Efficacy and safety of proton pump inhibitors for stress ulcer prophylaxis in critically ill patients: a systematic review and meta-analysis of randomized trials

Table 2 Grading of Recommendation, Assessment, Development, and Evaluation (GRADE) evidence profile

Quality assessment						Patients, number		Effect		Quality
Studies, number	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	PPIs	H2RAs	Relative (95 % CI)	Absolute (95 % CI)	Quality
Clinically important bleeding
14	Serious^a	Not serious	Not serious	Not serious^b	None	13/986 (1.3 %)	39/693 (5.6 %)	RR 0.39 (0.21, 0.71)	15 fewer per 1000 (7–20 fewer)	Moderate^a,b
Overt upper gastrointestinal bleeding
17	Serious^a	Not serious^c	Not serious	Not serious	None	53/1102 (4.8 %)	118/795 (14.8 %)	RR 0.48 (0.34, 0.66)	26 fewer per 1000 (17–33 fewer)	Moderate^a,c
Nosocomial pneumonia
13	Serious^a	Not serious^d	Not serious	Serious^e	None	119/862 (13.8 %)	92/709 (13.0 %)	RR 1.12 (0.86, 1.46)	16 more per 1000 (18 fewer to 60 more)	Low^a,d,e
Mortality
11	Not serious^f	Not serious	Not serious	Serious^e	None	151/874 (17.3 %)	120/614 (19.5 %)	RR 1.05 (0.87, 1.27)	10 more per 1000 (25 fewer to 53 more)	Moderate^e,f
ICU length of stay
7	Serious^g	Not serious	Not serious	Serious^h	None	371	373	-	MD 0.58 days fewer (2.03 fewer to 0.86 more)	Low^g,h

The Guideline Development Tool was used to summarize the quality of evidence for individual outcomes based on five main domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias. PPI proton pump inhibitor, H2RA histamine-2-receptor antagonist, MD mean difference, RR relative risk. ^aWe downgraded by one level, for risk of bias; most studies were unblinded. ^bAlthough the total number of events was small, we did not downgrade for imprecision. ^cSignificant inconsistency was not present (I ² = 6 %). ^dSignificant inconsistency was not present (I ² = 4 %). ^eWe downgraded by one level for imprecision; the confidence interval contains significant benefit and harm. ^fWe did not downgrade for risk of bias because mortality is an objective outcome that is less likely to be affected by lack of blinding in clinical trials. ^gWe downgraded by one level for risk of bias. ^hWe downgraded by one level for imprecision; the confidence interval contained significant benefit and harm

ISSN: 1364-8535