Purpose To combine the results of randomized controlled trials to provide overall estimates of the effect of estrogen treatment on fracture rates and measures of bone mass. Data sources Articles on estrogen treatment for osteoporosis published between 1977 and 1995 were identified. Study selection Studies selected were randomized controlled trials of the efficacy of estrogens in preventing loss of bone mass or fractures in postmenopausal women. Data extraction Data extraction and quality assessment were performed in duplicate, with assistance of a manual. Raters were blinded as to authors and their affiliations and the publication details. With estimates of bone mass, the treatment effect size was defined as the difference in the mean annual change in bone mass between the treatment and control groups divided by the pooled standard deviation for change. In the case of fractures, efficacy was measured as the reduction in the numbers of individuals experiencing new fractures with treatment. Effect sizes were pooled using the random effects model. Results Thirty-seven studies met the criteria for inclusion in the systematic review. Only one small secondary prevention trial contained evaluable data on vertebral fractures. This study found a fracture relative risk of 0.63 (95% confidence interval, CI 0.28-1.43) with estrogen treatment. There was more information on the effects of treatment on bone mass. Overall effect sizes ranged between 0.5 and 2.5 standard deviation (SD) units for change. A dose-response relationship was apparent but high doses of estrogens were not associated with effect sizes greater than those observed with recommended doses. There was no significant difference in efficacy between transdermal and oral administration of estrogens. Pooling of paired data from secondary prevention studies indicated that treatment effect sizes were smaller at the hip (0.92, 95% CI 0.3-1.5 SD units) than at the spine (2.1, 95% CI 0.9-3.3 SD units). No significant effects of co-intervention with calcium, progestogens or androgens were seen, although an additive effect of higher doses of calcium could not be ruled out.