A thorough range of univariate sensitivity analyses for each variable in the model was conducted. In the comparison with warfarin, the results were moderately sensitive to patient age and the cost of edoxaban. However, the interpretation of the results did not change in any scenario, since the highest reported incremental cost-utility ratio across the analyses was $30,658 per QALY. In the comparison with rivaroxaban, the results were moderately sensitive to the costs of edoxaban and rivaroxaban, but again the interpretation of the results did not change in any scenario; edoxaban remained dominant in most scenarios, and the highest reported incremental cost-utility ratio across the analyses was $7,256 per QALY.

A further range of scenarios was considered relating to the time horizon, the discount rate, and the patient population. In these analyses, for the comparison with warfarin, the results were moderately sensitive to the time horizon of the model (10 years) and the time in therapeutic range for warfarin (> 60%). However, the interpretation of the results did not change in any scenario. In the comparison with rivaroxaban, the results were not sensitive to any changes.

Within the probabilistic sensitivity analysis, at a threshold of $50,000 per QALY, there was a 96% probability that edoxaban was cost-effective compared with warfarin and a 92% probability that edoxaban was cost-effective compared with rivaroxaban.