Lisabeth, L., Sioban D. Harlow, X. Lin, B. Gillespie, and M. Sowers. 2004. "Sampling Strategies for Prospective Studies of Menstrual Function." American Journal of Epidemiology, 159(8): 795-802.
Little information is available about optimal sampling strategies for prospective studies of menstrual function. Sample size and study duration for menstrual studies have often been driven as much by feasibility and cost as by statistical principles, with follow-up lasting 6 months to 2 years and sample size ranging from 100 to 500 women. Whether these studies are sufficiently powered to address common study objectives has not been adequately evaluated, and sample size estimates rarely account for the repeated nature of menstrual cycle data. Using data from the Tremin Trust (a study of menstrual function across the reproductive life span initiated in Minneapolis, Minnesota, in 1935 with data collected through 1977), the authors determined sampling strategies for assessing differences in mean cycle length between two exposure groups and for assessing change in mean cycle length across the reproductive life span. Following a larger number of women for 1-2 years is optimal for studies of host and environmental exposures that alter menstrual function. In contrast, following fewer women for an extended period of time, for example, 4-5 years, is optimal when studying how menstrual patterns vary across the reproductive life span in different populations.