Trial-specific absolute disease-free survival effects of postoperative PD-1/PD-L1 blockade after complete resection of NSCLC: A systematic review.

INTRODUCTION

Postoperative PD-1/PD-L1 blockade after complete resection of non-small-cell lung cancer (NSCLC) is mainly reported using relative disease-free survival (DFS) effects. We estimated absolute effects for distinct completed trial policies.

METHODS

We systematically reviewed randomized phase III trials of postoperative single-agent PD-1/PD-L1 blockade. Published Kaplan-Meier curves were reconstructed to estimate trial-specific 36-month restricted mean DFS time (RMST) and DFS probability differences. Primary intervals used 2,000 arm-stratified bootstrap replicates. Sensitivity analyses examined reconstruction methods and time horizons; overall survival (OS) reconstruction was exploratory. Harms were described by trial, without pooling.

REGISTRATION

PROSPERO CRD420261469971.

RESULTS

Four trial families included 4,336 participants. Thirty-six-month RMST differences ranged from 1.34 to 1.97 months; DFS probability differences corresponded to 2.3-6.1 additional participants without a trial-defined DFS event per 100 randomized. Confidence intervals included zero for two RMST and three DFS probability estimates. The KEYNOTE-091 DFS probability interval crossed zero under alternative reconstruction; RMST interval interpretations were unchanged. The four-trial hazard-ratio synthesis was descriptive and model sensitive. Exploratory 36-month OS reconstructions in three trials yielded intervals allowing benefit and harm. In three trials with available between-arm safety data, severe and serious adverse events were more frequent with checkpoint blockade.

CONCLUSIONS

Reconstructed trial-specific estimates suggest average absolute DFS gains through 36 months, with substantial uncertainty. They complement the original trial results when discussing postoperative treatment; they do not establish a class-wide effect, treatment ranking, or individual benefit.