Methods: A semi-Markov model was used, which was implemented in TreeAge Software. Data from the randomised study (n=448) and literature were used. The reference case was a man aged 65, and the time horizon was two years. A societal perspective was adopted. Sensitivity analyses were performed to evaluate the effects of varying costs and effects.

Results: The sensitivity of MA was 0.56 (cystoscopy 0.95), which was too low to compensate for the ability to detect recurrent tumors in the upper urinary tract (which are missed by cystoscopy). The MA test also had worse specificity than cystoscopy (0.74 vs 0.97). The control arm hence led to better outcomes than the test arm, and lower costs (€2,691 vs €3,785 per capita over the course of two years). Although the Markov model indicated that the control arm dominated the test arm, patient preferences for the urinary test may be strong, accepting a slightly worse medical outcome.

Conclusions: Over the course of two years, surveillance in which cystoscopy is partly replaced by current MA does not provide a cost-effective alternative to the conventional surveillance. The search for a surveillance test, combining similar or better diagnostic accuracy than cystoscopy with lower patient burden, needs to be continued.