Observational epidemiologic studies suggest that the incidence of cardiovascular disease is reduced by about 50 % in users of unopposed estrogens, but the reduction may have been overestimated because of a greater tendency for women at lower risk to use estrogens. To minimize bias due to such behavior, the authors conducted a case-control study of first myocardial infarction among Massachusetts women aged 45-69 years during 1986-1990, in which each of 858 cases was age-matched with a control from the same geographic area, and important correlates of estrogen use and myocardial infarction were controlled by conditional logistic regression. The estimated relative risk was 0.9 for ever use of unopposed estrogen (95 % confidence interval 0.7-1.2); the estimate decreased with increasing duration of use to 0.6 for 5 or more years of use (p for trend = 0.08). The association with long-term use was stronger for recent use (p for trend