Purpose: To discuss the Canadian Common Drug Review (CDR) experience with economic models submitted by pharmaceutical manufacturers and the key limitations that have been identified by CDR reviewers.
Methods: A review of pharmacoeconomic submissions completed by CDR in 2004 was undertaken. Submissions that included economic models were identified. The authors reviewed these submissions and noted the key concerns with the models, paying attention to the appropriateness of the model structure, data sources used, analyses conducted, and transparency of methods. A summary of the key issues identified by the reviewers were summarized.
Results: 19 reviews had been completed by the CDR in 2004, 10 of which include an analysis using an economic model. 4 used complex Markov models and 6 were based on decision analytic models. The key concerns identified were the data sources used to populate the models. In 8 of the 10 submissions, treatment effect was not well modeled (model used benefits not seen in clinical trials, poorly conducted meta-analyses or indirect comparisons). Of the 7 models that did conduct sensitivity analyses, only 1 was deemed to be sufficient by the reviewers. In 4 of the 10 submissions, the methods used were not considered transparent by the reviewers. In 3 of the submissions, computational errors were identified, i.e., reviewers were unable to replicate the results. As a result, in 6 of the 10 submissions, the results of the economic evaluations on their own (without further analysis by CDR or modification to input parameters) were not sufficiently robust to inform recommendations on drug reimbursement.
Conclusions: New modeling techniques have the potential to facilitate better review of the economic models by the CDR and provide more accurate estimates of cost-effectiveness. However, the quality of the data used in these models, which is a key component of the validity of the results, is often lacking.