Issues in design, measurement, and analysis for menopause research.

Two broad classes of studies relating to menopause are reviewed: descriptive studies (cross-sectional or longitudinal); and comparative studies (observational or experimental). For descriptive studies the issues are: costs of sampling from a general population; measurement of last menstrual period (LMP) in the context of hormone therapy; unbiased symptom reporting in ethnically diverse populations; and in-home vs. clinic-based measurement. Comparative cohort studies of menopause-related therapies are somewhat more vulnerable to selection bias from medical care access than case-control studies with hospital or patient controls. Ideally, in clinical trials, no subjects should have previously used any of the trial therapies to ensure unbiased assessment of side effects and to maximize preservation of masking if a placebo is used. In terms of analysis, information (precision) depends on the proportion in the sample for whom LMP is defined, not the total number in the sample. The role of cigarette smoking as covariate in any analysis is of particular concern as it is the major determinant of the timing of LMP. These design and analysis issues, although identified from research to date, are still new. Some lack wide documentation and are thus not well understood. Others lack clear solutions.