ISSUES IN DESIGN, MEASUREMENT, AND ANALYSIS FOR MENOPAUSE RESEARCH

Two broad classes of studies relating to menopause are reviewed: descr iptive studies (cross-sectional or longitudinal); and comparative stud ies (observational or experimental). For descriptive studies the issue s are: costs of sampling from a general population; measurement of las t mentrual period (LMP) in the context of hormone therapy; unbiased sy mptom reporting in ethnically diverse populations; and in-home vs. cli nic-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 contro ls. Ideally, in clinical trials, no subjects should have previously us ed any of the trial therapies to ensure unbiased assessment of side ef fects 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 sam ple. 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 dear solutions.