Keywords: objective response rate, progression free survival, overall survival, surrogate endpoints, PD1/PDL1
Background: ORR and PFS are commonly used surrogates for long term clinical benefit in many disease settings. However, in recently reported immune-oncology studies, the conventional RECIST based outcomes does not appear to be well correlated with OS. The objective of this analysis was to further evaluate the association between ORR, PFS and OS in the PDL1 positive patients and evaluate differences with that in the all-comer population using data from recently reported trials.
Methods: For this analysis, we used 14 published anti PD1/PDL1 mono or combination therapy studies with reported OS, PFS and ORR results. Among these studies, 8 were from NSCLC trials, 2 from bladder trials, and one each from SCLC, melanoma, RCC and SCCHN. 11 out of the 14 studies provided OS, PFS, and ORR results by PD-L1 high or low expression levels. Pearson’s r was used to estimate the correlation between the PFS hazard ratio (HR) and the OS HR; the PFS HR and ?ORR (ORR of experimental arm – ORR of control arm); ?ORR and OS HR. A weighted linear regression model was used to evaluate the association between ?ORR, PFS HR and OS HR, where the weights were chosen based on the trial sizes. The adjusted R2 value was used to estimate the proportion of variance explained by the regression model. These analyses were performed across (i) tumor types regardless of PD-L1 status and (ii) across tumor types but in subgroup of patients who were PD-L1 positive. Definition of PD-L1 positivity was as per the individually reported trials.
Results: The correlation between OS HR and PFS HR was 0.65 (adjusted R2 = 0.58). The correlation between PFS HR and ?ORR was -0.83 (adjusted R2 = 0.70) and between OS HR and ?ORR was -0.51 (adjusted R2 = 0.44). In the PDL1+ subgroup, correlation between OS HR and PFS HR was 0.71 (adjusted R2 = 0.53). The correlation between PFS HR and ?ORR was -0.79 (adjusted R2 = 0.45) and between OS HR and ?ORR was -0.73 (adjusted R2 = 0.37).
Conclusion: A modest trial level correlation was observed between PFS and OS in the overall patient population. However, a stronger positive association was observed in the PD-L1 positive subgroup. No meaningful difference in association between ORR and PFS or ORR and OS was observed between the two populations, with a weak correlation of ORR with OS. These results indicate the need for further evaluating the relevance of biomarkers, such as PD-L1 expression, in understanding the relationship between PFS and ORR with OS.