Purpose: The efficacy of several novel combinations of anti-programmed cell death protein 1 or its ligand antibodies with chemotherapy, which have become the new standard first-line combination therapy with favorable outcomes, was still not certain in patients with different combined positive score (CPS) grades. This research aimed to evaluate the efficacy of immune checkpoint inhibitor immunotherapy or immunochemotherapy at different CPS grades, compared with chemotherapy.

Methods: Kaplan-Meier (KM) curve reconstruction was employed to assess the overall survival (OS) and progression-free survival (PFS) of patients with gastric cancer. The graphical reconstruction algorithm was used to estimate the time-to-event outcomes from Kaplan-Meier curves of the overall cohort or reported subgroups (depending on CPS). KMSubtraction was used to derive the unreported survival data by matching participants in the overall cohort and known subgroups.

Findings: This analysis included 5072 patients in 5 trials (CheckMate 649, KEYNOTE-859, ORIENT-16, KEYNOTE-062, and JAVELIN Gastric 100). Immunochemotherapy exhibited more effectiveness than chemotherapy in most cases. For the overall cohort, sintilimab + chemotherapy exhibited the best effect in OS (hazard ratio [HR], 0.65; 95% CI, 0.55–0.76). Nivolumab + chemotherapy (HR, 0.75; 95% CI, 0.67–0.84), sintilimab + chemotherapy (HR, 0.52; 95% CI, 0.41–0.65), and pembrolizumab + chemotherapy (HR, 0.68; 95% CI, 0.58–0.81) exhibited favorable outcomes in OS in patients with a CPS ≥1, 5, and 10, respectively, and similarly in PFS. Avelumab + chemotherapy performed similarly to chemotherapy in OS but had poor PFS in the reported subgroup.

Implications: Finding suggests that immune checkpoint inhibitors combined with chemotherapy could enrich patients with benefits regardless of CPS grades, though subtle efficacy in low CPS subgroups. This study compared the efficacy of different immunotherapies combined with chemotherapy in patients with gastric cancer, but we acknowledge some differences between reconstructed and original data. Hopefully there will be more research investigating comparisons between current therapies rather than with chemotherapy only in the future.