These results support findings from some studies that higher mercury exposure in pregnancy is associated with lower offspring cognitive scores, even at these relatively low levels of exposure. In addition, higher maternal fish intake was associated with higher mercury levels. However, higher maternal fish consumption was associated with better infant cognition. This benefit appeared greatest among infants whose mothers consumed more fish but had lower mercury levels.
The conflicting results from the two large-scale longitudinal studies in the Faroe and Seychelles Islands have led to disagreement about whether moderate mercury exposure from frequent seafood consumption may harm offspring development. The mean maternal hair mercury levels in the Faroe Islands (4.3 ppm) (Grandjean et al. 1997
) and Seychelles Islands (6.8 ppm) (Davidson et al. 1998
) studies were much higher than in our cohort (0.55 ppm), which had levels similar to other pregnant (Morrissette et al. 2004
; Stern et al. 2001
) and nonpregnant (Centers for Disease Control and Prevention 2001
) U.S. populations. Some data suggest that no threshold exists for adverse neuropsychologic effects from methylmercury exposure (Rice 2004
). A recent study in the United Kingdom did not show any adverse association of low levels of umbilical cord tissue mercury (median, 0.01 ppm) with child development (Daniels et al. 2004
). However, in the U.K. study, development at 15 months of age was assessed by parental self-report, which is not likely to be as sensitive to the adverse effects of mercury.
It may seem contradictory that, on the one hand, fish intake raises mercury levels and higher mercury levels lead to worse cognition but, on the other hand, higher fish consumption is associated with better cognition. The most likely explanation is that the benefit is conferred by consuming fish types with the combination of relatively little mercury and high amounts of beneficial nutrients. This explanation is supported by results from multivariable models in the present analysis, in which adjustment for mercury strengthened the observed positive association of fish intake and cognition. Similarly, in the stratified analysis, we observed the highest cognitive scores among offspring of mothers with higher fish intake but lower mercury levels.
The fish questions in the Project Viva food frequency questionnaire were designed to estimate intake of fatty acids, not mercury. Because this questionnaire does not assess intake of individual fish types, but rather groups of fish, we cannot report associations for specific types of fish. Mercury levels vary among different fish species. In general, white meat fish such as cod and haddock tend to have lower mercury levels but also lower levels of long-chain n-3 fatty acids, whereas dark meat fish, such as swordfish, mackerel, and other large long-lived predatory fish, tend to contain both more mercury and more n-3 fatty acids. Because mercury and n-3 fatty acids often travel together, it may be difficult to isolate the opposing influences of the two on child cognition. Small fatty fish such as sardines and canned light tuna (vs. albacore tuna) may contain relatively more fatty acids with less mercury. Future studies incorporating more detailed dietary information may help advise women about specific fish species that are better or worse for their children’s cognition.
Resolving this issue remains important because women may indiscriminately reduce fish consumption in response to concerns about mercury exposure, perhaps substituting fish with other, less healthful foods. In a previous study, we demonstrated that a different subset of pregnant women enrolled earlier in our cohort reduced consumption not only of dark meat fish, which are likely to have higher mercury levels, but also of canned tuna and white meat fish, which tend to have lower mercury levels, after dissemination of a 2001 U.S. federal mercury advisory (Oken et al. 2003a
). A recently updated federal advisory reiterated health warnings while encouraging women to consume up to two seafood meals per week (U.S. Department of Health and Human Services 2004
), but it is unclear to what extent women understand the details of the health message or simply hear that seafood contains mercury and therefore is harmful.
In Project Viva, we assessed infant cognition using the VRM protocol. This test of cognitive function has many advantages for studies of prenatal exposures: It can be performed early in infancy and assesses cognition isolated from motor function. VRM tests have predicted IQ in childhood and early adolescence as strongly as other standardized tests of infant development (e.g., the Bayley Scale of Infant Development) (Rose and Feldman 1995
; Rose et al. 1992
), with correlations with intelligence in later childhood ranging from 0.44 to 0.66 (Laucht et al. 1994
). However, similar to all tests of infant cognition, the VRM is most strongly correlated with later IQ when mental development is impaired, whereas the relationship is less strong when cognition is within the normal range. The VRM, which uses visual preference rather than motor skills to assess cognitive ability, might be particularly sensitive to the benefits of fish consumption: The marine fatty acid docosahexaenoic acid is an essential component of the retina and promotes infant vision (SanGiovanni et al. 2000
). VRM testing has been used to demonstrate associations of infant cognition with infant intake of marine n-3 fatty acids (O’Connor et al. 2001
; Uauy et al. 2001
) as well as prenatal exposure to environmental pollutants such as lead and polychlorinated biphenyls (Darvill et al. 2000
; Emory et al. 2003
). Inverse associations of methylmercury exposure and novelty preference scores have been seen among non-human primates (Gunderson et al. 1986
); however, other studies among human infants, most notably in the Seychelles Islands cohort, have not demonstrated an association of prenatal mercury exposure with VRM results (Darvill et al. 2000
; Myers et al. 1995
). Future cohort follow-up will help determine whether fish and mercury are associated with intelligence later in childhood.
Results should be generalized with some caution because our study population contained a high proportion of educated, white, and well-off mothers, all of whom lived in one area of the United States. Even within this sample, however, the main exposures and outcome were similar to those of other populations. For example, levels of hair mercury were similar to those in women of childbearing age in the most recent National Health and Nutrition Examination Survey (90th percentile, 1.4 ppm), which is nationally representative of the U.S. population (Centers for Disease Control and Prevention 2001
). The mean novelty preference score of 59.8 in our population was similar to that observed among children enrolled in the Seychelles Child Development Study (60.3) (Myers et al. 1995
) and among U.S. children living in New York State (59.8) (Darvill et al. 2000
As with all observational studies, it is possible that we did not completely adjust for potentially confounding factors. For example, we did not collect information on parental IQ or home environmental stimulation, because Project Viva includes many outcomes of interest in addition to cognition, and we wished to reduce participant burden. Both of these factors might be associated with fish intake and have been associated with child cognition in prior studies, although usually only beginning in the later half of the second year of life. We also did not measure exposure to persistent organic pollutants such as polychlorinated biphenyls, although we would expect that accounting for the adverse effect of these toxicants would strengthen the observed benefit of seafood. We assessed mercury exposure using total maternal hair mercury, which is a recommended biomarker for estimating the methylmercury dose received by the offspring’s brain (National Research Council 2000
) and strongly associated with both maternal blood and cord blood organic mercury (Morrissette et al. 2004
). More than 80% of the mercury in hair is in the methyl form; in addition, organic mercury can be partly transformed to inorganic mercury, so the concentration of total mercury more accurately represents the mercury entering the hair follicle from the blood stream (Davidson et al. 1998
; Myers et al. 2003
). Some mercury may have derived from the inorganic mercury in dental amalgams. If this were true, however, the beneficial effect of fish on cognition then would be even stronger than what we observed. Our sample was too small to address whether there are subgroups within the population that are more susceptible to mercury exposure because of genetic, nutritional, or social factors that may sensitize the developing brain to neurotoxicants.
Results from the present study among a U.S. cohort with moderate fish intake suggest that maternal fish consumption during pregnancy may benefit offspring cognition in infancy but that exposure to higher levels of mercury has adverse effects on child cognition. These findings, based on a relatively small group of women, merit further investigation and verification in other populations consuming moderate amounts of seafood. Meanwhile, we recommend that women continue to consume fish during pregnancy but seek out varieties with lower levels of mercury.