OBJECTIVES: In clinical trials, crossover from assigned control treatments onto other experimental treatments or alternatives often creates difficulties in estimating the underlying comparative efficacy and CE of experimental treatments. We examined the impact of adjusting crossover from everolimus to nivolumab in the third- or later-line treatment in the phase III CheckMate 025 trial on the estimated relative efficacy and CE of nivolumab in pre-treated patients with aRCC.

METHODS: We employed four crossover adjustment methods on the patient level data from CheckMate 025 (minimum follow-up of 60 months) to account for the 26% (n=106/411) of patients in the everolimus arm switching to nivolumab. An adjusted hazard ratio (HR) was estimated for overall survival (OS), employing the following crossover-adjustment methods: inverse probability of censoring weight, rank-preserving structural failure time model, iterative parameter estimation, and two-stage method. The impact of crossover on the CE of nivolumab versus everolimus was evaluated from a US payer perspective by calculating incremental total costs and quality-adjusted life-years (QALYs) both with and without the crossover adjustments.

RESULTS: The crossover adjustment methods used mostly resulted in lower OS HRs than the reported OS HR from the trial (HR: 0.73), ranging between 0.67-0.74. Adjusted OS was parametrically modelled using survival extrapolations, resulting in small changes in the estimated survival outcomes for everolimus. Consequently, over a lifetime horizon, the impact on total incremental costs and QALYs with nivolumab versus everolimus were both modest, implying minimal changes to the CE estimates.

CONCLUSIONS: The impact of crossover on economic results depends on the choice of the method; the majority of which produced lower OS HRs for nivolumab versus everolimus. However, CE estimates of nivolumab versus everolimus in patients with previously treated aRCC that do not account for crossover are likely to remain stable and slightly conservative.