Evaluation of Classical Clinical Endpoints As Surrogates for Overall Survival in Patients Treated with Immune Checkpoint Blockers: a Systematic Review and Meta-analysis

Purpose: Classical clinical endpoints [e.g., objective response rate (ORR), disease control rate (DCR), and progression-free survival (PFS)] may not be appropriate for immune checkpoint blockers (ICBs). We evaluated correlations between these endpoints and overall survival (OS) for surrogacy.

Methods: Randomized controlled trials (RCTs) of solid tumors patients treated with ICBs published between 01/2005 and 03/2017, and congress proceedings (2014-2016) were included. Arm-level analyses measured 6-month PFS rate to predict 18-month OS rate. Comparison-level analyses measured ORR odds ratio (OR), DCR OR, and 6-month PFS hazard ratio (HR) to predict OS HR. A pooled analysis for single-agent ICBs and ICBs plus chemotherapy vs chemotherapy was conducted. Studies of single-agent ICBs vs chemotherapy were separately analyzed.

Results: 27 RCTs involving 61 treatment arms and 10,300 patients were included. Arm-level analysis showed higher 6- or 9-month PFS rates predicted better 18-month OS rates for ICB arms and/or chemotherapy arms. ICB arms had a higher average OS rate vs chemotherapy for all PFS rates. Comparison-level analysis showed a nonsignificant/weak correlation between ORR OR (adjusted R = - 0.069; P = 0.866) or DCR OR (adjusted R = 0.271; P = 0.107) and OS HR. PFS HR correlated weakly with OS HR in the pooled (adjusted R = 0.366; P = 0.005) and single-agent (adjusted R = 0.452; P = 0.005) ICB studies. Six-month PFS HR was highly predictive of OS HR for single-agent ICBs (adjusted R = 0.907; P < 0.001), but weakly predictive in the pooled analysis (adjusted R = 0.333; P = 0.023).

Conclusions: PFS was an imperfect surrogate for OS. Predictive value of 6-month PFS HR for OS HR in the single-agent ICB analysis requires further exploration.