of marine mammals (15),
or intake of fatty fish (as in this study), might have an inverse
association with fetal growth. However, it is not clear
whether this inverse association is due to industrial contaminants, fatty acid composition, or some other constituent in
the fish fat.
One explanation for the inverse association observed for
fatty fish might be contamination by persistent organic pollutants. A role for methyl mercury is less likely, since
methyl mercury is also present in lean fish (36), because
of protein binding. With the exception of a few studies
(32, 33), most studies focusing on persistent organic pollutants such as polychlorinated biphenyls have failed to find
an association with fetal growth (18). Because of the complexity and cost of measuring these substances, these studies
usually contain few subjects and have low statistical power.
Am J Epidemiol 2007;166:687–696
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
for linear trend was significant for birth weight and length in
the adjusted analyses.
When we evaluated consumption of lean fish (table 5), the
results were almost opposite those for fatty fish. In the unadjusted analyses, the estimates showed an increase in birth
weight, birth length, and head circumference and a decreased risk of being small for gestational age for these
growth measures. After adjustment for covariates, the
estimates become nonsignificant and no association was
observed for either continuous growth measures or dichotomized growth measures.
To check the stability of our estimates for fatty and lean
fish and to account for their collinearity, we included those
variables simultaneously in the regression model. This mutual adjustment had only a minor impact on estimates, and
the same conclusions were reached for all three growth
measures (data not shown).
Fatty Fish Consumption and Fetal Growth Retardation
693
TABLE 4. Associations between consumption of fatty fish during pregnancy and measures of fetal growth (n ¼ 44,824), before and
after adjustment for covariates, Danish National Birth Cohort, 1996–2002
Fatty* fish intake
(no. of meals/month)
% of
subjects
Birth weight
Birth length
Head circumference
Continuous growth measures
Increase in birth
weight (g)
95% CIy
Increase in birth
length (cm)
95% CI
Increase in head
circumference (cm)
95% CI
Unadjusted
0
54.1
1
21.0
Referent
2–3
18.4
1.7
4
6.6
30.7
3.0
p for trendz
Referent
8.7, 14.7
13.9, 10.6
49.3, 12.0
Referent
0.09
0.03, 0.14
0.08
0.02, 0.14
0.02
0.06, 0.02
0.16, 0.02
0.12
0.18, 0.06
0.07
0.005
0.01
0.94
0.03, 0.04
<0.001
Adjusted§
54.1
Referent
1
21.0
4.8
15.1, 5.5
0.03
0.01, 0.08
0.01
0.04, 0.03
2–3
18.4
10.1
20.9, 0.8
0.00
0.05, 0.05
0.04
0.07, 0.00
4
6.6
27.5
43.8, 11.1
0.18, 0.03
0.11
0.16, 0.05
p for trend
Referent
0.10
Referent
0.03
<0.001
<0.001
Dichotomized growth measures
SGAy for birth weight
ORy
95% CI
SGA for birth length
OR
95% CI
SGA for head circumference
OR
95% CI
Unadjusted
0
54.1
1
1
21.0
0.90
0.82, 0.98
0.89
0.81, 0.98
0.97
0.89, 1.06
2–3
18.4
0.93
0.85, 1.02
0.99
0.90, 1.09
1.02
0.94, 1.12
4
6.6
1.12
0.98, 1.28
1.13
0.98, 1.30
1.10
0.97, 1.26
p for trend
1
0.42
1
0.18
0.15
Adjusted{
0
54.1
1
1
21.0
0.97
0.89, 1.06
0.97
0.88, 1.07
1.00
0.91, 1.09
2–3
18.4
1.00
0.91, 1.10
1.09
0.99, 1.21
1.05
0.96, 1.15
4
6.6
1.18
1.03, 1.35
1.22
1.05, 1.40
1.10
0.97, 1.25
p for trend
1
0.04
0.003
0.12
* Salmon, herring, mackerel, trout, and Greenland halibut (Reinhardtius hippoglossoides) were classified as fatty fish.
y CI, confidence interval; SGA, small for gestational age; OR, odds ratio.
z Two-sided p value.
§ Adjusted for gestational age, infant gender, parity, maternal age, maternal height, prepregnancy body mass index, energy intake, smoking,
familial socioeconomic status, and paternal height.
{ Adjusted for the same covariates as in the upper half of the table, apart from gestational age and gender, which were adjusted for in the
z scores.
In addition, the possibility cannot be excluded that certain
subgroups might be more sensitive to these substances than
others, which small studies would fail to detect (37).
Although many of the studies in the literature focus on
polychlorinated biphenyls, other compounds such as hexachlorobenzene have been shown to be associated with reduced fetal growth (38) and are potentially present in fish.
However, the absence of information regarding levels of
persistent organic pollutants in our study makes the above
arguments speculative.
Am J Epidemiol 2007;166:687–696
The inverse association observed for fatty fish might also
be related to the marine n-3 fatty acids. In postnatal feeding
trials with preterm infants, Carlson et al. (39) observed reduced growth among infants who were given formula fortified with marine n-3 fatty acids as compared with controls.
It was hypothesized that the growth-retarding effect was
mediated through EPA by displacement of arachidonic acid,
but this hypothesis has not been substantiated (40).
The strengths of this study include the facts that we had
prospective data on a large number of pregnant women, that
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
0
694 Halldorsson et al.
TABLE 5. Associations between consumption of lean fish during pregnancy and measures of fetal growth (n ¼ 44,824), before and
after covariate adjustment for covariates, Danish National Birth Cohort, 1996–2002
Lean* fish intake
(no. of meals/month)
% of
subjects
Birth weight
Birth length
Head circumference
Continuous growth measures
Increase in birth
weight (g)
95% CIy
Increase in birth
length (cm)
Increase in head
circumference (cm)
95% CI
95% CI
Unadjusted
0
24.7
Referent
1
20.1
30.1
Referent
15.8, 44.4
0.12
Referent
0.05, 0.19
0.09
0.04, 0.13
2–3
35.9
42.7
30.4, 55.0
0.16
0.10, 0.21
0.06
0.02, 0.10
4
19.3
44.7
31.3, 58.0
0.12
0.06, 0.19
0.07
0.03, 0.11
p for trendz
0.004
<0.001
0.04
Adjusted*
0
24.7
Referent
1
20.1
9.1
3.4, 21.6
0.03
0.03, 0.09
0.04
2–3
35.9
9.4
1.4, 20.2
0.01
0.04, 0.06
0.00
0.04, 0.04
4
19.3
1.9
10.0, 13.8
0.05
0.11, 0.01
0.01
0.05, 0.03
0.82
Referent
0.06
0.00, 0.09
0.16
Dichotomized growth measures
SGA for birth weight
OR
95% CI
SGA for birth length
OR
95% CI
SGA for head circumference
OR
95% CI
Unadjusted
0
24.7
1
1
20.1
0.88
1
0.79, 0.97
0.91
1
0.81, 1.01
0.84
0.76, 0.93
2–3
35.9
0.81
0.74, 0.88
0.87
0.79, 0.96
0.92
0.84, 1.00
4
19.3
0.79
0.71, 0.87
0.93
0.84, 1.03
0.87
0.79, 0.96
p for trend
0.33
<0.001
0.06
Adjusted§
0
24.7
1.00
1
20.1
0.96
0.86, 1.06
0.99
0.88, 1.11
0.89
0.79, 1.00
2–3
35.9
0.94
0.86, 1.03
1.01
0.91, 1.11
1.00
0.92, 1.09
4
19.3
0.96
0.86, 1.06
1.10
0.99, 1.22
0.98
0.89, 1.08
p for trend
1.00
0.51
1.00
0.05
0.66
* Cod, pollack, plaice, flounder, garfish, and similar species were classified as lean fish.
y CI, confidence interval; SGA, small for gestational age; OR, odds ratio.
z Two-sided p value.
§ Adjusted for gestational age, infant gender, parity, maternal age, maternal height, prepregnancy body mass index, energy intake, smoking,
familial socioeconomic status, and paternal height.
{ Adjusted for the same covariates as in the upper half of the table, apart from gestational age and gender, which were adjusted for in the z
scores.
cohort members were recruited throughout Denmark, and
that we gathered detailed information on both fish consumption and maternal characteristics. However, the results of
this study are subject to certain limitations. As with all
observational studies, we cannot rule out the possibility that
the observed association resulted from the influence of unadjusted or unmeasured confounders. Another limitation is
that fish consumption in our cohort was strongly associated
with lifestyle (smoking), socioeconomic status, and maternal factors such as parity and prepregnancy body mass index, which are strong predictors of fetal growth. We did
check the stability of our estimates after covariate adjustment by using restricted cubic spline regression for continuous variables as an alternative (41), but changes in
estimates were minor. The consistency between the unadjusted and adjusted results for fatty fish also makes it less
likely that our results are due to residual confounding.
In this large cohort of pregnant Danish women, women
who consumed fatty fish four times per month or more
had a higher risk of giving birth to children who were
small for gestational age with respect to birth weight, birth
length, and head circumference, while consumption of
Am J Epidemiol 2007;166:687–696
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
p for trend
Referent
Fatty Fish Consumption and Fetal Growth Retardation
lean fish had no association with these growth measures.
We conclude that for pregnant women, the type of fish
consumed is important, and moderate consumption of fatty
fish should be encouraged. Further studies distinguishing
between fatty fish and lean fish in relation to fetal growth
are warranted.
ACKNOWLEDGMENTS
REFERENCES
1. Olsen SF, Hansen HS, Sorensen TI, et al. Intake of marine fat,
rich in (n-3)-polyunsaturated fatty acids, may increase birth
weight by prolonging gestation. Lancet 1986;2:367–9.
2. Olsen SF, Olsen J, Frische G. Does fish consumption during
pregnancy increase fetal growth? A study of the size of the
newborn, placental weight and gestational age in relation to
fish consumption during pregnancy. Int J Epidemiol 1990;
19:971–7.
3. Olsen SF, Grandjean P, Weihe P, et al. Frequency of seafood
intake in pregnancy as a determinant of birth weight: evidence
for a dose dependent relationship. J Epidemiol Community
Health 1993;47:436–40.
4. Olsen SF, Secher NJ. Low consumption of seafood in early
pregnancy as a risk factor for preterm delivery: prospective
cohort study. BMJ 2002;324:447.
5. Smuts CM, Huang M, Mundy D, et al. A randomized trial
of docosahexaenoic acid supplementation during the third
trimester of pregnancy. Obstet Gynecol 2003;101:469–79.
6. Olsen SF, Secher NJ. A possible preventive effect of lowdose fish oil on early delivery and pre-eclampsia: indications
from a 50-year-old controlled trial. Br J Nutr 1990;64:
599–609.
7. Olsen SF, Secher NJ, Tabor A, et al. Randomised clinical
trials of fish oil supplementation in high risk pregnancies.
Fish Oil Trials In Pregnancy (FOTIP) Team. BJOG 2000;
107:382–95.
8. Olsen SF, Sorensen JD, Secher NJ, et al. Randomised controlled trial of effect of fish-oil supplementation on pregnancy
duration. Lancet 1992;339:1003–7.
Am J Epidemiol 2007;166:687–696
9. Helland IB, Saugstad OD, Smith L, et al. Similar effects on
infants of n-3 and n-6 fatty acids supplementation to
pregnant and lactating women. Pediatrics 2001;108:E82.
(Electronic article).
10. Bulstra-Ramakers MT, Huisjes HJ, Visser GH. The effects of
3g eicosapentaenoic acid daily on recurrence of intrauterine
growth retardation and pregnancy induced hypertension. Br J
Obstet Gynaecol 1995;102:123–6.
11. Onwude JL, Lilford RJ, Hjartardottir H, et al. A randomised
double blind placebo controlled trial of fish oil in high risk
pregnancy. Br J Obstet Gynaecol 1995;102:95–100.
12. Knudsen VK, Hansen HS, Osterdal ML, et al. Fish oil in
various doses or flax oil in pregnancy and timing of spontaneous delivery: a randomised controlled trial. BJOG 2006;
113:536–43.
13. Oken E, Kleinman KP, Olsen SF, et al. Associations of seafood
and elongated n-3 fatty acid intake with fetal growth and
length of gestation: results from a US pregnancy cohort. Am
J Epidemiol 2004;160:774–83.
14. Thorsdottir I, Birgisdottir BE, Halldorsdottir S, et al. Association of fish and fish liver oil intake in pregnancy with
infant size at birth among women of normal weight before
pregnancy in a fishing community. Am J Epidemiol 2004;
160:460–5.
15. Grandjean P, Bjerve KS, Weihe P, et al. Birth weight in
a fishing community: significance of essential fatty acids
and marine food contaminants. Int J Epidemiol 2002;30:
1272–8.
16. Jackson A, Key T, Williams C, et al. Advice on fish consumption: benefits and risks. Committee on Toxicity,
Scientific Advisory Committee on Nutrition. London, United
Kingdom: The Stationery Office, 2004. (http://www.food.
gov.uk/multimedia/pdfs/fishreport2004full.pdf). (Accessed
February 20, 2007).
17. Rylander L, Stromberg U, Hagmar L. Lowered birth weight
among infants born to women with a high intake of fish
contaminated with persistent organochlorine compounds.
Chemosphere 2000;40:1255–62.
18. Kimbrough RD, Krouskas CA. Polychlorinated biphenyls,
dibenzo-p-dioxins, and dibenzofurans and birth weight and
immune and thyroid function in children. Regul Toxicol
Pharmacol 2001;34:42–52.
19. Bergqvist P-A, Bergek S, Hallbäck H, et al. Dioxins in cod
and herring from the seas around Sweden. Chemosphere
1989;19:513–16.
20. Olsen J, Melbye M, Olsen SF, et al. The Danish National Birth
Cohort—its background, structure and aim. Scand J Public
Health 2001;29:300–7.
21. Nohr EA, Frydenberg M, Henriksen TB, et al. Does low
participation in cohort studies induce bias? Epidemiology
2006;17:413–18.
22. Olsen SF, Mikkelsen TB, Knudsen VK, et al. Data
collected on maternal dietary exposures in the Danish
National Birth Cohort. Paediatr Perinat Epidemiol 2007;21:
76–86.
23. Overvad K, Tjonneland A, Haraldsdottir J, et al. Development
of a semiquantitative food frequency questionnaire to assess
food, energy and nutrient intake in Denmark. Int J Epidemiol
1991;20:900–5.
24. Mikkelsen TB. Validation of the food frequency questionnaire used in the Danish National Birth Cohort. (Doctoral
thesis). Copenhagen, Denmark: Statens Serum Institute, 2004.
(http://www.ssi.dk/sw15332.asp).
25. Andersen LT, Jensen H, Haraldsdottir J. Typiske vægte for
madvarer. Scand J Nutr 1996;40(suppl 32):S129–52.
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
Financial support for this study was obtained from the
Nordic Academy for Advanced Study, the Nordic Working
Group on Fishery Research, and the Early Nutrition Programming Project (EARNEST) (European Union Sixth
Framework Programme, project FOOD-CT-2005-007036).
The March of Dimes Birth Defects Foundation supported
the collection of the dietary data. Financial support for the
Danish National Birth Cohort was obtained from the
Danish National Research Foundation, the Pharmacy
Foundation, the Egmont Foundation, the Augustinus Foundation, the Health Foundation, the European Union (grant
QLK1-2000-00083), the Danish Medical Research Foundation, and the Heart Foundation.
The managerial team of the Danish National Birth
Cohort consists of Drs. Jørn Olsen (Chair), Mads Melbye,
Anne Marie Nybo Andersen, Sjurdur F. Olsen, Thorkild I.
A. Sørensen, and Peter Aaby.
Conflict of interest: none declared.
695
696 Halldorsson et al.
35. Fernandes AR, Rose M, White S, et al. Dioxins and polychlorinated biphenyls (PCBs) in fish oil dietary supplements:
occurrence and human exposure in the UK. Food Addit
Contam 2006;23:939–47.
36. Fromberg A, Larsen EH, Hartkopp H, et al. Chemical
contaminants. Food monitoring, 1998–2003. Part 1. Søborg,
Denmark: Danish Veterinary and Food Administration,
Ministry of Family and Consumer Affairs, 2005. (http://
www.food.dtu.dk/Admin/Public/Download.aspx?file¼Files/
Filer/Publikationer/Rapport_Chemical_contaminants.pdf).
(Accessed February 5, 2007).
37. Schade G, Heinzow B. Organochlorine pesticides and polychlorinated biphenyls in human milk of mothers living in
northern Germany: current extent of contamination, time
trend from 1986 to 1997 and factors that influence the levels
of contamination. Sci Total Environ 1998;215:31–9.
38. Ribas-Fito N, Sala M, Cardo E, et al. Association of
hexachlorobenzene and other organochlorine compounds
with anthropometric measures at birth. Pediatr Res 2002;
52:163–7.
39. Carlson SE, Werkman SH, Tolley EA. Effect of long-chain n-3
fatty acid supplementation on visual acuity and growth of
preterm infants with and without bronchopulmonary dysplasia. Am J Clin Nutr 1996;63:687–97.
40. Makrides M, Gibson RA, Udell T, et al. Supplementation of
infant formula with long-chain polyunsaturated fatty acids
does not influence the growth of term infants. Am J Clin
Nutr 2005;81:1094–101.
41. Durrleman S, Simon R. Flexible regression models with
cubic splines. Stat Med 1989;8:551–61.
Am J Epidemiol 2007;166:687–696
Downloaded from http://aje.oxfordjournals.org/ by guest on April 13, 2016
26. Oken E, Kleinman KP, Berland WE, et al. Decline in fish
consumption among pregnant women after a national mercury advisory. Obstet Gynecol 2003;102:346–51.
27. National Food Institute (Denmark). Danish Food Composition
Databank. Søborg, Denmark: Department of Nutrition, Technical University of Denmark, 2007. (http://www.foodcomp.dk/
fcdb_default.asp).
28. Freeman JV, Cole TJ, Chinn S, et al. Cross sectional stature
and weight reference curves for the UK, 1990. Arch Dis
Child 1995;73:17–24.
29. Cole TJ, Freeman JV, Preece MA. British 1990 growth
reference centiles for weight, height, body mass index and
head circumference fitted by maximum penalized likelihood. Stat Med 1998;17:407–29.
30. Willet W. Nutritional epidemiology. 2nd ed. New York, NY:
Oxford University Press, 1998.
31. Olsen SF, Hansen HS, Sandström M, et al. Erythrocyte levels
compared with reported dietary intake of marine n-3 fatty
acids in pregnant women. Br J Nutr 1995;73:387–95.
32. Fein GG, Jacobson JL, Jacobson SW, et al. Prenatal exposure to polychlorinated biphenyls: effects on birth size and
gestational age. J Pediatr 1984;105:315–20.
33. Patandin S, Koopman-Esseboom C, de Ridder MA, et al.
Effects of environmental exposure to polychlorinated
biphenyls and dioxins on birth size and growth in Dutch
children. Pediatr Res 1998;44:538–45.
34. Foldspang A, Hansen JC. Dietary intake of methylmercury
as a correlate of gestational length and birth weight
among newborns in Greenland. Am J Epidemiol 1990;
132:310–17.