Discussion
Limiting maternal exposure to mercury to decrease potential adverse neurodevelopmental effects on the fetus has been the subject of much discussion in the medical and environmental literature (e.g., Holmes et al. 2009). The concerns have influenced public policy, and efforts to reduce maternal mercury exposure have focused on limiting seafood consumption, which has been presumed to be the chief source of exposure. However, our findings suggest that seafood accounted only for an estimated 6.98% of the variation in blood mercury levels in the pregnant women included in the analysis, who were representative of the general ALSPAC population in regard to seafood intake. This accounted for less than half of the variability in blood mercury explained by the dietary factors included in our analysis.
The estimated proportion of food-related intake associated with seafood in our study population was slightly higher than estimated from UK dietary surveys for dietary consumption of mercury over 1 week, which suggested that 25% of dietary mercury came from seafood [based on 1994 survey data (Ysart et al. 1999)] and 33% [based on 1997 data (Ysart et al. 2000)]. Their measures, however, did not take protective dietary or absorption factors into account. The higher estimate of dietary mercury consumption from seafood in the present study may reflect different forms of mercury (e.g., methylmercury) and hence different absorption rates of the different foodstuffs. Nevertheless, our findings suggest that although seafood is a component of dietary mercury exposure, it may contribute less than half of the overall mercury intake from dietary sources. Importantly, a large proportion of the blood mercury variance was not associated with any dietary variable including seafood.
Herbal teas were unexpected dietary predictors of total blood mercury in our study population. Mercury is found at relatively high levels in some folk and patent preparations (Espinoza et al. 1995; Liu et al. 2008), which are often found in health food shops in the United Kingdom, and herbal preparations such as herbal teas may have similar contaminants. However, although herbal tea consumption was a significant predictor in several models, it contributed less to the overall variance than seafood consumption because only 18% of participants reported that they drank herbal teas, whereas 88% consumed seafood.
Some dietary factors were negative predictors of total blood mercury in our study population, including white bread, whole milk, sugar, french fries, baked beans, and meat pies/pasties. Consistent with these findings, Bates et al. (2007) reported negative associations between total blood mercury and white bread, whole milk, sugar, and chips (french fries) in a study of 1,216 British adults 19–64 years of age. Although positive associations are interpreted as evidence of contamination with mercury, explanations for negative associations are less clear, but might reflect the effects of dietary constituents that limit the absorption or accelerate the elimination of mercury from the body. Alternatively, people who are more likely to eat these foods may be less likely to consume foods that are sources of dietary mercury. In this context, food items that are negative predictors of total blood mercury may be serving as a proxy indicator of low consumption of food items that are positively associated with blood mercury levels.
We have analyzed the dietary factors contributing to total blood mercury levels among pregnant women residing in Avon, an area of the United Kingdom that is largely representative of England as a whole. The analyses of maternal blood were performed in the same laboratories as the NHANES (National Health and Nutrition Examination Survey) surveys conducted in the United States (Mahaffey et al. 2004). A comparison of blood mercury concentrations of the 4,134 pregnant women who donated blood samples in 1991–1992 for the present study with those of 286 pregnant women in the 1999–2000 NHANES study suggests a marked difference. Specifically, the median value in the present UK population of pregnant women was twice that reported for the American pregnant population (1.86 vs. 0.89 μg/L, respectively), as were the 10th (0.81 vs. 0.15 μg/L) and 25th percentiles (0.99 vs. 0.38 μg/L). However, the 90th and 95th percentiles for the U.S. study (4.83 μg/L and 5.98 μg/L, respectively) were higher than in the United Kingdom (3.33 μg/L and 4.02 μg/L).
The proportion of women of childbearing age with blood mercury levels above the recommended level for adult women (5.8 μg/L) was 8% in the NHANES study (Schober et al. 2003) compared with 1.9% in a study in North Carolina (Miranda et al. 2011) and 0.9% in the present study. The difference between the high ends of the distributions of blood mercury in NHANES and ALSPAC is unlikely to be attributable to differences in the consumption of seafood, as this is less in the United States than in the United Kingdom, and the mercury levels of seafood eaten in the United Kingdom are generally higher than those of seafood eaten in the United States (Hibbeln et al. 2007).
Relatively few studies have evaluated demographic predictors of blood mercury levels. There have been reports of positive associations between blood mercury and age (Batariova et al. 2006; Caldwell et al. 2009), and maternal blood mercury levels have been associated with higher education, income, and ethnicity (Mahaffey et al. 2009; Miranda et al. 2011). Although it has been suggested that these associations may just reflect differences in the amounts and types of seafood consumed, our analysis suggested that associations with sociodemographic factors persist when adjusted for dietary factors, and vice versa.
The diet is not the only contributor to blood mercury levels. Mercury can also be absorbed from water and air, and from nondietary products such as dental amalgam fillings, beauty products, social drugs such as cigarettes and alcohol, illicit drugs, and medications. Mercury vapor in the atmosphere is absorbed mainly through the respiratory tract (Holmes et al. 2009). Once absorbed, the mercury is widely distributed to fat-rich tissues, and is readily transferred across the placenta and blood brain barriers. Major sources include refuse incineration, fossil fuel combustion, and fungicides/pesticides (Hutton and Symon 1986). It has been estimated that 9.9 tons of mercury are deposited on the United Kingdom from the atmosphere each year (41% from sources in the United Kingdom, 33% from elsewhere in Europe, and 25% from other parts of the northern hemisphere) (Lee et al. 2001).
Although there are a number of strengths to our study, including the large sample size and representativeness of the local population, there are some potential weaknesses. The blood samples assayed for mercury were stored for 18–19 years before funding was available to process them, and methods to process the clotted blood samples had to be developed de novo in the CDC laboratories. Although the physical integrity of the samples was maintained by ALSPAC, and the analytical methodology was verified by CDC, the age of the samples may have resulted in some degree of analytical inaccuracy.
Information on the mother's diet was collected using a self-completed FFQ rather than weighed intakes. This method has been shown to be appropriate for estimating the intake of foods that are not eaten daily (such as seafood) (Emmett 2009), and the method has been demonstrated to provide adequate assessments of trace metal intake when compared with the duplicate diet method (Liu et al. 2010). We were not able to account for cooking methods or possible joint effects of foods and drinks consumed at the same time. Ouédraogo and Amyot (2011) have demonstrated in vitro that mercury bioaccessibility is reduced if the fish is fried or consumed with black tea or coffee. These findings have not been tested in vivo yet, but it is tempting to suggest that similar mechanisms may account for some of our protective findings. We acknowledge that the dietary measures used in this study are estimates, with a wide error component. This is likely to reduce the amount of variance explained overall. Although we believe that the errors should be similar for all dietary items, we cannot rule out the possibility that the ratio of the variance associated with seafood to the variance associated with all dietary items may be biased due to differences in the accuracy of the intake estimates for different food items.
Most studies have measured mercury in maternal hair rather than in blood on the assumption that it would give a long-term cumulative measure of fetal exposure (e.g., Davidson et al. 1998, 2008; Grandjean et al. 1997). Cord tissue mercury levels were more closely related to maternal blood levels (r = 0.85) than to maternal hair levels (r = 0.65) in a Japanese study population (Sakamoto et al. 2007), which suggests that blood mercury levels may provide a more accurate measure of fetal exposure in late pregnancy. However, maternal hair level may be a better proxy for fetal exposure early in pregnancy, corresponding to our maternal blood measures that were drawn in the first half of pregnancy. We measured total blood mercury levels rather than methylmercury levels that are likely to be the major form of mercury in fish. However, Sakamoto et al. (2007) estimated that approximately 90% of total blood mercury was methylmercury in their study population of 116 mother–infant pairs in three districts of Japan, and that correlations between total mercury and methylmercury were very high (r = 0.98 for maternal blood and 0.97 for umbilical cord tissue). If these estimates apply to our UK study population, total blood mercury levels are likely to be similar to blood methylmercury levels.