Many surveys have shown that women using postmenopausal hormone therapy have a lower incidence of cardiovascular diseases and lower overall mortality. The purpose of this study was to compare past and non-users with current users of hormone therapy in regard to characteristics known to, or assumed to, predict poor subsequent health (indicators). The main data source was a survey in 1989 of a random sample (n = 2000, 86% response rate) of 45-64 year-old Finnish women. Among women with their uterus, after adjusting for age and urbanism, of the 21 indicators studied, 10 suggested a poorer and none a better health prognosis for the non-users than for current users. Many differences were greater among older women, suggesting a cohort effect or long-term users being an especially selected group. Among hysterectomized women, differences between users and non-users were similar or smaller than among women with uteri. The past users were more similar to non-users than current users. Our study suggests that women with a better health prognosis are selected or select themselves for hormone therapy, and that may impede interpretation of observational studies on hormone therapy and health.