External Validation of Health Economic Decision Models for Chronic Obstructive Pulmonary Disease (COPD): Report of the Third COPD Modeling Meeting.

Abstract

To validate outcomes of presently available chronic obstructive pulmonary disease (COPD) cost-effectiveness models against results of two large COPD trials-the 3-year TOwards a Revolution in COPD Health (TORCH) trial and the 4-year Understanding Potential Long-term Impacts on Function with Tiotropium (UPLIFT) trial. Participating COPD modeling groups simulated the outcomes for the placebo-treated groups of the TORCH and UPLIFT trials using baseline characteristics of the trial populations as input. Groups then simulated treatment effectiveness by using relative reductions in annual decline in lung function and exacerbation frequency observed in the most intensively treated group compared with placebo as input for the models. Main outcomes were (change in) total/severe exacerbations and mortality. Furthermore, the absolute differences in total exacerbations and quality-adjusted life-years (QALYs) were used to approximate the cost per exacerbation avoided and the cost per QALY gained. Of the six participating models, three models reported higher total exacerbation rates than observed in the TORCH trial (1.13/patient-year) (models: 1.22-1.48). Four models reported higher rates than observed in the UPLIFT trial (0.85/patient-year) (models: 1.13-1.52). Two models reported higher mortality rates than in the TORCH trial (15.2%) (models: 20.0% and 30.6%) and the UPLIFT trial (16.3%) (models: 24.8% and 36.0%), whereas one model reported lower rates (9.8% and 12.1%, respectively). Simulation of treatment effectiveness showed that the absolute reduction in total exacerbations, the gain in QALYs, and the cost-effectiveness ratios did not differ from the trials, except for one model. Although most of the participating COPD cost-effectiveness models reported higher total exacerbation rates than observed in the trials, estimates of the absolute treatment effect and cost-effectiveness ratios do not seem different from the trials in most models.