Extended Data Fig. 1: CONSORT diagram for patient disposition.

^aIncluded death (n = 36), adverse events (n = 24), poor/noncompliance (n = 15), administrative reasons (n = 5), pregnancy (n = 1) and additional reasons (n = 43); ^bIncludes patients concurrently randomized to the nivolumab plus chemotherapy, nivolumab plus ipilimumab, and chemotherapy groups. Relevant protocol deviations were noted in 21 (1%) patients concurrently randomized to nivolumab plus chemotherapy versus chemotherapy: usage of prohibited on-treatment anti-cancer therapy (n = 12), baseline ECOG PS >1 (n = 5), incorrect cancer diagnosis (n = 2), prohibited prior anti-cancer therapy (at study entry) (n = 1) and no baseline (measurable or evaluable) disease (n = 1); Relevant protocol deviations were noted in 10 (1%) patients concurrently randomized to nivolumab plus ipilimumab versus chemotherapy: usage of prohibited on-treatment anti-cancer therapy (n = 5), incorrect cancer diagnosis (n = 2), no baseline PD-L1 result (n = 2) and baseline ECOG PS > 1 (n = 1); ^c363 patients overlapped between the two chemotherapy groups from the 1:1:1 randomization period; ^dThe median follow-up for survival (time from concurrent randomization to last known date alive or death) was 13.1 months (range 0.1–49.5) and 11.2 months (range 0.0–47.9) in the nivolumab plus chemotherapy versus chemotherapy groups, respectively, and 11.4 months (range 0.0–52.1) and 11.5 months (range 0.0–52.8) in the nivolumab plus ipilimumab versus chemotherapy groups, respectively; ^eIncluded adverse events unrelated to study treatment (n = 47), maximum clinical benefit (n = 11), lost to follow-up (n = 2), patient no longer met trial criteria (n = 1), poor/noncompliance (n = 1) and other reasons (n = 9); ^fIncluded adverse events unrelated to study treatment (n = 35), maximum clinical benefit (n = 30), poor/noncompliance (n = 4), other reasons (n = 8), lost to follow-up (n = 2) and death (n = 1); ^gIncluded adverse events unrelated to study treatment (n = 21), death (n = 3), poor/noncompliance (n = 1), other reasons (n = 6) and not reported (n = 5); ^hIncluded adverse events unrelated to study treatment (n = 15), maximum clinical benefit (n = 13), lost to follow-up (n = 2), poor/noncompliance (n = 2) and other reasons (n = 4). ECOG PS, Eastern Cooperative Oncology Group performance status; OS, overall survival; PD-L1, programmed death ligand 1; PFS, progression-free survival.