OBJECTIVES

: Observed plateaus in overall survival (OS) for immuno-oncology (I-O) therapies imply a more complex hazard function than can be captured by standard parametric survival models, indicating a subset of patients (e.g. responders) that may achieve long-term survival. Therefore, a response-based modelling approach was conducted to model long-term progression-free survival (PFS) and OS separately for responders and non-responders for the economic evaluation of nivolumab versus relevant comparator therapies (paclitaxel, docetaxel and best supportive care [BSC]) for advanced/metastatic urothelial carcinoma in the UK.

METHODS

: Individual patient-level data from CheckMate 275 and 032 were used to model nivolumab efficacy; comparator efficacy was based on time-varying hazard ratios from a simulated treatment comparison. Parametric distributions were fitted for both responders and non-responders to predict long-term PFS and OS. To overcome immortal time bias, an 8-week landmark point was applied for nivolumab (median time to response in the CheckMate trials). Scenario analyses explored the adoption of standard parametric survival models. Costs included drug acquisition/administration and monitoring costs; utility values were based on EQ-5D data from the CheckMate trials.

RESULTS

: With the response-based modelling approach, incremental cost-effectiveness ratios (ICERs) for nivolumab were £28,263, £23,497 and £24,285 per quality-adjusted life year (QALY) gained versus paclitaxel, docetaxel and BSC, respectively. When adopting a standard parametric survival approach, ICERs were higher at £54,895, £41,195, and £45,451 versus paclitaxel, docetaxel and BSC, respectively.

CONCLUSIONS

: Previous health technology assessment (HTA) appraisals have highlighted that standard survival modelling approaches may not accurately reflect the mechanism of action (MOA) of I-O therapies. Whilst the presentation of response-based models has increased in recent years, UK HTA bodies criticised this approach, opting in favour of standard parametric survival models. Given the ongoing development of therapies with novel MOAs and survival profiles, HTA bodies may need to be open to newer approaches, such as response-based modelling, for future appraisals.