OBJECTIVES:

Regularly several models are developed to estimate the cost-effectiveness of the same drug, but the models are based on different underlying assumptions. The aim of this study was to evaluate the influence of assumptions that derive the model structure and input data on the case of cost-effectiveness analysis of dabigatran compared to warfarin.

METHODS:

Different cost-effectiveness analyses based on three models were compared. One was previously developed (Janzic et al., Pharmacoeconomics, 2015, 33(4):395-408) in house and two (Freeman et al., Ann. Intern. Med., 2011, 154(1):1-11 and Sorensen et al., Thromb. Haemost., 2011, 105(5):908-19) were re-build as far as possible based on the published data. A step wise approach was used to test the assumptions, adjust the model structures and unify the input data. The models outputs (total cost and QALYs) were assessed in each step.

RESULTS:

Additional assumptions were necessary during rebuilding the two models based. Up to 6% difference in estimates of QALYs and up to 42% difference in estimated costs between published and our simulated results were observed. At the baseline the results among the three models varied significantly, the difference in QALY was almost 40%, while the differences in total costs were more than 10-fold. When unifying the input data, the highest impact had cost data, especially costs of events, and mortality tables. According to the assumptions underlying model structure, the most important were the number of health states and their definition, clinical events considered and treatment strategy after discontinuation. Other assumptions, such as age dependent adjustment for bleedings, had minor effect.

CONCLUSIONS:

The assumptions underlying model structure in addition to input data significantly affected the results. More emphasis should be focused on critical evaluation of the model assumptions.