Analysis of early mortality in randomized clinical trials evaluating anti-PD-1/PD-L1 antibodies: A systematic analysis by the United States Food and Drug Administration (FDA).

Abstract

2516 Background: Many studies exhibit what seems to be dis-proportionately higher early mortality (EM) in anti-PD-1/PD-L1 containing arms (IO) when compared to active control arms (AC), resulting in early crossing of the Kaplan-Meier overall survival curves. We examine if EM with the use of IO is specific to certain demographic and disease characteristics. Methods: Data from 16 randomized AC trials submitted to FDA containing 6055 IO and 3604 AC patients in HNSCC, Melanoma, NSCLC, RCC, and Urothelial Carcinoma were evaluated for signs of EM. Study-specific and pooled piecewise hazard ratios (HRs) were used to quantify EM from 0 to 60 and > 60 days. Additionally, HRs up to 60 days were used to assess the extent specific subgroups account for EM. Results: Piecewise HRs comparing OS between IO and AC changed direction, > 1 to < 1 in 11 trials; melanoma (5/6), NSCLC (3/7), HNSCC (1/1), RCC (1/1), and urothelial cancer (1/1). When pooled, NSCLC studies retained this EM pattern, although attenuated, with HR (95% CI) of 1.12 (0.91, 1.38) ≤60 days and 0.66 (0.61, 0.72) after 60 days. This was not observed in the pooled melanoma studies: 0.88 (0.63, 1.24) ≤ 60 days and 0.59 (0.53, 0.67) after 60 days. EM in both arms was associated with poor ECOG performance status (PS), increased LDH, and high tumor burden. Comparing EM patients in the IO and AC arms, a larger proportion were female in the melanoma trials (41% vs. 28%), a smaller proportion had squamous histology in the NSCLC trials (32% vs. 41%), and a larger proportion were PD-L1 negative (56% vs. 36% melanoma; 60% vs. 43% NSCLC). Analysis of the pooled melanoma studies suggests PD-L1 negative melanoma patients with high baseline tumor burden and PS played a role in EM with HR before 60 days of 1.49 (0.75, 2.97). However, these results were not reproducible in NSCLC. Conclusions: Potential risk factors for EM were assessed in individual and pooled trials. While several factors—negative PD-1/PD-L1 status and high ECOG, LDH and tumor burden—seem to play a role in EM, these high-risk subgroups do not fully explain the EM patterns observed in the IO treated patients.