Five RCTs (n=979 calculated, range 40 to 536 participants) were included in the meta-analysis. There was some discrepancy between the figures reported in the text and tables and we have reported the figures from the tables. Three RCTs scored 5 on the quality score and two scored 4.
Second-look endoscopy with injections (three RCTs) did not significantly reduce re-bleeding compared to single endoscopy (RR 0.85, 95% CI 0.63 to 1.14). There were no significant differences in blood transfusion (mean difference -0.15, 95% CI -1.42 to 1.13), requirement of surgical intervention (RR 0.69, 95% CI 0.37 to 1.28) and all-cause mortality (RR 0.83, 95% CI 0.44 to 1.60). The authors stated that there was significant statistical heterogeneity among trials that assessed unit of blood transfusion, but the statistical value reported suggested otherwise (p>0.1). Sensitivity analyses did not significantly alter the results. Subgroup analyses showed that neither epinephrine nor fibrin glue injection significantly reduced re-bleeding.
Second-look endoscopy with thermal coagulation (two RCTs) showed a statistically significant reduction in re-bleeding compared with single endoscopy (RR 0.29, 95% CI 0.11 to 0.73), but not in all cause mortality (RR 0.75, 95% CI 0.17 to 3.30). One RCT reported no statistically significant difference in blood transfusion and the requirement of surgical intervention.