Data extraction and analysis

Clinical outcomes included objective remission rate (ORR), disease control rate (DCR), median progression-free survival (mPFS), median overall survival (mOS) and adverse reactions. All adverse reactions were classified into grades ‘any’ and ‘3–4’. Joint analysis was performed using STATA 14.0 software (StataCorp LP). The heterogeneity of the data was quantified using the I² statistics. I²≥50% was considered to indicate a significant heterogeneity. For I²≥50%, the combined proportion and 95% CI for the outcomes of interest were calculated using the random effects model, whereas the fixed effects model was used for I²<50%. The heterogeneity of results was reduced through sensitivity analysis using a one-by-one elimination method and subgroup analysis according to different trial stages. P<0.05 was considered to indicate a statistically significant difference.