In total, 1581 patients with previously untreated advanced GC/GEJC/EAC were randomized to receive nivolumab plus chemotherapy or chemotherapy in the CheckMate 649 study; of these patients, 1297 had PD-L1 CPS ≥ 1 and 955 had PD-L1 CPS ≥ 5. With 48-month minimum follow-up, nivolumab plus chemotherapy continued to demonstrate OS benefit versus chemotherapy in all randomized patients (13.7 vs. 11.6 months; HR, 0.79 [95% CI, 0.71–0.88]), patients with PD-L1 CPS ≥ 1 (13.8 vs. 11.4 months; HR, 0.75 [95% CI, 0.67–0.85]), and patients with PD-L1 CPS ≥ 5 (14.4 vs. 11.1 months; HR, 0.70 [95% CI, 0.61–0.81]) (Supplementary Fig. 1). PFS benefit with nivolumab plus chemotherapy versus chemotherapy was also maintained in all randomized patients (7.7 vs. 6.9 months; HR, 0.80 [95% CI, 0.71–0.89]), patients with PD-L1 CPS ≥ 1 (7.5 vs. 6.9 months; HR, 0.77 [95% CI, 0.68–0.88]), and patients with PD-L1 CPS ≥ 5 (8.3 vs. 6.1 months; HR, 0.71 [95% CI, 0.61–0.82]) (Supplementary Fig. 2). Grade 3 or 4 treatment-related adverse events occurred in 60% and 45% of patients in the nivolumab plus chemotherapy and chemotherapy arms, respectively.

With 4 years of minimum follow-up, patients treated with nivolumab plus chemotherapy had longer mean durations of OS, PFS per BICR, TOX, TWiST, and PROG compared with patients who received chemotherapy (Table 1), and the mean differences were statistically significant for all health states except PROG across all evaluated PD-L1 CPS expression levels. In all randomized patients, the mean duration of Q-TWiST was 16.9 months with nivolumab plus chemotherapy versus 13.6 months with chemotherapy, corresponding to an absolute Q-TWiST gain of 3.4 months (Table 1) and relative gain of 20.5%. The TOX, TWiST, and PROG partition survival curves for all randomized patients are shown in Supplementary Fig. 3.

Patients with PD-L1 CPS ≥ 1 had a mean Q-TWiST duration of 17.4 months with nivolumab plus chemotherapy versus 13.2 months with chemotherapy, leading to a higher absolute Q-TWiST gain of 4.2 months (Table 1) and higher relative gain of 26.1% compared with all randomized patients.

Furthermore, in patients with PD-L1 CPS ≥ 5, the mean Q-TWiST duration with nivolumab plus chemotherapy was 18.7 months versus 13.3 months with chemotherapy (Table 1), resulting in the highest absolute and relative Q-TWiST gains (5.4 months and 33.4%, respectively).

The Q-TWiST benefit with nivolumab plus chemotherapy versus chemotherapy in all randomized patients and in patients with PD-L1 CPS ≥ 1 was observed in most prespecified subgroups (Fig. 1). Nivolumab plus chemotherapy demonstrated Q-TWiST benefit versus chemotherapy alone across all subgroups of patients with PD-L1 CPS ≥ 5 (Fig. 1). Across all randomized patients, patients with PD-L1 CPS ≥ 1, and patients with PD-L1 CPS ≥ 5, the greatest relative Q-TWiST gains were observed in the microsatellite instability-high subgroup (140.2%, 180.0%, and 200.1%, respectively).

Base-case Q-TWiST subgroup analysis. a All randomized patients. b Patients with PD-L1 CPS ≥ 1. c. Patients with PD-L1 CPS ≥ 5. Chemo chemotherapy, CI confidence interval, CPS combined positive score, EAC esophageal adenocarcinoma, ECOG PS Eastern Cooperative Oncology Group performance status, FOLFOX leucovorin fluorouracil and oxaliplatin, GC gastric cancer, GEJC gastroesophageal junction cancer, MSI microsatellite instability, MSI-H microsatellite instability-high, MSS microsatellite stable, NIVO nivolumab, PD-L1 programmed death ligand 1, Q-TWiST quality-adjusted time without symptoms or toxicity, XELOX capecitabine and oxaliplatin

Sensitivity analysis at follow-up times from 3 to 66 months showed an increase in the relative Q-TWiST gain with longer follow-up time for all randomized patients, patients with PD-L1 CPS ≥ 1, and patients with PD-L1 CPS ≥ 5 (Fig. 2). All randomized patients had a clinically important relative Q-TWiST gain after 26 months and clearly clinically important gain after 44 months. Clinically important and clearly clinically important relative Q-TWiST gains were seen earlier in patients with PD-L1 CPS ≥ 1 (22 and 34 months, respectively). Patients with PD-L1 CPS ≥ 5 had the earliest relative Q-TWiST gains with a clinically important gain achieved after 14 months and a clearly clinically important gain after 21 months. Among all randomized patients, the duration of TWiST showed continued increase with nivolumab plus chemotherapy versus chemotherapy with longer follow-up while the duration of TOX remained stable (Supplementary Fig. 3).

Q-TWiST sensitivity analysis. a All randomized patients. b Patients with PD-L1 CPS ≥ 1. c Patients with PD-L1 CPS ≥ 5. CPS combined positive score, PD-L1 programmed death ligand 1, Q-TWiST quality-adjusted time without symptoms or toxicity

In the second sensitivity analysis, the absolute Q-TWiST gains with nivolumab plus chemotherapy decreased slightly with the inclusion of grade 2 adverse events lasting ≥ 28 days (2.6, 3.3, and 4.3 months, in all randomized patients, patients with PD-L1 CPS ≥ 1, and patients with PD-L1 CPS ≥ 5, respectively). Nonetheless, relative Q-TWiST gains remained clearly clinically important (15.7%, 20.3%, and 26.4%, respectively).

In the exploratory analysis, Q-TWiST gain was estimated using the pre-calculated UK EQ-5D-3L utility values from CheckMate 649. The UK EQ-5D-3L utility values for nivolumab plus chemotherapy were higher compared with chemotherapy across the three cohorts, indicating a better QoL during each health state. The absolute Q-TWiST gains for nivolumab plus chemotherapy in all randomized patients, patients with PD-L1 CPS ≥ 1, and patients with PD-L1 CPS ≥ 5 were 3.8, 4.6, and 6.1 months, respectively, with corresponding relative Q-TWiST gains of 23.0%, 28.4%, and 37.6% (Table 2).