Influence of sampling on the reproducibility of ovarian follicle counts in mouse toxicity studies

Different ovarian follicle counting procedures were investigated to reduce labor while retaining statistical power. Intact ovaries of untreated CD-1 mice (20/group) from National Toxicology Program Reproductive Assessment by Continuous Breeding (RACB) studies were serially sectioned at 6 microm. Mean numbers of small and growing follicles were used to assess sampling efficiency. In 10 mice per group, comparisons were made between 10% nonrandom samples from every 10th section starting at either the first or sixth section having follicles (approximately 40 sections per ovary). These 10% counts were compared with 5% (20 sections) and 20% (80 sections) nonrandom samples and with 1% (4 sections), 5%, or 10% random samples from the same 10 animals. For two studies, a 10% nonrandom sample was analyzed from 20 mice per group. Follicle counts for each group were comparable regardless of the sampling paradigm. Four to 10 animals provided 90% confidence that a 20% difference in mean counts would be detected. The 1% sample had a larger error term and, thus, slightly reduced statistical power. These data suggest that follicle counts from 1% or 5% random samples may provide a suitable screen for ovarian toxicity.