The February 27, 1997 issue of the New England Journal of Medicine prints a letter from the Alaska Division of Public Health that casts new doubt on a study often cited by activists for tighter regulation of water quality.
John P. Middaugh M.D. and Grace M. Egeland Ph.D. write, ” Drs. Jacobson and Jacobson reported that low-level exposure to PCBs in utero is associated with lower IQ scores (by an average of 6.2 points) among school-age children. The results seam implausible given the fact that among Taiwanese children who were exposed prenatally to levels of PCBs that were 10 to 20 times higher and to levels of certain congeners of polychlorinated dibenzofurans that were 100,000 times higher than background levels, the IQ score was only 5 points lower than that in unexposed children.
“Several methodologic flaws cast doubt on the validity of the Jacobsons’ findings. They studied only 212 children (47 percent) of the 452 mothers originally invited to participate. This group was a highly selected subgroup of the initial study population of 8,482 women. Descriptive statistics were given for 178 children according to PCB concentration in maternal milk, yet PCBs were measured in only 113 samples of maternal milk. No data were presented to support the value of PCB concentration in breast milk as an accurate measure of prenatal exposure.
“Data on two important risk factors and potential confounders, alcohol ingestion and cigarette smoking, appear to be inconsistent. Although 37 percent of the mothers smoked before and 28 percent during pregnancy, virtually none drank during pregnancy. In contrast, data from the Behavior Risk Factor Surveillance System suggested that a high proportion (14 to 21 percent) of women of childbearing age in Michigan consumed alcohol frequently. Furthermore, the method used to control for potential confounders may not be adequate. The authors stated that a variable’s ‘association with either exposure or outcome can be used as the criteria for inclusion’ in a model. However, standard epidemiologic analytic methods do not advocate the use of this method for model development.
“Given these methodologic issues, we think that this study provides little evidence that in utero exposure to low levels of PCBs affects intellectual function.”
The Middaugh/Egeland letter, and the Jacobsons’ response to it, can be found on the Internet at http://www.nejm.org/public/1997/0336/0009/0660b/1.htm.