Reply

Abstract

Listen

Translate article

We reported that negative effects of prenatal exposure to polychlorinated biphenyls (PCBs) on cognitive and motor abilities at school age were seen when parental and home environmental characteristics were less optimal, although these effects were not measurable in children raised in more optimal environments.1Vreugdenhil HJI Lanting CI Mulder PGH Boersma ER Weisglas-Kuperus N Effects of prenatal PCB and dioxin background exposure on cognitive and motor abilities in Dutch children at school age.J Pediatr. 2002; 140: 48-56Abstract Full Text Full Text PDF PubMed Scopus (177) Google Scholar The authors of the letter state that this conclusion is erroneous and misleading and illustrate their arguments using the Figure, A, that demonstrates the interaction between PCB exposure and maternal age. We fitted a log dose-response relationship to our data. This is a very plausible relationship, also in toxicology. Application to our data justifies the statistical assumptions underlying the regression analyses more properly than when not log-transforming the dose. Our model simply specifies that the extent of the response will vary with percentage changes of the dose. To some children, a given PCB exposure seems to do more damage than to other children. The purpose of our essentially explorative paper was to try to identify the more vulnerable children. From the results of our explorative linear regression analyses in Table III there is evidence that the log dose-response relation is modified by maternal age, adjusted for the other variables in the regression model. The estimated effect of lnPCB (the log dose) on the response (eg, general cognitive index [GCI]) is estimated to be a significant quadratic function of maternal age in the fitted regression model. The result that, corrected for the other variables in the model, no (significant) overall (ie, without assuming interaction) effect of lnPCB on GCI was found, does not conflict with this. For the sake of argument and illustration, we took a doubling of PCB dose to see how, according to the estimated model, the GCI changes result from this as a quadratic function of maternal age (Figure, A ). We consider this illustration realistic, because the maternal PCB levels ranged from 0.59 to 7.35 μg/L, a more than 10-fold difference. For low maternal ages, there appeared to be a negative effect of doubling the PCB dose on GCI, which effect appeared to level off towards a level near zero as maternal age increases. We could have equally well taken any percent increase in PCB, then only the scale of the y-axis of the Figure, A, (change in GCI points) would have to be changed, the form of the curve remaining the same. A statistical model should be considered an abstraction of nature. Predictions based on a statistical model should always be interpreted with care, especially if these predictions are based on “drastically different” low or high levels of the dose. This is not a particular property of our model; it is a well known property of any statistical model. Moreover, within our cohort we are dealing with environmental levels of PCB exposure (not particularly low or high). Our data provide evidence that parental and home environmental circumstances influence the neurotoxic effects of prenatal exposure to PCBs and dioxins. Animal studies show the biologic basis for such an effect. YMPD177