Bayesian Evaluation of Treatment Effect of Avelumab Plus Axitinib for Advanced Renal Cell Carcinoma

Despite not achieving statistical significance, the JAVELIN Renal 101 trial indicated a potentially clinically relevant effect size (hazard ratio [HR], 0.88; 95% confidence interval, 0.75 to 1.04) on overall survival (OS) favoring avelumab plus axitinib over sunitinib for advanced renal cell carcinoma (aRCC). To better interpret these findings, we performed a Bayesian evaluation.

We reanalyzed individual participant data from the JAVELIN Renal 101 trial using a Bayesian framework. Prior distributions were specified for varying levels of enthusiasm and skepticism for avelumab plus axitinib, and a prior from a meta-analysis of similar trials. The posterior probability of any OS benefit with avelumab plus axitinib, Pr(HR < 1.0), and of benefit exceeding the minimal clinically important difference (MCID; HR = 0.75), Pr(HR < MCID), were estimated using Bayesian Cox models. Bayesian hierarchical models evaluated treatment effect heterogeneity across International Metastatic Renal Cell Carcinoma Database Consortium risk groups and programmed death-ligand 1 (PD-L1) status.

Regardless of the prior assumptions used, Pr(HR < 1.0) was consistently high (93.0% to 95.5%). However, Pr(HR < MCID) was low (1.0% to 3.8%). Results were similar when the prior was informed by pooled data from similar trials. In subgroup analyses, Pr(HR < MCID) varied by risk group: 63.4% for poor-risk, 34.9% for favorable-risk, and 0.9% for intermediate-risk aRCC. This probability was low regardless of PD-L1 status.

This exploratory Bayesian reanalysis complements the interpretation of the JAVELIN Renal 101 trial and offers a probabilistic perspective beyond a dichotomous (i.e., significant/nonsignificant) interpretation.