Estrogen replacement therapy (ERT) has not been associated with epithelial ovarian cancer in most reported epidemiologic studies that have looked for an association. Some studies may have found weak statistically nonsignificant associations because the number of cases or number of women who reported estrogen use was small. We performed a meta-analysis of data from 15 case-control studies that provided data on ERT and risk of epithelial ovarian cancer. The 15 combined studies were statistically heterogeneous (chi(2) (14) = 26. 3, P < 0.05) in terms of the effect they found. When we combined these studies using a random effects model, we did not find a significant association of ERT with ovarian cancer (odds ratio = 1.1, 95% confidence interval = 0.9-1.3). There was no clear evidence of a dose-response relation with increasing duration of estrogen use in a subset of five studies that reported estrogen use by duration (overall slope = 0.0012, 95% confidence interval = -0.0055 to 0. 0080). The influences of statistical outliers, study design (hospital or clinic controls vs. community controls), and location (U.S. and Canada vs. Europe and Australia) were examined. The odds ratio was 1.3 (95% confidence interval = 1.0-1.6) in the relatively homogeneous subset of four U.S. case-control studies with community controls, but we cannot rule out the possibility of uncontrolled confounding. The odds ratios for estrogen use for other subgroups defined by geographic location and type of control group were not significantly different from one.