Estimating Exposure to Methylmercury: Effects of Uncertainties

Uncertainties in exposures can lead to biased estimates of slopes and thresholds in the exposure-response relationships that are developed from regression analysis. This paper reviews published exposure and epidemiological studies of methylmercury (MeHg) from the perspective of the accuracy and precision of the estimates used to represent the actual doses received. Sources of such uncertainties, collectively referred to as “exposure errors”, include instrumental and analytical errors, sampling and survey uncertainties, and individual variability in the relationships between the exposure metrics and the actual doses to target organs. Because the relationship between maternal intake and the consequent dose to the fetal brain varies among individuals, epidemiological studies of the effects of prenatal exposure must necessarily be accompanied by larger exposure uncertainties than comparable studies of effects on the mothers. The increased exposure errors typically result in attenuated slopes of the dose-response functions and under-estimates of thresholds, so that part of the apparent increased sensitivity of the fetus that has been developed from epidemiological studies may in fact be due to their inherently less certain exposures. Sources and magnitudes of exposure error found in the literature are discussed and their statistical ramifications are explored with Monte Carlo simulations. The paper also finds that, after adjusting for exposure error, the relationship between dietary intake and blood concentration is consistent with an average half-life shorter than has typically been used and that using population averages yields a consistent but sub-linear relationship between dietary intake of Hg and hair concentration. Investigators are urged to obtain (and present) data on more than one exposure metric, so that their relative uncertainties may be assessed independently.