Background: Immunotherapy, especially immune-checkpoint inhibitors (PD-1 and PD-L1 inhibitors), is now one of the mainstays of cancer treatment. Several studies have analyzed treatment-related toxicities of immunotherapy. However, small sample size, rough and unspecific stratification, and lack of comparison (pure sing-arm studies) are common limitations. Detailed organ- and system-specific toxicities remain not clear enough. Methods: We systematically searched PubMed, Embase, Cochrane Library, Web of Science, and others (CNKI), from database inception to Mar 31, 2020, for randomized controlled trials (RCTs) related to PD-1/PD-L1 inhibitors that had available toxicity data. We excluded non-randomized trials. The primary endpoint was to assess the difference in the incidences of toxicities between cancer patients who did and did not receive PD-1/PD-L1 inhibitors. We calculated the pooled relative risks (RRs) and corresponding 95% confidence intervals (95% CIs) using a random-effects model and assessed the heterogeneity between different groups. The subgroup analyses were conducted based on toxicity grade (severity), system and organ, treatment regimens in the intervention arm and control arm, PD-1/PD-L1 inhibitor drug type, and cancer histotype. We applied the five-point Jadad ranking system to evaluate the quality of the selected studies. We performed multivariate meta-regression analyses to explore the proportion of between-study variance. Results: A total of 29 eligible RCTs including 8067 patients were selected for the meta-analysis based on specified inclusion and exclusion criteria. Patients treated with PD-1/PD-L1 inhibitors were at lower risks of overall toxicities (all grades: RR 0.91, 95% CI 0.89-0.92; grade 3~4: RR 0.76, 95% CI 0.74-0.78), including gastrointestinal toxicity (all grades: RR 0.68, 95% CI 0.60-0.77; grade 3~4: RR 0.71, 95% CI 0.43-1.20), hematologic toxicity (all grades: RR 0.66, 95% CI 0.51-0.85; grade 3~4: RR 0.55, 95% CI 0.37-0.83), and treatment event leading to discontinuation (all grades: RR 0.78, 95% CI 0.72-0.84; grade 3~4: RR 0.58, 95% CI 0.49-0.67); but were at higher risks for respiratory toxicity (all grades: RR 1.74, 95% CI 1.33-2.28; grade 3~4: RR 1.92, 95% CI 1.45-2.55) and endocrine toxicity (all grades: RR 1.70, 95% CI 0.62-4.69; grade 3~4: RR 1.29, 95% CI 0.45-3.69). The subgroup analyses indicated that when compared with the control, toxicity comparison tendency for PD-1/PD-L1 inhibitors varied with the toxicity grade, affected system and organ, treatment regimens in the intervention arm and control arm, drug type, and cancer histotype. The male-female ratio was a statistically significant variable in the Meta-Regression analysis (I2=89.1,τ2=0.01, and P=0.001). Conclusion: For most toxicity types based on system and organ, the incidence proportions for patients in the intervention arm were lower than those in the control arm, which suggested the general safety of PD-1/PD-L1 inhibitors against conventional chemotherapy and CTLA-4 inhibitors. However, for some specific toxicities including respiratory, cutaneous, and endocrine toxicities, the case was the opposite. The toxicity grade, system and organ, treatment regimens, drug type, and cancer histotype were all influencing factors. To our knowledge, this was by far the most comprehensive meta-analysis of RCTs on toxicities of immune-checkpoint inhibitors. Future research should focus on taking effective targeted measures to decrease the risks of different toxicities for different patient populations.