Ovarian follicle loss in humans and mice: lessons from statistical model comparison.

BACKGROUND Mammalian oocyte stocks reach maximum size in early development and begin depletion immediately thereafter. This depletion ends women's fertility by midlife. Here we compare five models proposed to characterize human follicular depletion, highlight underlying variation in atresia, and use oocyte counts from laboratory mice to illustrate possible effects of known covariates. METHODS We compared statistical models, of human data, from five well-known sources and also compared the models' fit to data from four genetically distinct strains of mice. RESULTS A model first published by Hansen et al. (2008) fit the human data better than any of the alternatives. Best-fit models of oocyte loss in the four strains of mice differed substantially from the best-fit model of the aggregated mouse data. CONCLUSIONS Although the power model published by Hansen et al. (2008) fit the human data best, Faddy and Gosden's (1996) differential equation model may be a more useful characterization of human follicular atresia. However, these models leave a great deal of variation unexplained. Mouse strain comparisons show that follicle loss in genetically distinct subpopulations can differ substantially from the pattern in the aggregate population. This indicates that differences in follicular stock size between and within populations depend upon more than a single predictor (i.e. age or follicle stocks at previous time points). Our reliance upon data from Western populations represents this study's most important limitation. Expanding data collection to include likely covariates and a wider range of human populations would improve the basis for predicting individual trajectories of follicle loss as more women worldwide opt to delay childbearing and risk aging beyond their own windows of fertility.