Exploring nonlinear association between prenatal methylmercury exposure from fish consumption and child development: evaluation of the Seychelles Child Development Study nine-year data using semiparametric additive models.

Studies of the association between prenatal methylmercury exposure from maternal fish consumption and neurodevelopmental test scores in the Seychelles Child Development Study have not found adverse effects through age 9 years. The analysis for the most recent 9-year data (Lancet 361 (2003) 1686) employed conventional linear regression models. In this study we reanalyzed the same Seychelles 9-year data using semiparametric additive models with different degrees of smoothing to explore whether nonlinear effects of prenatal exposure were present. Of 21 endpoints in the linear analysis, we chose only those with a two-tailed P value less than 0.2 for the effect of prenatal exposure. Six endpoints met the criterion. A nonlinear effect was identified with the more smooth model for only one endpoint. The test for an overall effect of prenatal exposure was also significant, with a P value of 0.04, while the corresponding P value in the linear regression analysis was 0.08. The nonlinear curve appeared to be nearly flat when the level was below approximately 12 ppm in maternal hair, with a linear trend above that level, suggesting a possible adverse effect in the uppermost range of prenatal exposure included in this cohort. Because of the descriptive nature of semiparametric additive models, the P values are not precise, and certainly there are fewer data above 12 ppm. We conclude that this reanalysis supports the primary linear analysis, showing little evidence for a prenatal adverse effect.