Estimating survival for cost-effectiveness analyses: A case study in atherothrombosis

Background: Economic evaluations typically require estimates of survival beyond the limited follow-up in clinical trials. The objective of this study was to demonstrate a data-driven approach to deriving these estimates. Methods: To provide survival estimates for analyses of the CAPRIE trial, data were obtained from Saskatchewan on more than 50,000 patients like those in the trial: diagnosed with peripheral arterial disease (PAD), myocardial infarction, or ischemic stroke between 1985 and 1995; follow-up to December 31, 2000. Mean survival was estimated by integrating the full survival curve derived by applying hazard functions over time. Cox proportional hazards analyses were carried out in each of four periods defined to ensure proportionality. Results: Adjusting for mean age in CAPRIE, mean survival ranged from 12.1 years after index stroke to 13.6 years after diagnosis of PAD. Comorbidities reduced mean survival by 1 to 2 years. Subsequent events had a marked detrimental effect, decreasing life-expectancy by 50% or more, and disease in multiple vascular beds led to survival of less than 5 years. Discussion: These analyses demonstrate the analytic methods required to accurately estimate survival. The trial ages were much lower than in the observational study. Thus, the estimates are optimistic for the general population. Conclusions: Accurate valuation of interventions depends on valid survival estimates. These analyses confirm that survival is significantly reduced in patients with atherothrombotic disease, particularly with additional comorbidities.