Background: We performed a meta‐analysis to quantify toxic death, adverse events (AEs) and treatment discontinuation due to AEs from checkpoint inhibitors (CI).

Methods: We searched for randomized trials with adequate reporting for toxicity outcomes. Pooled risk ratios were estimated for CI versus chemotherapy or different combinations of these agents.

Results: Twenty trials of five different cancers with 10 794 patients with performance status 0 or 1 were identified. Toxic deaths from CI were infrequent (0.6%). Treatment discontinuations were less frequent for programmed‐death‐1 (PD‐1) or PD‐ligand‐1 (PD‐L1) inhibitors (5.8% vs 13.3%, P < 0.001) and cytotoxic‐T‐lymphocyte‐associated‐antigen‐4 (CTLA‐4) inhibitors (6.2% vs 11.4%, P = 0.002) than chemotherapy. PD‐1/PD‐L1 inhibitors had less grade 3, 4, and 5 (G3/4/5) AEs than chemotherapy (13.8% vs 39.8%, P < 0.001) or CTLA‐4 inhibitors (13.4% vs 22.8%, P < 0.001). Combination CI had higher discontinuation (37.8% vs 11.6%, P < 0.001) and higher G3/4/5 AEs (55.3% vs 21.9%, P < 0.001) than CI monotherapy. Endocrinopathy (11.2% vs 0.9%), rash (10.1% vs 4.3%) and pneumonitis (3.1% vs 0.7%) were associated with CI, and alopecia (25.9% vs 1.0%), neutropenia (16.6% vs 0.6%) and neuropathy (7.6% vs 3.0%) with chemotherapy.

Conclusions: CI inhibitors have different toxicity profiles to chemotherapy. These findings are useful for patient counselling and planning of future trials.