Effects of Cohort Selection on the Results of Cost-Effectiveness Analysis of Disease-Modifying Drugs for Relapsing-Remitting Multiple Sclerosis

BACKGROUND: Decision-analytic cost-effectiveness models are used to determine the most cost-effective treatment option on the basis of the best available data. Guidelines for pharmacoeconomic model development indicate that models should be updated as new evidence becomes available. OBJECTIVES: To evaluate the appropriateness of the clinical data that were selected for Goldberg et al.s 2009 model of cost-effectiveness in multiple sclerosis and calculate results based on a revised cohort selection method for intramuscular (IM) interferon (IFN) beta-1a. METHODS: The original model compared cost per relapse avoided for IM IFN beta-1a, subcutaneous (SC) IFN beta-1a, IFN beta-1b, and glatiramer acetate (GA) based on intent-to-treat (ITT) results from clinical trials. However, due to lower-than-expected subject dropout rates, the IM IFN beta-1a trial had sufficient statistical power to be terminated early and was subsequently found to have met its primary endpoint, time to sustained 1.0-point Expanded Disability Status Scale progression. Within the all-patient (ITT) cohort (n=301), approximately 43% of patients were followed for less than 2 years; 172 patients were followed for 2 years or more. In contrast, the proportions of patients followed for at least 2 years in the clinical trials of IFN beta-1b, SC IFN beta-1a, and GA were 92%, 90%, and 86%, respectively. To test the impact of data selection on the cost-effectiveness model results, we recreated the original model using both the all-patient and 2-year cohorts from the IM IFN beta-1a pivotal trial. We then compared our results with those of the original model. RESULTS: In the original model, costs per relapse avoided were $141,721 for IM IFN beta-1a, $80,589 for SC IFN beta-1a, $87,061 for SC IFN beta-1b, and $88,310 for GA. In the reanalysis using the 2-year completer data for IM IFN beta-1a, costs per relapse avoided were $77,980 for IM IFN beta-1a, $80,121 for SC IFN beta-1a, $86,572 for IFN beta-1b, and $87,767 for GA. The cost per relapse avoided for IM IFN beta-1a was approximately 45% lower than in the original analysis, whereas the recreated results for the other 3 therapies differed from the original results by less than 1%. Sensitivity analyses showed that the recreated model was robust and that the rank order of cost-effectiveness results was unaffected by changes to any univariate parameter. CONCLUSIONS: The current study highlights the importance of data selection in cost-effectiveness analyses. After limiting the pivotal trial data for IM IFN beta-1a to patients followed for at least 2 years, we found that IM IFN beta-1a was more cost-effective than in the original analysis, while results for the other first-line disease-modifying drugs remained stable.