Extended Data Fig. 1: Treatment difference in objective response in subgroups of the efficacy population.

Response was assessed per RECIST, version 1.1, by blinded, independent central review. The estimated treatment difference between the pembrolizumab and placebo groups in the overall population was calculated using the Miettinen and Nurminen method stratified by geographic region (Australia/Europe/Israel/North America [Aus/Eur/Isr/NAm] vs Asia vs rest of world), PD-L1 combined positive score ([CPS]; ≥1 vs <1), and chemotherapy choice (5-fluorouracil plus cisplatin [FP] vs capecitabine plus oxaliplatin [CAPOX]); differences in subgroups were calculated using the unstratified Miettinen and Nurminen method. The efficacy population included the first 264 participants randomly allocated to treatment. The treatment regimen included trastuzumab and chemotherapy in both groups. Diamonds represent the estimated treatment difference in objective response, the error bars represent the 95% confidence interval (CI) of the estimated treatment difference, and the region shaded in dark grey represents the 95% CI for the treatment difference in the overall population. ECOG, Eastern Cooperative Oncology Group; GEJ, gastro-oesophageal junction; IHC, immunohistochemistry; ISH, in situ hybridization.