OBJECTIVES : To determine the comparative effectiveness of 1L aRCC treatments utilized in France. In a Bayesian framework, a multi-dimensional treatment effect network meta-analysis (NMA) was conducted to account for time-varying hazards ratios.

METHODS : CheckMate 9ER, CheckMate 214, COMPARZ, and KEYNOTE-426 formed a connected network of randomized controlled trials, and KM curves were evaluated. Sunitinib was selected as the anchor treatment as it was the common comparator. Pseudo-individual patient data (IPD) was reconstructed using Guyot’s algorithm when IPD was unavailable. Outcomes assessed were progression-free survival (PFS) and overall survival (OS). Fixed effect first and second order fractional polynomials (FP; 44 models in total) were fitted to data for each outcome after which goodness of fit was assessed via statistical fit criteria and clinical plausibility of the time-varying hazard ratios. The predictive accuracy of survival extrapolations was validated with observed median survival, restricted mean survival time and landmark survival at 24 months.

RESULTS : For PFS, 35 models converged, with the second order model (P1=-1, P2=-2) considered the most viable based on clinical plausibility of the hazards. Model extrapolations showed nivolumab+ipilimumab to be the most effective therapy for PFS from month 70 to 15 years. For OS, 29 models converged, and the first order model (P1=-1) was considered the most viable. Over the time horizon, nivolumab+ipilimumab was the most effective therapy, with the highest OS estimations after 50 months, followed by nivolumab+cabozantinib, pembrolizumab+axitinib, pazopanib, and sunitinib.

CONCLUSIONS : Due to the flexible functional form of FP models, clinical plausibility should be considered in model selection in addition to statistical fit criteria. NMAs with multidimensional treatment effects offer a sophisticated alternative for evidence synthesis. Selected models provided accurate estimates for PFS and OS survival extrapolations relative to empirical data, with nivolumab+ipilimumab estimated as the most effective 1L aRCC treatment over a 15-year horizon in France for both outcomes.