Effect of different sampling designs on outcome of endocrine disruptor studies.

In this article, we demonstrate how sampling strategy can influence the outcome of endocrine disruptor studies. In a study of the weak xenoestrogen bisphenol A (BPA), possible treatment-related effects on ventral prostate (VP) fresh weight were found in rat offspring at 6 months of age when only one or two male pups were randomly selected from each litter. In subsequent BPA and di-n-butyl phthalate studies, large intralitter variability of this specific end point was apparent when the VP weights from entire litter complements were examined. We modeled the effects of sampling 1, 2, or 3 pups from each litter using the full-litter complement data. When one pup was randomly selected, a substantial percentage of incorrect conclusions about the presence or absence of treatment effects occurred. These statistical modeling analyses raise significant concern about the selection of one pup per litter for highly variable end points.