Using several different metrics for recent postnatal MeHg exposure we evaluated the SCDS main cohort for associations with children’s developmental outcomes. We found a number of associations present at the 66 and 107 month evaluations in the primary linear analyses that examined the covariate adjusted association between prenatal exposure and outcomes. Some of the associations in the primary analyses were in the direction of declining performance as postnatal exposure increased and with others the performance improved. Some of these associations were sex specific. One alternative postnatal metric we evaluated showed an association between postnatal THg exposure and IQ present in males only. The associations we found were intriguing, but we were not able to discern a recognizable pattern of associations between the postnatal MeHg exposure metrics we studied and children‘s development.
In the primary analysis at 66 months there were three postnatal associations present. All associations indicated improved performance as exposure increased (Davidson et al., 1998
). For the BG-ES, the improved performance was present in males only. When instead of the 66-month THg level we substituted the19-month THg or the sum of the 19-month and 66-month THg, we also found improved performance on the MSCI-GCI, the PLS-TS, and the WJ-AP as exposure increased.
In the primary analysis at 107 months there were four postnatal associations present. All were in the direction of declining performance as exposure increased. Three were present only in females, a finding we are unable to explain. There is limited evidence regarding postnatal exposure, but prenatal MeHg exposure is generally thought to affect males more than females (Marsh, 1995). The postnatal associations seen here were not consistent across psychological domains since some tests that measured similar cognitive domains showed no association with exposure. If THg affects a psychological domain, it would seem likely that all the tests examining that function would be affected. The absence of consistent findings across ages and psychological domains and concern about continuing brain development and exposure postnatally were factors that led us to explore postnatal exposure and develop alternative metrics.
The reversal of associations between postnatal exposure and endpoints from improved performance to deteriorating performance between the 66 and 107 month exams has not been previously reported, was not expected, and its significance is not clear. The adverse association with the Conner’s Teacher Rating Scale ADHD index is especially intriguing given the prevalence of ADHD and the reports of behavioral changes with other toxicants such as lead (Bellinger, 2008
). However, we are cautious in interpreting this finding since there is no clear evidence that behavioral changes should be expected with MeHg. In addition, we have found some associations with both improving and deteriorating performance on varying endpoints in the past, but no consistent or clear pattern of associations has emerged. A consistent pattern of associations would support a causal relationship. However, inconsistent findings could occur if there is an exposure threshold and if some of the cohort subjects being studied were at or above that level. Presently it is not known if there is a threshold for postnatal MeHg exposure. Varying findings might also occur if the interplay between Hg exposure and the nutrients present in fish such as long chain polyunsaturated fatty acids, iodine, selenium, or other factors were more complex than we presently understand or more important earlier in development (Davidson et al., 2008
; Strain et al., 2008
We examined the association of the alternative postnatal metrics only with the IQ measured at 107 months. A significant association was present only when the model included a postnatal THg by sex interaction, only for males, and only for the H-L metric with the most extreme cut points. This model suggests that boys who are consistently exposed to higher levels of THg postnatally did better on IQ testing than boys consistently exposed to lower THg within the range we are studying in Seychelles. However due to the small sample size used for this model, this suggestion should be interpreted with caution. The models using the alternative metrics did show the expected positive association between maternal IQ and 9-year child IQ (p<0.005). The HOME score in all models was a significant predictor of 9 year IQ (p<0.005). Other covariates known to be significant predictors of 9 year IQ such as SES and maternal age were also significant in some models. These finding suggest that the data are robust enough to detect associations known to be associated with children’s IQ and might have detected an association with postnatal THg exposure if its effect size was similar to that of these covariates.
Each of the alternative postnatal metrics could be constructed using recent THg measured at multiple time points. Unfortunately, we had only samples obtained at the time of evaluations for postnatal analysis and there were significant amounts of missing data. Consequently we were limited to using at most only 3 time points from which to calculate each metric. Two of the alternate postnatal metrics (AUC and BGW) are weighted averages of the recent postnatal THg values at each time point. These two metrics are scaled in different ways and weight the time of exposure differently. The AUC metric weights the THg measurement at each time point in proportion to the period of time covered by the measurement (which is determined by how close in time other THg measurements were taken). However the AUC metric does not treat any time period as being more important than any other time period. In contrast, the BGW metric assumes that THg exposure that occurs during periods of rapid brain development are the most detrimental to the child. When determining weights for the BGW metric based on changes in brain volume, the 6-month THg level accounted for nearly 100% of the weight, suggesting that THg exposure that occurs much later is less important in comparison to the 6-month exposure. Had we extended this thinking to derive a BGW metric that included the prenatal period, it seems likely the prenatal exposure would have had a much greater weight than even the 6-month THg exposure. The rationale for the BGW metric is therefore consistent with the theory that prenatal THg exposure is more detrimental to neurodevelopment than postnatal exposure, because the prenatal period is the time of most rapid brain growth. Our BGW metrics assume that the changes in head area or volume are reasonable proxies for overall brain development, but this may not be entirely accurate. The metrics we used might not account for some developmental brain processes such as myelination that can continue into early adulthood.
An advantage of the AUC and BGW metrics as compared to the H-L metric is that all the data on postnatal THg levels could be included as well as all subjects with postnatal data. The H-L metric differs in that an increasing number of subjects were excluded as the “high” and “low” categories moved farther apart. This reduction in sample size for the H-L models is one reason we might expect results using this metric to differ from the other models. However, if consistent high exposure to THg is detrimental, this metric may be better than the other metrics to detect an association. Indeed, at the highest cut points (4 and 8) there was a significant postnatal THg by sex interaction and a significant increase in IQ present in males.
This study has a number of strengths. The cohort size was large and over 500 children had postnatal exposure measured at the time points we evaluated. The population studied consumes fish daily and does not consume marine mammals. The average postnatal exposure for the cohort ranged from 6.6 ppm at 6 months of age to 4.8 ppm at 66 months. The analyses did show the expected effects of covariates known to influence child development such as maternal IQ, HOME, SES, and maternal age, suggesting they might have detected an association with postnatal THg if one was present.
The study also has limitations. The primary goal of the SCDS was to study prenatal MeHg exposure and there was incomplete collection of postnatal hair samples at some ages. Measuring continuous postnatal exposure would have been preferable to the one centimeter recent exposure that we measured, but we had neither the hair samples nor the resources to recapitulate more continuous postnatal exposure. The optimum postnatal metric that most closely reflects the brain exposure is not presently known. Although we selected three postnatal metrics with biological rationales, other metrics or combinations of metrics might have resulted in different associations. Similarly, if we had examined the association of postnatal exposure with other outcomes, the results might have been different. Our metrics that recapitulated exposure over time assumed that the exposures measured at the three time points were representative of other unobserved exposure over this time period. However, this assumption would not be warranted if exposure were episodic as might occur in societies where the source of THg exposure is different than daily consumption of fish. There may also be differences in exposure effects on development at different ages related to different events in the developing central nervous system, even if the exposure is not episodic. The mean THg exposure present in the SCDS may have been too low to find more than inconsistent associations with the endpoints measured. Postnatal exposure could also have been influenced by nutritional variables that were not available for these studies. Recent evidence suggests that long chain polyunsaturated fatty acids and other nutrients present in fish may significantly modify the influence of exposure (Davidson et al., 2008
; Strain et al., 2008
In summary, there are biological reasons to believe that postnatal exposure to MeHg might influence children’s development. We measured recent postnatal exposure to MeHg at several ages in the SCDS main cohort and examined the association of several different postnatal metrics with some of the subjects developmental test scores. We found a number of associations between postnatal exposure metrics and test outcomes, but the results varied across ages and psychological domains. These findings are consistent with our earlier findings in the SCDS and do not provide clear evidence for an adverse association between the levels of THg exposure studied in this cohort and the children’s development. However, the findings do raise intriguing possibilities and suggest that postnatal exposure should be studied prospectively.