To review the effect of raloxifene on bone density and fractures in postmenopausal women.We searched MEDLINE from 1966 to 2000 and examined citations of relevant articles and the proceedings of international osteoporosis meetings.We included seven trials that randomized women to raloxifene or placebo, with both groups receiving similar calcium and vitamin D supplementation, and measured bone density for at least one year.For each trial, three independent reviewers abstracted the data and assessed the methodological quality using a validated tool.Data from one large dominating trial suggest a reduction in vertebral fractures with a relative risk (RR) of 0.60 [95% confidence interval (CI) 0.50-0.70, P0.01]. The RR of nonvertebral fractures in patients given 60 mg or more of raloxifene in the larger study was 0.92 (95% CI 0.79-1.07, P = 0.27). Raloxifene resulted in positive effects on the percentage change in bone density, which increased over time and was independent of dose. At the final year, point estimates and 95% CIs for the differences in percent change in bone density (95% CI) between raloxifene and placebo groups were 1.33 (95% CI 0.37-2.30) for total body, 2.51 (95% CI 2.21-2.82) for lumbar spine, 2.05 (95% CI 0.71-3.39) for combined forearm, and 2.11 (95% CI 1.68-2.53) for combined hip (P0.01 at all four sites). Results were similar across studies, and formal tests of heterogeneity did not approach conventional statistical significance. Raloxifene slightly increased rates of withdrawal from therapy as a result of adverse effects (RR 1.15, 95% CI 1.00-1.33, P = 0.05). The pooled RR was significant for hot flashes 1.46 (95% CI 1.23-1.74, P0.01) and nonsignificant for leg cramps 1.64 (95% CI 0.84-3.20, P = 0.15).Raloxifene increases bone density, and the effect increases over 2 yr. The data suggest a positive impact of raloxifene on vertebral fractures. There was little effect of raloxifene on nonvertebral fractures.