Article Text
PDF
Abstract
Background Studies on maternal seafood consumption during pregnancy and the risk of small for gestational age (SGA) births have yielded inconsistent results. As few studies have examined associations with specific seafood subtypes or accounted simultaneously for exposure to persistent organic pollutants (POPs), it is uncertain to what extent intakes of seafood subtypes with variable contaminant or fatty acid content may explain these inconsistencies.
Methods A cohort of 657 women recruited during the first trimester of pregnancy from a Mediterranean area with high seafood intakes was followed through birth. Dietary intakes were estimated using a validated questionnaire. Multivariable logistic regression was used to estimate associations between SGA and intakes of fatty fish, lean fish, canned tuna, crustaceans and other shellfish, adjusting for parity, child sex, parental anthropometry, socio-economic factors and serum levels of several POPs, including several polychlorinated biphenyls (PCBs), hexachlorobenzene (HCB) and dichlorodiphenyldichloroethylene (DDE).
Results Overall, 7.8% of infants were SGA. Maternal consumption of crustaceans (more than once/week) and canned tuna (more than once/week was associated with a significantly increased risk of SGA. Fatty fish (more than once/week) was associated with weaker and generally non-significant increases in risk, while lean fish and other shellfish were not associated with SGA. Adjusting for contaminants did not meaningfully change results.
Conclusions Higher maternal intakes of crustaceans and canned tuna, but not other types of seafood, were associated with increased risk of SGA independently of several POPs. Future studies exploring seafood subtypes and additional contaminants are needed to determine whether these associations are causal and identify mechanisms involved.
- Small for gestational age
- birth weight
- seafood
- pregnancy
- organochlorine compounds
- nutrition
- perinatal epidemiology
Statistics from Altmetric.com
- Small for gestational age
- birth weight
- seafood
- pregnancy
- organochlorine compounds
- nutrition
- perinatal epidemiology
Introduction
Poor fetal growth is associated with adverse short- and long-term health outcomes.1 2 Seafood is an important source of nutrients thought to be beneficial for fetal growth and development, notably the n-3 fatty acid docosahexanoic acid, as well as other nutrients such as protein, selenium, iodine and vitamin D.3 4 However, seafood may also be an important source of contaminants hypothesised to adversely affect fetal growth and other reproductive outcomes.5–8 In some countries, pregnant women have been advised to balance these risks and benefits by limiting their intakes of seafood to a maximum of three times weekly.9 10
Currently, the relationship between maternal seafood consumption and fetal growth is uncertain. While several studies have found maternal intakes of seafood or marine fatty acids to be associated with higher birth weight or reduced likelihood of small size for gestational age (SGA),11–16 others have reported null17 18 or negative associations.4 19 One study reported increases in birth weight up to the recommended threshold of three meals per week, with small declines at intakes exceeding this level.11 However, as most analyses have been conducted in populations with the highest level of consumption on the order of twice per week, few studies have been able to explicitly examine effects of seafood intakes above this threshold.4 12–19 Moreover, given the modest levels of intake, the majority of studies have examined fetal growth in relation to overall seafood consumption. Two recent articles suggest that the heterogeneous results of earlier studies may be partly explained by varying effects of different types of seafood. One study reported a non-significant reduction in SGA births associated with fish consumption, but a significant increase in SGA associated with higher intakes of crustaceans.20 While that study was unable to examine effects of different types of fish, another recent publication reported a significant increase in SGA associated with eating fatty but not lean fish at least once/week.4 Though direct measures of contaminants were not assessed, the authors hypothesised that these patterns of association may be partly attributable to higher concentrations of pollutants in fatty fish and crustaceans than in other types of seafood.
This study examines the relationship between maternal seafood consumption during pregnancy and SGA birth in a cohort from the Mediterranean coast of Spain. As this population has high seafood intakes, we are able to examine associations with different types of fish and shellfish, as well as overall seafood intakes, including levels exceeding the recommended threshold of three times per week.9 10 Using measures in maternal serum, we also examine whether associations between seafood intakes and fetal growth are confounded by levels of several pollutants, including PCBs, DDE and other compounds previously studied in relation to fetal growth, with mixed results.7 8 21–28
Methods
Population and study design
Data come from the Sabadell (Catalonia, Spain) cohort of the Childhood and Environment study (INfancia y Medio Ambiente: ‘INMA’ by its Spanish acronym), a population-based longitudinal study with enrolment during the first trimester of pregnancy. Details of the study design have been published previously.29 Briefly, between July 2004 and July 2006, 657 women were recruited from the main public health centre during the first ultrasound visit (mean±SD 13.5±1.7 weeks gestation); public health services are predominately used in this population.30 Questionnaires regarding diet, health behaviours, sociodemographic characteristics, reproductive history and other factors were administered by trained interviewers at recruitment, during the third trimester and at delivery. Maternal blood was collected at recruitment, and cord blood at delivery. Birth outcomes were available for 616 women (94%), including 596 term births (gestational age ≥37 weeks); maternal characteristics and dietary data were available for n=610 and 592 of these women, respectively. This analysis focused on term births, as there were too few (n=20) preterm births to assess separately, and seafood or fish oil intakes may be related to length of gestation.15 31 32 The IMIM-IMAS institutional review board provided ethical approval, and informed consent was obtained from all subjects.
Dietary assessment
Dietary intakes used in this analysis were assessed at recruitment using a 101-item semiquantitative food-frequency questionnaire previously validated against four dietary records in Spanish adults, with correlations similar to other studies (eg, r=0.56 for polyunsaturated fats).33 Correlations between fasting plasma docosahexanoic acid (subset of n=45) and reported fish consumption were similar to other studies, supporting the validity of intake estimates (r=0.34 for fresh fatty fish).34–37 Women were asked to report usual intakes since the start of pregnancy using reference portion sizes and nine frequency categories ranging from never/less than once month to 6+ times per day. Higher intake frequencies may represent larger habitual portion sizes relative to reference amounts as well as more frequent consumption. Separate questions about supplements, including consumption of fortified foods, were used to identify women using fish oil supplements.
The food frequency questionnaire included 10 seafood items classified a priori into subtypes as follows: (i) fatty fish species (two fresh fish items, ‘baked or steamed smaller fatty fish such as mackerel, sardines, anchovies,’ and ‘baked or steamed larger fatty fish such as albacore, swordfish;’ 1 canned item—‘tinned sardines/mackerel’); (ii) other fish (two items including ‘varied fried fish’; ‘baked or steamed lean fish such as hake, sole, or bream’); (iii) ‘canned tuna or albacore’ (one item); (iv) crustaceans (one item: ‘shrimp, prawns, lobster or crab’), (iv) other shellfish (two items including bivalves (‘clams, mussels, oysters’) and mollusks (‘squid, octopus, cuttlefish’)); and (v) smoked or salted fish (one item with very low intakes (‘salted or smoked fish: anchovies, cod, salmon’)—mean 1.5 g/day—which was therefore not analysed separately). All 10 items were summed to estimate overall seafood intakes. Intakes were similar to other Spanish studies.38 39
Reported intakes of each item were converted to estimated weekly frequencies using the midpoint of each category (eg, 1–3 times/month: 2/30 or 0.067/day, or 0.469/week), and gram amounts were estimated using the reference portions. Particularly for shellfish, typical reference portion sizes for seafood vary across food cultures (eg, one serving shrimp/other crustaceans=100 g in this study or approximately n=18 large pieces,40 vs 89 g41 and 200 g20 assumed elsewhere). Therefore, to facilitate interpretation of overall seafood intakes relative to other studies, associations with amounts are presented in addition to frequencies. Intake frequencies for each seafood subtype were categorised as high versus low based approximately on the median intake: fish intakes as ≤3 versus >3 servings/week, and shellfish and other fish subtypes as ≤1 versus >1 serving/week. Overall seafood intakes were analysed in frequency categories of ≤3, 3–6 and >6 servings/week, as well as gram amounts categorised in tertiles.
In addition to seafood, intakes of other foods (fruits, vegetables, meats and eggs, dairy products, alcoholic beverages, coffee) were estimated to assess potential confounding by these foods. Energy and nutrient intakes were also estimated using a database compiled based on US, UK and Spanish nutrient tables.40 42 43 Energy intake under-reporters were identified by calculating total estimated energy requirements using published equations44 45; women with intakes below 2 SDs of requirements were defined as under-reporters (19.9%).
Birth outcomes
Birth weight and length were recorded by specially trained midwives at delivery. Length of gestation was estimated from the date of the last menstrual period recorded at recruitment, and confirmed using first trimester ultrasound measures.46 When measures differed by >7 days (16.5%), ultrasound estimates were used. Infants below the 10th percentile of a Spanish reference population were classified as SGA.47 Additionally, to better distinguish intrauterine growth restriction from constitutionally small infants (ie, with parents of small stature), we used published methods to develop an adjusted SGA variable (SGAadj) which excluded infants classified as SGA but whose birth weight exceeded the 10th percentile of predicted values.48–50 The goal was to develop an SGA variable more likely to include only infants with intrauterine growth retardation, excluding infants whose smaller size may be largely a consequence of lower growth potential. Fractional polynomials were used to develop the best-fitting model to predict birth weight using child sex, gestational age, maternal age, prepregnancy weight, height and parity.48 N=8 infants with actual birth weights above the 10th percentile of predicted values were excluded from the SGA category using this approach.
Other variables
Maternal age, education, occupation, prepregnancy weight and reproductive history, as well as maternal reports of paternal weight, height and occupation, were collected in the first trimester; information on tobacco use throughout pregnancy was collected in the third trimester. Lipids (cholesterol and triglycerides) and several organochlorine pollutants were measured in maternal serum by the Reference Laboratory of the Basque Country (n=97% of the analysis sample), including: PCBs 118, 138, 153, and 180; hexachlorobenzene (HCB), beta-hexachlorohexane, DDT (2,2-bis(p-chlorophenyl)-1,1,1-trichloroethane), and DDE (2,2-bis(p-chlorophenyl)-1,1-dichloro-ethylene).51 Mercury was measured in cord blood (79% of the sample) by atomic absorption spectrometry with the amalgamation technique, using the advanced mercury analyser (AMA-254, Altec). Internal quality controls were used to ensure validity.
Statistical analysis
In descriptive analyses, size at birth (SGA, SGAadj and birth weight), parental characteristics and pollutant levels were described by levels of maternal seafood intake, using χ2, ANOVA or Kruskal–Wallis tests to evaluate statistical significance. For contaminants, we calculated geometric means using models for censored data (due to the presence of subjects with levels below detection limits) and assessed significance across levels of intake using likelihood ratio tests. Both overall intakes and seafood subtypes were explored. In multivariable models, logistic regression was used to assess the association between seafood measures and the likelihood of SGA birth; linear regression was used to assess associations with birth weight. Results are reported as ORs or coefficients (coef.) with 95% CIs. Covariates in the final models were child sex, nulliparity, paternal BMI (weight in kg/height in m2), and maternal age, BMI, education level (≥secondary vs ≤primary), energy intakes and dietary intake under-reporting; birth weight models included gestational age, and maternal height and weight rather than BMI, as these variables were more strongly associated with this outcome. Models for seafood subtypes adjusted for other types of seafood. To assess confounding by other aspects of the diet, we assessed the effect of adjusting for maternal intakes of meats and eggs, vegetables, fruits, legumes, dairy products, dietary fat, alcohol and coffee; as associations were virtually unchanged, these variables were omitted. Other potential covariates (dietary supplement use; marital status; parental occupational class based on International Standard Classification of Occupations coding; and tobacco use) were excluded for parsimony, as they did not meet criteria for confounding (change-in-estimate ≥10%).
In separate models, we examined how seafood intake–birth size associations were affected by adjusting for organochlorine pollutants and mercury. Variables were log-transformed, and the linearity of associations was examined using tertiles. Models adjusted for total lipids as prescribed by Phillips et al.52 In supplementary models, we confirmed that results were similar after excluding women using fish oil supplements (n=20), large for gestational age infants (LGA: birth weight >90th percentile, n=53), or excluding versus adjusting for under-reporters (not shown). All analyses used STATA version 10.1.
Results
Sample characteristics
Overall, 7.8% of births (n=46) were classified as SGA, with 6.5% (n=38) classified as SGAadj (ie, excluding children whose actual birth weights were ≥10th percentile of predicted values) (table 1). As shown, mean intakes of each type of fish and shellfish increased with rising levels of overall seafood intake. However, intakes of different types of seafood were only moderately correlated, with Spearman correlations ranging from 0.13 to 0.36. Maternal seafood consumption was not strongly related to sociodemographic or other parental characteristics including BMI.
Sample characteristics by overall seafood intake category: mean (SD) or proportion
Seafood intakes and size at birth
There was a modest increase in SGA prevalence among women reporting >6 versus ≤3 servings of seafood per week (table 1). When specific subtypes of seafood were examined (table 2), higher intakes of crustaceans (χ2 p<0.05), canned tuna (χ2 p<0.05) and, to a lesser extent, fatty fish (χ2 p<0.10) were associated with increased SGA, but there was no association with other types of seafood (χ2 p>0.10). Similarly, mean birth weight was lower among women reporting higher overall seafood intakes (table 1), with the most marked reduction (−115 g, p=0.007) associated with higher intakes of crustaceans (table 2). Neither fatty fish nor canned tuna was associated with lower mean birth weight.
Prevalence of SGA and mean (SD) birth weight by level of maternal seafood consumption
In multivariable models (table 3), there was a dose–response relationship between overall seafood consumption and SGA, with a more than twofold increase in odds of SGA associated with intakes >6 servings/week. Associations were stronger with SGAadj than with the unadjusted SGA variable, but neither reached statistical significance. Using intakes categorised as amounts rather than frequencies, there was a significant increase in SGAadj associated with the highest tertile of intake (≥100 g), albeit not with the intermediate (64–99.9 g) tertile (adjusted ORs for the 2nd and 3rd vs lowest tertile=2.30, 0.87–6.07 and 2.92, 1.12–7.65, respectively). As in the descriptive analysis, when seafood subtypes were examined, there was a significant increase in SGA associated with higher intakes of crustaceans and canned tuna after adjustment. Associations with fatty fish were weaker, and were strongly attenuated and non-significant using the adjusted SGA variable. Additionally adjusting for organochlorine pollutants did not meaningfully change associations (table 3). Similarly, adjusting for cord blood mercury had virtually no effect (not shown).
Multivariable associations between seafood consumption in times per week and SGA
After multivariable adjustment, the highest level of overall seafood intakes was associated with a non-significant reduction in mean birth weight; this association was further attenuated after adjusting for pollutants.
Discussion
In this population, higher maternal intakes of crustaceans and canned tuna during pregnancy were associated with a significant increase in risk of SGA after multivariable adjustments including biomarkers of numerous contaminants, with no increase in risk associated with consumption of lean fish or other types of shellfish. Relationships with fatty fish consumption were less clear, with largely non-significant increases in SGA associated with this type of seafood. In addition to particular seafood subtypes, because of the high levels of intake, we were able to examine associations between size at birth and overall seafood intakes at levels exceeding current recommendations for pregnant women in the UK and USA. Increases in SGA were observed only at intakes well above the recommended threshold of three servings per week. However, for the specific seafood subtypes for which associations were observed, relatively low intakes—at least once per week—were associated with SGA.
Interestingly, these results are to a large extent consistent with two recent studies which also examined fetal growth in relation to selected seafood subtypes.4 20 Like us, one study reported an increase in SGA associated with crustacean but not fish intakes.20 Although we did not clearly confirm an increase in risk associated with consumption of fatty but not lean fish,4 it is important to note that this analysis reflects a higher range of intakes than earlier studies, where mean seafood intakes were in the order of 20–40 g/day, with infrequent (intakes <1/month) or non-consumers used as the referent.4 19 20 Mean intakes in this sample were twice as high, with only 1% reporting no consumption.
While several earlier studies have reported inverse associations between seafood consumption during pregnancy and size at birth, reasons for this negative association—if causal—are unclear.4 19 20 The associations observed with crustaceans and canned tuna—for which consumption in Spain is high53 54—do not necessarily suggest involvement of marine fatty acids, since levels in these foods are relatively low compared with fatty fish.40 42 It has been suggested that adverse effects on fetal growth might be attributable in part to organic pollutants such as PCBs, for which relatively high levels have been reported for some types of seafood, including tuna, in Spain.21 55 56 Mechanisms linking such pollutants to fetal growth are not established27 but may involve effects on endocrine,23 28 thyroid57 or immune function.58 However, associations with SGA remained similar after adjusting for serum levels of several pollutants, suggesting that associations between may be more complex than a simple increase in the burden of these compounds. Not surprisingly, as it is not clearly established to be a risk factor for small size at birth, confounding by mercury also did not explain the associations observed, although seafood is an important source of this contaminant.59 Further study is needed to elucidate whether these relationships observed are causal and to identify the mechanisms involved, which might be related to other contaminants or contaminant mixtures.26 58
Recent literature has emphasised the need to identify intrauterine growth retardation based on factors other than population-based norms for size at birth, using characteristics such as maternal size to develop measures, which may help distinguish small but normal children from those with growth retardation.49 50 Similar to our findings for overall seafood and crustacean intakes, several studies have reported stronger associations with various exposures after adopting individualised measures of SGA which attempt to reduce misclassification of IUGR.50 60 61 In contrast, the association with canned tuna for SGAadj was slightly attenuated compared with the standard SGA measure, suggesting that this association may be less robust.
We were unable to examine associations with preterm birth or low birth weight because of the low prevalence of these outcomes and modest sample size. In addition, as this is an observational study, we cannot rule out possible confounding by unmeasured factors. However, the similarity of descriptive and multivariable-adjusted results, as well as the consistency with previous studies, suggests that associations may not be attributable simply to residual confounding. In contrast to settings where previous studies were conducted, consumption of seafood as a whole, as well as canned tuna and other fish, was not related to socio-economic status. Shellfish intakes were weakly but positively related to socio-economic measures, suggesting that any residual socio-economic confounding might lead to under- rather than overestimated associations with SGA. While we cannot rule out the possibility that crustaceans or canned tuna are markers of diets or lifestyle patterns related to SGA, associations did not change after adjusting for numerous dietary factors.
If linked to contaminant accumulation, seafood consumption before, as well as during, pregnancy may play a role in this relationship. Using women's reports on changes in seafood consumption since the start of pregnancy, we restricted the analysis to those reporting no differences in their intake (58% of the sample): there was no meaningful change in findings for crustaceans or most other types of seafood. However, associations with canned tuna were further attenuated and non-significant (not shown). Moreover, consistent with stronger effects of longer-term consumption, associations with crustaceans were strengthened after excluding women who reported having increased their seafood intakes since the start of pregnancy (26% of the sample; intakes during pregnancy may over-represent long-term intakes in these women), and slightly attenuated excluding those reporting having reduced consumption (18%) (not shown).
In conclusion, this study suggests that higher maternal intakes of certain types of seafood, in particular crustaceans and canned tuna, may be associated with increased risk of SGA. Further study is needed to confirm these findings and clarify the mechanisms that may be involved, including analyses of additional contaminants, other congeners or metabolites of the compounds studied, and contaminant mixtures. Future analysis relating longer-term intakes of specific types of seafood vs diet in pregnancy alone may also be useful to elucidate the mechanisms involved.
What is already known on this subject
Recent studies suggest that inconsistencies in studies on maternal seafood consumption during pregnancy and fetal growth may be partly attributable to heterogeneous effects of different types of seafood consumed. Previous studies have been unable to simultaneously explore associations with both different subtypes of fish and shellfish, or to assess to what extent persistent organic pollutants (POPs) appear to explain any adverse associations.
What this study adds
In this cohort, maternal consumption of crustaceans (at least once per week) and canned tuna (at least once per week), but not fatty fish, lean fish or other shellfish, was consistently associated with significant increases in risk of small size for gestational age (SGA). Adjusting for several POPs in maternal serum had little effect on these associations, and more research is needed to elucidate the mechanisms involved in these associations.
Acknowledgments
The authors would like to acknowledge all the study participants for their generous collaboration. We are also indebted to S Folchs, A Sànchez, M López and N Pey for their assistance in contacting the families and administering the questionnaires. A full roster of the INMA-Sabadell Study Investigators can be found at http://proyectoinmania.org/en_index.html
References
Footnotes
Linked articles 093328.
Funding Spanish Ministry of Health (FIS-FIS-PI041436), Instituto de Salud Carlos III (Red INMA G03/176 and CB06/02/0041), the Generalitat de Catalunya-CIRIT 1999SGR 00241, and the European Union sixth framework project EARNEST FOOD-CT-2005-007036.
Competing interests None.
Ethics approval Ethics approval was provided by the IMIM-IMAS (Instituto Municipal de Investigación Medica-Instituto Municipal de Asistencia Sanitaria) ethics review committee.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.