OBJECTIVES: This study investigates compliance with hormone replacement therapy (HRT), bisphosphonate medications and raloxifene in the treatment of osteoporosis. METHODS: Data from a large health insurer were used to identify 59,902 osteoporosis patients who initiated drug therapy between January 1, 1998 and August 30, 2000. Multivariate statistical models were developed for duration of therapy, uninterrupted therapy over one year, time to discontinuation, time to a change in therapy and health care costs over one year. Separate models were estimated for the effect of drug use patterns (compliance, switching) and initial drug therapy (raloxifene, bisphosphonates vs. HRT). Other independent variables included age, gender, type of insurance, history of fractures, and patient diagnostic and drug profiles at baseline. Sensitivity analyses were conducted to investigate if the impact of alternative drugs varied with age. RESULTS: Bisphosphonate patients were more compliant than HRT patients, though compliance rates were below 25% for all drugs. Drug use patterns for raloxifene patients did not differ significantly from HRT patients. Compliance was correlated with a reduced risk of hip fractures of 43% (p<0.05) and lower health care costs of -$213 (p<0.01), while switching increased the risk of hip fractures by 84% (p<0.05) and increased costs by $278 (p<0.01). Bisphosphonate patients were twice as likely as HRT patients to experience vertebral, Colles and other fractures and experienced higher health care costs relative to HRT patients of $510 (p<0.0001). The estimated impacts of raloxifene and bisphosphonates improve significantly with patient age. For example, raloxifene's impact on total costs improved from +$314 for patients under 55 to -$570 for patients over 65. CONCLUSIONS: Compliance with drug therapies for osteoporosis over a 1-year period is poor (<25%) leaving patients at risk for fractures and higher health care costs. Alternatives to HRT were associated with better patient outcomes, especially for older patients.