Assessing human polychlorinated biphenyl contamination for epidemiologic studies: lessons from patterns of congener...

Humans are always exposed to mixtures of polychlorinated biphenyls (PCBs), so assessment of their health effects is complicated. Because the original sources are relatively standard mixtures that change in predictable ways while traversing the environment, there is substantial uniformity in the congener mixtures people carry. To the extent that concentrations are highly correlated, measuring multiple congeners within correlated groups would be unnecessary and estimation of separate biologic effects would be impossible. We examined correlation patterns in previously collected data on 38 congeners (and 14 other organochlorines) from 497 human milk samples from Canada from 1992. Congeners 138, 153, 156, 157, 170, 183, 187, 194, 199, and 203 were highly intercorrelated; 180 had slightly lower correlations with this group. Congeners 74, 105, and 118 were highly intercorrelated and moderately to highly correlated with the first group. Congener 99 had moderate correlations with both these groups, and congener 66 had lesser correlations with the primary group. In contrast, congeners 28, 44, 49, 60, 90/101, 128, 137, and 193 showed little correlation with any other congeners. The remaining 14 congeners were uninformative; they were quantified in fewer than 30% of samples, and varying lipid concentrations meant that those quantified were not necessarily at higher concentrations than those not quantified. In study of human health effects of PCBs, the congener pattern present in the population under study should be examined when deciding which congeners to measure; instead of solely redundant or uninformative congeners, attention should be given to other congeners that may be more useful in addressing the question of interest.