Analytical approach:
A decision model was used to extend cost and effect results from a one-year trial to a lifetime horizon. The authors reported that a third party healthcare payer perspective was adopted.
Effectiveness data:
Most clinical and effectiveness data were derived from a randomised controlled trial (RCT) conducted in large emergency departments and fracture clinics in Alberta (see Other Publications of Related Interest). Patients aged at least 50 with universal healthcare were randomised to receive the intervention (137 patients) or usual care (135 patients). The primary outcome of the trial was starting bisphosphonate therapy within six months of fracture (22% for intervention group compared to 7% for controls). Estimates of reduction in fracture risk with alendronate were obtained from systematic reviews. Other model inputs were derived from USA population statistics and published literature.
Monetary benefit and utility valuations:
Age-specific utility scores were derived for each of the model health states (post-wrist fracture, wrist fracture, post-hip fracture, hip fracture, spine fracture and death). The utility estimates were based on utility weights and their proposed multipliers from two published papers.
Measure of benefit:
Quality-adjusted life-years (QALYs) and numbers of fractures discounted at an annual rate of 3%.
Cost data:
Costs associated with the intervention, osteoporosis treatment and subsequent fractures were considered. Costs and resource use estimates were derived from various published sources including micro-costing studies, regional and national databases and life tables. All costs were reported in 2006 Canadian dollars ($). Future costs were discounted at an annual rate of 3%.
Analysis of uncertainty:
Analysis of uncertainty included a series of one-way and probabilistic sensitivity analyses. One-way analysis was conducted on key model inputs that included intervention and treatment costs, treatment effects and the discount rate. Probabilistic analysis involved assigning probability distributions to model parameters and subsequently running 5,000 simulations.