Data synthesis and statistical analysis

For each safety and tolerability outcome, we calculated the odds ratio (OR) and respective standard error (SE) and/or 95% confidence intervals (CI). ORs were calculated by the Mantel Haenszel method using Review Manager Version 5.3 (The Cochrane Collaboration, Copenhagen, Denmark). Data were pooled only in disease sites where more than one anti-PD-1/PD-L1 antibody was approved. Pair-wise comparisons were performed to explore differential efficacy (as measured by the HR for OS) and safety and tolerability (as measured by the OR for each safety and tolerability measure described above). When data were available from more than one pair-wise comparison, they were pooled using generic inverse variance. Due to the substantial clinical heterogeneity, analyses were performed using random effects modeling irrespective of statistical heterogeneity. Network meta-analysis was performed using WINBUGS within Microsoft Excel (Microsoft Corp, Redmond WA). Statistical tests were two-sided, and statistical significance was defined as p < 0.05. No correction was made for multiple statistical testing.