OBJECTIVES: To highlight the impact of post-fracture treatment costs and time horizon on the cost-effectiveness of hypothetical interventions. METHODS: A Markov model was used to simulate the natural history of osteoporosis and the clinical, economic and health-related quality of life impacts of a hypothetical post-menopausal osteoporosis intervention. The intervention was assumed to reduce all fractures (50% vertebral and hip; 25% wrist and other), and cost $650 (US$) annually. The model tracked the annual incidence of all fractures (vertebral, hip, wrist, other), as well as mortality due to natural causes and hip fractures. Age-specific mortality and fracture incidence rates were derived from published US data sources. Fracture rates were adjusted to reflect a patient population at increased risk due to low BMD (T-score –2.5) and an existing vertebral fracture. Acute fracture costs (within year of fracture) and annual fracture costs in all years subsequent to the hip or vertebral fracture were taken from the NOF status report (1998). RESULTS: When patients were treated from 70 years of age for five years and costs collected over a similar time frame, the cost per QALY gained was estimated as approximately $20,000. When the treatment length remained 5 years but the time horizon over which costs were accumulated was increased to 30 years (or until death), the incremental ratio approached a win-win (reduced costs, increased QALYs). Several analyses were conducted to determine the components contributing to this trend. CONCLUSIONS: Dramatic change in the cost-effectiveness of an intervention can be produced by modifying time horizon. The Canadian and other guidelines for economic evaluations recommend a time horizon sufficient to capture all downstream therapy effects. This work provides an example of the importance of time horizon selection. In addition, it highlights the need for accurate long-term costs of hip and vertebral fractures.