Thirty-five RCTs (n=4,843) were included.
Monotherapy with PPIs versus monotherapy with placebo (3 RCTs, 1,045 patients, 29% with active bleeding): the studies were a high quality (Jadad score 5). There was a reduced risk of re-bleeding with PPIs compared with placebo, but statistical heterogeneity was found; the RD was 13.7% (95% CI: 0.9, 27), the OR was 0.50 (95% CI: 0.26, 0.96), the NNTB was 7 (95% CI: 4, 114) and the fail-safe N was 8. One RCT was an outlier on the Galbraith plot. Only one RCT reported results by bleeding stigmatas; it showed RDs varying from 20% in Forrest 1A to 48% in Forrest 2A. There was no statistically significant difference between treatments for surgery and mortality.
Monotherapy with PPIs versus monotherapy with H2RAs (8 RCTs, 816 patients, 23% with active bleeding): the studies were generally of a poor quality (4 scored 2 points on the Jadad scale). There was a reduced risk of re-bleeding with PPIs compared with H2RAs, but statistical heterogeneity was found; the RD was 20% (95% CI: 7, 33), the OR was 0.31 (95% CI: 0.15, 0.66), the NNTB was 5 (95% CI: 3, 15) and the fail-safe N was 20. Two RCTs were outliers on the Galbraith plot. There was a reduced risk of surgery with PPIs, (OR 0.53, 95% CI: 0.29, 0.98) but not mortality. No statistically significant heterogeneity was found.
Monotherapy with PPIs versus endotherapy plus H2RAs (4 RCTs, 357 patients, none had active bleeding): the studies were generally of a poor quality (mean Jadad score 2.75). There was no statistically significant difference between PPI and endotherapy plus H2RA in the risk of re-bleeding (OR 1.30, 95% CI: 0.7, 2.17). No statistically significant heterogeneity was found.
Monotherapy with PPIs versus endotherapy plus PPI (5 RCTs, 816 patients, 91% with nonbleeding): the studies were of a high quality (Jadad score 4 or 5). There was a reduced risk of re- bleeding with PPIs compared with H2RA, (OR 0.19, 95% CI: 0.09, 0.37; NNTB 6, 95% CI: 4, 10). No statistically significant heterogeneity was found. There was a reduced risk of surgery with PPIs compared with H2RAs (OR 0.17, 95% CI: 0.06, 0.45; NNTB 505, 95% CI: 38, -45). There was no significant difference between treatments for mortality.
PPIs plus endotherapy versus endoscopic monotherapy (5 RCTs, 986 patients, 46% with bleeding lesions): the studies were of a high quality (mean Jadad score 4.4). There was a reduced risk of re-bleeding with PPIs plus endotherapy compared with endotherapy alone (OR 0.51, 95% CI: 0.37, 0.71). No statistically significant heterogeneity was found. There was no significant difference between treatments for surgery or mortality.
PPIs plus endotherapy versus H2RAs plus endotherapy (8 RCTs, 947 patients, 42% with active bleeding): the studies were generally of a poor quality (mean Jadad score 2.7). There was no statistically significant difference between treatments for rebleeeding, surgery or mortality. No statistically significant heterogeneity was found. The subgroup analysis showed a significantly reduced risk of re-bleeding with PPI plus endotherapy in nonbleeding patients; the OR (2 RCTs) was 0.24 (95% CI: 0.07, 0.76) and the NNTB was 5 (95% CI: 3, 24).
Infusion PPIs versus bolus PPIs used in addition to endotherapy (2 high-quality RCTs, 310 patients): there was no statistically significant difference between treatments for re-bleeding or mortality, and the data were homogeneous (OR 1.21, 95% CI: 0.59, 2.52).
Details of the subgroup analyses were also presented in the paper.