Eating fish has been associated with reductions in preterm birth, increased duration of pregnancy, or both reductions in preterm birth and increased duration of pregnancy in some1–4 but not all studies5–8; most studies have been conducted in unselected or low-risk women. Several reports have found the association between fish consumption and pregnancy outcome to be nonlinear. In one study, mean birth weight increased up to three fish dinners per week, but leveled off or decreased thereafter.9 In another study, gestational age was shortened and preterm birth was increased2 only among women who consumed no fish. However, among women who ate fish, frequency of consumption was not associated with these outcomes, suggesting a threshold for fish intake. One study reported that, among women with high marine omega-3 fatty acid consumption, the ratio of omega-3 fatty acids to arachidonic acid in erythrocytes was not associated with shortened gestation, whereas among women with lower fish consumption, gestation was prolonged with increases in this ratio. These results suggest that any benefit of increasing omega-3 long-chain polyunsaturated fatty acid levels may occur only in gravidas with chronically low intake and have no effect in those with chronically high intake.10
We recently conducted a randomized, double-blind placebo-controlled clinical trial of omega-3 supplementation, beginning at 16–21 weeks of gestation, for the prevention of recurrent preterm birth in a group of high-risk women.11 We were interested in estimating the association between omega-3 fatty acids early in pregnancy before study enrollment and preterm birth, and whether this exposure modified the response to supplementation. Omega-3 exposure was estimated by dietary history of fish intake (including tuna and shellfish) at study enrollment and by erythrocyte omega fatty acid levels collected at study enrollment.
MATERIALS AND METHODS
The data for this report are from the Eunice Kennedy Shriver National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network randomized clinical trial of omega-3 long-chain polyunsaturated fatty acid supplementation to prevent recurrent preterm birth.11 The trial recruited (at 13 Network centers from January 2005 to October 2006) women who had a history of at least one previous spontaneous singleton preterm birth and randomized 434 women to receive daily supplementation of 1,200 mg of eicosapentaenoic acid (20:5n-3) and 800 mg of docosahexaenoic acid (22:6n-3), and randomized 418 to matching placebos, beginning at 16–21 6/7 weeks of gestation and continuing until 36 6/7 weeks of gestation or delivery, whichever occurred first. As part of the trial, all enrolled women also received weekly injections of 17 alpha-hydroxyprogesterone caproate. Women currently taking fish oil or omega-3 supplements were ineligible for the trial; detailed inclusion and exclusion criteria are reported elsewhere.11 Delivery before 37 completed weeks of gestation occurred in 37.8% and 41.6% of women randomized to omega-3 supplementation and placebo, respectively.11 Omega-3 supplementation was associated with a significant change in plasma levels of docosahexaenoic acid, eicosapentaenoic acid, and arachidonic acid from baseline to 25–28 weeks of gestation compared with placebo.11 The study was approved by the Institutional Review Boards of the biostatistical coordinating center and all participating clinical centers, and this secondary analysis was determined to be exempt from Institutional Review Board review by the Office of Human Subjects Research, National Institutes of Health; all enrolled women gave written informed consent.
At randomization, women underwent an interview, including a four-item food frequency questionnaire designed to assess fish intake during the time from last menstrual period to randomization. The four items were dark-meat fish, canned tuna, other fish, and shellfish12; response options for frequency of consumption were never or less than one serving per month, one to three servings per month, one serving per week, two to four servings per week, five to six servings per week, one serving per day, two to three servings per day, four to five servings per day, and six or more servings per day. In this analysis, intake of each type of fish was converted to servings per week, and all four were summed and considered as fish intake. When the response option was a range, the midpoint was used; one to three servings per month was assumed to represent one-half serving per week; never or less than one serving per month was assumed to represent zero servings per week. Owing to small numbers, the four individual types of fish were not considered separately. Women received no specific dietary advice as part of this study, although taking nonstudy omega-3 supplements or prenatal vitamins containing omega-3 long-chain polyunsaturated fatty acids was prohibited during the trial.
Blood was collected at randomization, before dispensing study drug. Erythrocytes were separated from plasma, snap frozen, and shipped to a central laboratory for fatty acid analysis by gas chromatography. Individual polyunsaturated fatty acids were assayed using previously described methods and expressed as percent of total fatty acids.13 Erythrocyte measures studied were the sum of omega-3 fatty acids, docosahexaenoic acid, and eicosapentaenoic acid; the sum of omega-6 polyunsaturated fatty acids, linoleic acid, and arachidonic acid; and the ratio of these two sums.
Gestational age at birth was determined from the ultrasonographically confirmed gestational age at randomization and the elapsed time from randomization to delivery11; preterm birth was defined as birth at less than 37 completed weeks (259 days) of gestation. The associations between preterm birth and fish intake and erythrocyte polyunsaturated fatty acids were evaluated graphically by locally weighted scatterplot smoothing plots.14 Locally weighted scatterplot smoothing is a form of data smoothing and plotting that performs locally weighted regression of preterm birth on fish intake at every point of fish intake and combines the regressions to form a curve. It is thus a “nonparametric” regression that does not constrain the association between fish intake and preterm birth to take any prespecified mathematical relationship. We used this technique to evaluate whether the association between preterm birth and fish intake appeared linear, and if not then how best to model the association. Continuous variables were compared using the Wilcoxon or Kruskal-Wallis test and categorical variables were compared using the χ2 test. When categories were ordered, significance was assessed with the Cochran-Armitage test for trend.15 We tested whether the effect of omega-3 supplementation on preterm birth differed according to baseline erythrocyte fatty acid concentration using the Breslow–Day test.16 Multiple logistic regression was used to adjust the association between fish intake and preterm birth for confounders selected a priori: study center, number of previous preterm births, gestation of earliest previous spontaneous preterm birth, receipt of omega-3 compared with placebo supplement, smoking, age, education, body mass index, race, and ethnicity. When we considered linear and quadratic terms in evaluation of dose-responses, we used the likelihood ratio test with 1 and 2 degrees of freedom, respectively, to determine statistical significance. Tests for interaction between fish consumption as both a linear and quadratic term and study treatment assignment utilized the likelihood ratio test with 2 degrees of freedom. Statistical significance was defined as a two-tailed P<.05, with no adjustment for multiple comparisons. Data were analyzed using SAS (version 8.2) and R (version 2.9.0).
Gestational age at delivery and baseline (16–21 completed weeks) fish consumption were available for all 852 randomized women. There were 253 (29.7%) women who reported consuming fish never or less than once per month; 524 (61.5%) who consumed one-half to three servings per week, and 75 (8.8%) who consumed fish more than three times per week. The association between fish consumption and characteristics of the study population is presented in Table 1. Because the association between fish consumption and the study outcomes appeared U-shaped (see below), fish consumption is presented in Table 1 as none or less than one per month, one-half to three servings per week, and more than three servings per week. However, in subsequent modeling, fish consumption was considered as a continuous variable. African-American and Hispanic women ate fish more frequently than non–African American and non-Hispanic women, respectively. Fish consumption did not differ significantly by any of the other characteristics in Table 1.
Preterm birth occurred for 48.6% of the 253 women who ate fish once per month or less as reported at 16–21 weeks of gestation, compared with 35.9% of the 599 women who ate fish more often (P<.001). Locally weighted scatterplot smoothing plots of the association between servings of fish per week and preterm birth (Fig. 1) demonstrated that the probability of preterm birth declined with increasing fish consumption and then increased again, although few women ate fish more than several times per week and thus the confidence bands around the association were wide for women eating fish very frequently. The shape of the association was similar among women assigned to omega-3 supplements and to placebo (data not shown). Because of this U shape, fish intake was modeled as both a linear and a quadratic term. In unadjusted analysis, both linear and quadratic terms for fish intake were statistically significant (P<.001 and P=.001, respectively), and a model with a term for (number of fish servings)2 fit the data significantly better than one containing only a linear term for number of fish servings. In addition the P value for the two terms taken together was also significant (P=.002).
In a model adjusting for study center, number of previous preterm births, gestation of earliest previous spontaneous preterm birth, receipt of omega-3 compared with placebo supplement, smoking, age, education, body mass index, race, and ethnicity, the combination of linear and quadratic terms for fish consumption in the first half of pregnancy remained statistically significant (P=.02). Both the linear (P=.01) and quadratic terms (P=.008) were individually statistically significant and were therefore retained in calculation of odds ratios. Furthermore, the association between fish consumption and preterm birth was similar among women receiving omega-3 supplementation or placebo (P value for the interaction between treatment and fish consumption was .95).
The modeled odds ratios and their 95% confidence intervals for the association between total fish consumption and preterm birth are presented in Table 2. Note that the odds ratios in Table 2 are derived from modeling the entire dose-response curve for fish consumption and preterm birth rather than directly from the individual reported amounts. Increasing fish consumption was associated with a decreasing odds ratio for preterm birth, reaching a predicted minimum of 0.6 at approximately three servings of fish per week. The modeled odds ratio rose as fish consumption increased beyond this point, reaching the value observed in nonconsumers at approximately seven servings per week, although the confidence limits were wide.
Red cell fatty acid values at baseline were available for 701 (82.3%) of the 852 women. The mean red blood cell omega-3 fatty acids as a percentage of all fatty acids was 3.55% for women randomized to omega-3 supplements and 3.73% for women randomized to placebo (P=.59). A locally weighted scatterplot smoothing plot of the association between preterm birth and erythrocyte docosahexaenoic acid plus eicosapentaenoic acid was generated (not shown). Owing to the irregularity of the curve, erythrocyte fatty acids were analyzed as quartiles. Table 3 shows the association between quartiles of erythrocyte docosahexaenoic acid plus eicosapentaenoic acid and preterm birth. Women in the lowest quartile had an increased risk of preterm birth (47.2%) compared with women in the three highest quartiles combined (38.1%, P=.03), but there was no consistent trend among women in the higher quartiles; the overall association (by χ2 test) among all four quartiles of erythrocyte docosahexaenoic acid plus eicosapentaenoic acid and preterm birth was of borderline statistical significance (P=.054). After adjusting for study center, number of previous preterm births, gestation of earliest previous spontaneous preterm birth, receipt of omega-3 compared with placebo supplement, smoking, age, education, body mass index, race, and ethnicity, the odds ratio for preterm birth among women in the lowest quartile compared with women in the three highest quartiles combined was 1.41 (0.97–2.05). When the top three quartiles were compared individually with the lowest quartile, the adjusted odds ratio for women in quartile 2 indicated a statistically significant reduction, but quartiles 3 and 4 were not significantly different from quartile 1 (Table 3).
Compared with placebo, the effect of omega-3 supplementation was similar among women in the lowest quartile of erythrocyte omega-3 (45.2% compared with 48.9%, respectively, P=.63) and among women in the top three quartiles of erythrocyte omega-3 (35.7% compared with 40.6%, respectively, P=.24). The P value for interaction between treatment assignment and erythrocyte omega-3 was .86 (Breslow–Day test for homogeneity of the odds ratios).
The odds ratios for quartiles of erythrocyte omega-6 fatty acids and preterm birth were not statistically significant. Results for quartiles of the omega-3:omega-6 ratio were similar to those for omega-3, but of slightly smaller magnitude. In no case did the interaction between treatment assignment and any of the measures of erythrocyte fatty acids (omega-3, omega-6, or the ratio of omega-3:omega-6) approach statistical significance (lowest P value for interaction was .27).
Erythrocyte docosahexaenoic acid plus eicosapentaenoic acid levels correlated weakly but significantly with reported frequency of fish intake (Spearman r=0.22, P<.001). Women in the lowest quartile of docosahexaenoic acid plus eicosapentaenoic acid were more likely to report consuming fish never or less than once per month (40.3%) than were women in the highest three quartiles (26.3%, P<.001). When actual omega-3 intake was estimated according to the formula of Hu et al,12 the correlation between omega-3 intake and erythrocyte docosahexaenoic acid plus eicosapentaenoic acid was essentially identical (Spearman r=0.22, P<.001).
Our findings may be summarized as follows: 1) In our high-risk population of women with a previous preterm birth, those who reported the lowest fish consumption at 16–21 weeks of gestation were at elevated risk of recurrent preterm birth compared with comparable women who ate fish more frequently. 2) However, the “dose-response” of fish consumption for preterm birth was statistically significantly nonlinear. The lowest occurrence of preterm birth was seen among women who ate fish approximately two to three times per week. More frequent fish consumption was not associated with further reductions in preterm birth, and our data indicate that preterm birth might become more common with further increases in fish consumption, although our estimates were imprecise. Similarly, the lowest occurrence of preterm birth was observed among women in the second quartile of erythrocyte omega-3 levels. 3) The lack of benefit of omega-3 polyunsaturated fatty acid supplementation was similar regardless of either baseline fish consumption or erythrocyte omega-3 concentration.
The association between low fish consumption and preterm birth was of similar magnitude among women assigned to omega-3 supplementation (47.6% compared with 33.9%, P=.008) or placebo (49.6% compared with 38.1%, P=.03); the Breslow–Day P value for homogeneity was .74.11 Why should moderate fish intake and higher erythrocyte omega-3 concentration in the first half of pregnancy be associated with a reduced recurrence of preterm birth while omega-3 supplementation did not reduce the recurrence? One reason for our discrepant findings might be the timing of supplementation. Our erythrocyte omega-3 measures were obtained at 16–21 6/7 weeks of gestation and reflect omega-3 status over 2–3 months17; the fish questionnaire asked about consumption from conception to randomization. Thus, both measures reflect exposure around conception and during early gestation. In contrast, supplementation began from 16 to 21 6/7 weeks of gestation. Even though supplementation raised plasma omega-3 concentrations,11 it is possible that low early-pregnancy levels are associated with preterm birth and supplementation occurred too late to have an effect. Olsen and Secher2 also noted that seafood consumption assessed at 16 weeks of gestation but not at 30 weeks7 was associated with preterm birth, providing support for this explanation.
A second possible explanation for the result is that another nutrient contained in fish, thus sharing a dietary source with docosahexaenoic acid and eicosapentaenoic acid, is responsible for the reduction in preterm birth. Several types of fish are rich dietary sources of vitamin D,18 deficiency of which has been associated with preterm birth19 and preeclampsia.20 A third explanation is that supplementation does not alter dietary habits and individuals who do not eat seafood may have higher dietary intake of omega-6 fatty acids,12 the precursors of the uterotonic prostaglandins, leukotrienes, and thromboxanes involved in cervical ripening and myometrial contraction.21 A fourth explanation is that there may be unmeasured characteristics of women who do not consume seafood (or consume very little dietary docosahexaenoic acid and eicosapentaenoic acid) that are the true cause of the increase in preterm birth. We found few differences between women who ate fish rarely or never, occasionally, or frequently, and Oken et al reported similar results in a cohort of pregnant women in Boston.22 Nevertheless, our measure of socioeconomic status was limited to education and we did not collect data on other components of diet, behavioral factors such as physical activity, psychosocial characteristics, or genital tract infection that might be associated with fish consumption and preterm birth.
We also observed that preterm birth may increase at high levels of fish consumption and failed to decrease further beyond the second quartile of erythrocyte omega-3 long-chain polyunsaturated fatty acids. Our results of a nonlinear association between fish intake and preterm birth are similar to those of Olsen et al, who reported that gestational age was shorter and preterm birth increased2,9 only among women who consumed little or no fish, but among women who ate fish, the actual frequency of consumption was not strongly associated with these outcomes. Olsen et al also reported that mean birth weight increased with increasing fish consumption up to three meals per week, but leveled off or decreased with more frequent consumption.9 In addition to being a source of nutrients, many types of fish contain environmental contaminants such as methylmercury23 and polychlorinated biphenyls,24 which have been associated in some studies with shortened gestation.25 Fish is a rich source of protein, and high-protein supplementation was associated with increased occurrence of preterm birth in a randomized trial conducted among women in Harlem.26
Our study has limitations. In addition to having few measures of socioeconomic factors that might be associated with fish consumption, we collected no data on diet other than frequency of fish consumption up to the time of randomization, nor did we assess biomarkers of environmental contaminants or nutrients other than omega-3 and omega-6 fatty acids. Thus we cannot evaluate what other components or correlates of fish intake might account for the associations we observed with preterm birth. Fish consumption is likely to be reported with error, although because the report was obtained before the outcome of the pregnancy was known, the error is likely to be unbiased with respect to preterm birth. As part of the trial protocol, all women received weekly injections of 17 alpha-hydroxyprogesterone caproate and all women had a previous spontaneous preterm birth. Therefore, our population was at unusually high risk of having preterm birth, and our results may not apply to other populations of pregnant women.
However, from a clinical perspective our data suggest that docosahexaenoic acid and eicosapentaenoic acid supplementation cannot take the place of fish consumption with respect to any benefits there might be for preterm birth. Our results also support the recommendations of both the American Congress of Obstetricians and Gynecologists27 and the Food and Drug Administration28 for consumption of up to two servings of fish per week.
1.Olsen SF, Østerdal ML, Salvig JD, Kesmodel U, Henriksen TB, Hedegaard M, et al. Duration of pregnancy in relation to seafood intake during early and mid pregnancy: prospective cohort. Eur J Epidemol 2006;21:749–58.
2.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.
3.Guldner L, Monfort C, Rouget F, Garlantezec R, Cordier S. Maternal fish and shellfish intake and pregnancy outcomes: a prospective cohort study in Brittany, France. Environ Health 2007;6:33.
4.Grandjean P, Bjerve KS, Weihe P, Steuerwald U. Birthweight in a fishing community: significance of essential fatty acids and marine food contaminants. Int J Epidemiol 2001;30:1272–8.
5.Oken E, Kleinman KP, Olsen SF, Rich-Edwards JW, Gillman MW. 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.
6.Rogers I, Emmett P, Ness A, Golding J. Maternal fish intake in late pregnancy and the frequency of low birth weight and intrauterine growth retardation in a cohort of British infants. J Epidemiol Community Health 2004;58:486–92.
7.Olsen SF, Hansen HS, Secher NJ, Jensen B, Sandström B. Gestation length and birth weight in relation to intake of marine n-3 fatty acids. Br J Nutr 1995;73:397–404.
8.Thorsdottir I, Birgisdottir BE, Halldorsdottir S, Geirsson RT. 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.
9.Olsen SF, Grandjean P, Weihe P, Viderø T. 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.
10.Olsen S, Hansen HS, Sommer S, Jensen B, Sørensen TI, Secher NJ, et al. Gestational age in relation to marine n-3 fatty acids in maternal erythrocytes: a study of women in the Faroe Islands and Denmark. Am J Obstet Gynecol 1991;164(5 pt 1):1203–9.
11.Harper M, Thom E, Klebanoff MA, Thorp J Jr, Sorokin Y, Varner MW, et al. Omega-3 fatty acid supplementation to prevent recurrent preterm birth: a randomized controlled trial. Obstet Gynecol 2010;115(2 pt 1):234–42.
12.Hu FB, Bronner L, Willett WC, Stampfer MJ, Rexrode KM, Albert CM, et al. Fish and omega-3 fatty acid intake and risk of coronary heart disease in women. JAMA 2002;287:1815–21.
13.Reece MS, McGregor JA, Allen KG, Harris MA. Maternal and perinatal long-chain fatty acids: possible roles in preterm birth. Am J Obstet Gynecol 1997;176:907–14.
14.Cleveland WS, Grosse E. Computational methods for local regression. Stat Comput 1991;1:47–62.
15.Armitage P. Statistical methods in medical research. London: Blackwell Scientific Publications; 1971. p. 363–5.
16.Breslow NE, Day NE, editors. Statistical methods in cancer research. Lyon, France: International Agency for Research on Cancer; 1980. The analysis of case-control studies, Vol. 1. p. 142.
17.Farquhar JW, Ahrens EH Jr. Effects of dietary fats on human erythrocyte fatty acid patterns. J Clin Invest 1963;42:675–85.
18.Hyppönen E, Power C. Hypovitaminosis D in British adults at age 45 y: nationwide cohort study of dietary and lifestyle predictors. Am J Clin Nutr 2007;85:860–8.
19.Scholl TO, Chen X. Vitamin D intake during pregnancy: association with maternal characteristics and infant birth weight. Early Hum Dev 2009;85:231–4.
20.Bodnar LM, Catov JM, Simhan HN, Holick MF, Powers RW, Roberts JM. Maternal vitamin D deficiency increases the risk of preeclampsia. J Clin Endocrinol Metab 2007;92:3517–22.
21.Hagve TA, Christophersen BO. Linolenic acid desaturation and chain elongation and rapid turnover of phospholipids n-3 fatty acids in isolated rat liver cells. Biochem Biophys Acta 1983;753:339–49.
22.Oken E, Radesky JS, Wright RO, Bellinger DC, Amarasiriwardena CJ, Kleinman KP, et al. Maternal fish intake during pregnancy, blood mercury levels, and child cognition at age 3 years in a US cohort. Am J Epidemiol 2008;167:1171–81.
23.Smith KM, Barraj LM, Kantor M, Sahyoun NR. Relationship between fish intake, n-3 fatty acids, mercury and risk markers of CHD (National Health and Nutrition Examination Survey 1999-2002) Public Health Nutr 2009;12:1261–9.
24.Stahl LL, Snyder BD, Olsen AR, Pitt JL. Contaminants in fish tissue from US lakes and reservoirs: a national probabilistic study. Environ Monit Assess 2009;150:3–19.
25.Xue F, Holzman C, Rahbar M, Trosko K, Fischer L. Maternal fish consumption, mercury levels, and risk of preterm delivery. Environ Health Perspect 2007;115:42–7.
26.Rush D, Stein Z, Susser M. A randomized controlled trial of prenatal nutritional supplementation in New York City. Pediatrics 1980;65:683–97.
27.American Congress of Obstetricians and Gynecologists. Nutrition during pregnancy. Patient Education Pamphlet AP001. Washington, DC: American Congress of Obstetricians and Gynecologists; 2010.