OBJECTIVES: Objectives were: 1) to determine the effectiveness of osteoporosis medications; 2) to identify risk factors and other covariates that predict osteoporotic fractures; and 3) to determine the cost-effectiveness of osteoporosis medications. METHODS: A retrospective cohort study was conducted using DOD claims from fiscal years 2000 to 2003. Using an intent-to-treat study design, a total of 49,851 women >= age 50 were followed. Differences in effectiveness (fracture/no fracture) were determined by both logistic and direct Cox proportional hazard regressions. To assess cost-effectiveness, a net-benefit regression method was employed. RESULTS: Findings showed that the three-year cumulative incidence of an osteoporotic fracture in this cohort was 2.5% (0.4% in patients without an osteoporosis diagnosis; 6.1% in patients with an osteoporosis diagnosis). The medication effectiveness results obtained from both regression models were consistent and suggested that women treated with the combination of alendronate and HRT were at lower risk for any fracture, hip fracture, and vertebral fracture when compared to no treatment, while comparisons of the individual medications with no treatment did not show a significant decrease in risk. Variables that significantly increased the risk of fracture were: prior fracture, increasing age, and oral corticosteroid use >= year. The net-benefit regression showed that the use of osteoporosis medications was not cost-effective overall in the short-term compared to no treatment. However, among high-risk populations, such as patients with a prior osteoporotic fracture or those >= 65, medications became more cost-effective. CONCLUSIONS: Combination therapy with HRT and alendronate was more effective than no treatment in DOD women >= 50. None of the treatment options were cost-effective in the short-term for the overall population, but some were more cost-effective in subsets of high-risk patients. The results of this study were potentially influenced by the presence of selection bias, therefore propensity scoring will be conducted.