Maternal Fish Consumption, Mercury Levels, and Risk of Preterm Delivery

Fei Xue; Claudia Holzman; Mohammad Hossein Rahbar; Kay Trosko; Lawrence Fischer


Environ Health Perspect. 2007;115(1):42-47. 

In This Article


Women enrolled in the POUCH Study resided in communities surrounded by the Great Lakes, but their levels of fish consumption in the first 6 months of pregnancy would be considered moderate to low relative to populations that subsist on fish. Despite the modest levels of fish consumption, there was strong evidence that mercury levels in maternal hair were higher when fish consumption levels were higher.

A positive correlation between fish consumption and mercury levels in pregnant women has been reported in studies conducted in European countries (Bjornberg et al. 2003; Daniels et al. 2004; Oskarsson et al. 1994) and the Amazon basin of South America (Bruhn et al. 1995; Hacon et al. 2000). Fewer data are available from pregnant populations in North America. Two studies from Canada in areas close to the Great Lakes showed a positive association between fish consumption and mercury levels in maternal hair (Morrissette et al. 2004; Muckle et al. 2001). Other studies from the Great Lakes area have observed a positive relationship between fish consumption and mercury levels in nonpregnant women of child-bearing age (Nadon et al. 2002), anglers and fish eaters (Cole et al. 2004), Native Americans (Gerstenberger et al. 1997), Montreal sportfishers of Asian origin (Kosatsky et al. 1999b), and the general population (Kosatsky et al. 2000; Mahaffey and Mergler 1998). The concordance of the information provided from these and other studies indicates that in diverse regions of the globe, consumption of fish is a major source of methylmercury exposure in humans.

In one study, methylmercury was the most commonly identified pollutant in sport-caught fish (Anderson et al. 2004). Canned tuna has been implicated as one of the main foods contributing to total mercury intake (Legrand et al. 2005; U.S. Department of Health and Human Services and U.S. Environmental Protection Agency 2006). Because of variation in methylmercury levels by fish type and location, studies in different settings have tried, as we tried in this study, to link mercury levels to consumption of particular types of fish. In anglers and Native Americans residing in the Great Lakes region, levels of mercury have been shown to be positively associated with consumption of sport-caught fish (Cole et al. 2004; Gerstenberger et al. 1997; Kosatsky et al. 1999a; Nadon et al. 2002). A study by the Wisconsin Division of Public Health reported that women of child-bearing age who ate sport-caught fish had higher mercury levels in hair than those of women who did not eat sport-caught fish, but the difference was not statistically significant (Knobeloch et al. 2005). Morressette et al. (2004) reported that bought fish, including canned fish, were important sources of mercury exposure during pregnancy. These studies are consistent with our findings of higher hair mercury levels in pregnant women who consumed greater amounts of canned fish, sport-caught fish, and bought fish.

As anticipated, hair mercury levels in POUCH Study mothers were lower than levels found in heavily exposed, fish-eating communities. Across studies the unit of measure used to report hair mercury levels varies -- parts per million, milligrams per kilogram, or micrograms per gram -- but these units are equivalent and can be directly compared. The median hair mercury level in the Faroe Islands sample was 4.5 ppm, and in the Seychelles 5.9 ppm, almost double the highest mercury level observed in the POUCH Study (Davidson et al. 1999; Grandjean et al. 1992). In a study of Swedish women of child-bearing age, 127 women reported consuming, on average, four fish meals/week and had a median hair mercury level of 0.70 mg/kg (Bjornberg et al. 2005). A study of 150 pregnant women from varied locations in Alaska found somewhat lower levels: The median hair mercury was 0.47 mg/kg and the mean was 0.72 mg/kg (Arnold et al. 2005). Two studies -- one from Massachusetts, the other from southwest Quebec -- reported mercury levels more similar to those in our findings. The first study included 135 pregnant women recruited from an HMO in eastern Massachusetts (Oken et al. 2005). On average, women in that study consumed slightly more fish (1.2 meals/week) than did women in the POUCH Study (0.82 meals/week), and had slightly higher hair mercury levels (mean = 0.55 ppm; 10% exceeded 1.2 ppm) than did those in the POUCH Study (mean = 0.29 µg/g; 10 % exceeded 0.54 µg/g). In the Quebec study of 159 pregnant women, mercury was assessed in sequential centimeters of hair to represent levels during different months of pregnancy (Morrissette et al. 2004). Among women who consumed two or more fish meals/month, mean hair mercury levels were about 0.20 µg/g at the 5th-6th month of pregnancy. Differences in maternal mercury levels across these studies may be explained by differences in both amount and types of fish consumed.

In the POUCH Study, maternal characteristics of older age, ethnicity of white and other, not being insured by Medicaid, and residing in three of the five communities were significantly associated with higher maternal mercury levels after adjusting for fish consumption. The reasons for these associations are unclear and may include other lifestyle factors, mercury exposures not related to fish, and residual confounding from incomplete adjustment for fish types, portion size, and method of preparation. The observed higher mean mercury level among women without Medicaid insurance -- a marker of higher socioeconomic status -- is consistent with a recently released report by the Wisconsin Division of Public Health, which also found higher mercury levels in association with higher socioeconomic status (i.e., college education and annual household income > $75,000) (Knobeloch et al. 2005). It is unlikely that the community differences in mercury levels within our study reflect variations in pollutant levels in local fish, because most of the fish consumed are not from local waters. About 10% of women in the POUCH Study reported not consuming fish during pregnancy but had hair mercury levels in the top two quintiles. This suggests that their diet histories may have been inaccurate or there may be sources of exposure, other than fish, that are uncommon but deserve consideration.

Our study is the first to report an association between delivery at < 35 weeks' gestation and maternal hair mercury levels ≥ 0.55 µg/g (upper 10th percentile). The biologic mechanism supporting this finding requires further investigation. It has been shown that methylmercury produces oxidative stress at the cellular level (Sanfeliu et al. 2003), which may be a contributing factor. In addition, methylmercury can influence shape, aggregation, and levels of platelets (Hornberger and Patscheke 1989; Macfarlane 1981) and thromboxane (Caprino et al. 1983). These effects may potentiate underlying pathology related to maternal vascular diseases in pregnancy. Within the POUCH Study we plan to examine maternal biomarkers of endothelial dysfunction and will explore whether they are related to maternal mercury levels.

Several methodologic differences could account for the inconsistency between our study results and those of previous studies that did not detect an association between mercury levels and gestational age at delivery (Foldspang and Hansen 1990; Fu 1993; Grandjean et al. 2001; Lucas et al. 2004). Our study correlating fish consumption and maternal mercury levels in hair is the largest in the United States to date, providing greater statistical power to detect moderate associations (OR = 3.0) with an outcome that is less frequent, such as delivery at < 35 weeks. Studies using blood levels of mercury (maternal, cord) reflect variations in recent exposure (Foldspang and Hansen 1990; Grandjean et al. 2001; Lucas et al. 2004), whereas mercury in hair is relatively stable and mid-pregnancy hair levels better represent average exposure across the first half of pregnancy. The POUCH Study women had considerably lower mercury levels relative to those of women in other studies, allowing for the testing of lower thresholds. Only one study assessed preterm delivery as an outcome (Fu 1993), and the other three considered mean gestational age at delivery which may have obscured mercury effects in the tail of the distribution (i.e., very preterm) (Foldspang and Hansen 1990; Grandjean et al. 2001; Lucas et al. 2004). In addition, other studies did not adjust for fish consumption, a potential confounder in populations with high levels of fish consumption. Researchers have hypothesized that omega-3 fatty acids in fish can prolong gestation by down-regulating the synthesis of prostaglandin (PG) E2 and PGF2a, and promoting the synthesis of PGI2 and PGI3, thus leading to a more relaxed myometrium (Ferretti et al. 1988; Hansen and Olsen 1988).

Major strengths of the present study include the large number of pregnant women participating, the prospective design, and the use of hair as an index of methylmercury exposure. Hair levels of total mercury represent a longer window of exposure than those of blood levels. This study also had several limitations: First, information on fish consumption was based on recall, which can lead to inaccurate estimates. However, the recall period was focused on eating habits during pregnancy, a time when women are often more aware of their dietary patterns. Because the reports of fish consumption were obtained well before the outcome of the pregnancy, there is little concern about differential recall bias. Second, maternal interviews did not include information on portion size of fish meals and methods of preparation. Factoring in these details might further strengthen correlations between levels of fish consumption and levels of mercury in hair. Third, women were not asked about number of dental amalgam fillings. Total mercury measured in maternal hair may include 10-30% inorganic mercury, the major form of mercury in amalgam fillings (Spencer 2000). However, it is widely accepted that in populations similar to those we have studied, food -- particularly fish -- is the major exposure source of methylmercury, the primary component of total mercury found in hair. In addition, several studies have reported little or no relationship between hair total mercury and exposure to amalgams (Berglund et. al. 2005; Bjornberg et al. 2003; Hansen et al. 2004; Pesch et al. 2002; Tulinius 1995).

An important limitation is our lack of information on maternal levels of other pollutants commonly found in fish (e.g., organochlorines). Women with higher mercury levels in hair may have been exposed to higher levels of other contaminants, such as polychlorinated biphenyls (PCBs) and dichlorodiphenyldichloroethylene (DDE), a metabolite of DDT. Hair mercury levels and blood PCB and DDE levels have all been found to be elevated in high-level fish consumers compared with people who eat less fish (Kosatsky et al. 1999a, 1999b). Based on the estimate of average daily dietary exposure to contaminants for approximately 120,000 adults from the Nurses' Health Study and the Health Professional Follow-up Study, intraindividual exposures to mercury and DDE were found to be moderately correlated among women (r = 0.27) and men (r = 0.17) (MacIntosh et al. 1996). We tried to adjust for other pollutants in fish and for the potentially beneficial aspects of fish (e.g., omega-3 fatty acids) by including total fish consumption in our final analyses. After adjustment, the positive association between maternal mercury levels in hair and risk of very preterm delivery remained, but we cannot rule out residual confounding by other pollutants. Previous studies have failed to find a link between preterm delivery risk and levels of PCBs (Berkowitz et al. 1996; Longnecker et al. 2005) or DDE (Farhang et al. 2005; Fenster et al. 2006; Torres-Arreola et al. 2003). The one exception is a report of excess preterm delivery among women with high blood levels of DDE (Longnecker et al. 2001). This study used stored blood samples from the Collaborative Perinatal Project, a U.S. cohort assembled in 1959 through 1966. The DDE levels associated with preterm birth were considerably higher than DDE levels found in U.S. women of reproductive age today.

Our study reinforced previous findings suggesting that fish consumption is a major source of mercury exposure for pregnant women. Although much attention has been focused on pollutants in locally caught fish, and fish advisories are not uncommon, we found that only a small percentage of pregnant women, < 10%, consumed sport-caught fish during pregnancy. The greatest fish source for mercury exposure appeared to be canned fish, both because it was consumed more and, per meal, it was among the fish categories associated with the highest levels of mercury in maternal hair. The observed relationship between elevated mercury levels and increased risk of very preterm delivery is a new finding and requires caution in interpretation. Although the study sample was large, the number of women who delivered before 35 weeks of pregnancy was small (n = 44), and more studies are needed to test this association.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.