Intended for healthcare professionals

Practice Rapid Recommendations

Gastrointestinal bleeding prophylaxis for critically ill patients: a clinical practice guideline

BMJ 2020; 368 doi: https://doi.org/10.1136/bmj.l6722 (Published 06 January 2020) Cite this as: BMJ 2020;368:l6722

Visual summary of recommendation

Population Critically ill patients Does not apply to: Including: Patients receiving gastricacid suppression for another therapeutic indication Patients admitted to intensive care units On average, 4% of critically ill patients develop gastrointestinal bleeding. One cause is physiologic stress leading to stress ulcers in the oesophagus, stomach, or duodenum, but critical illness is also associated with other forms of upper gastrointestinal bleeding.
Recommendation 1 We suggest using acid suppression prophylaxis for people with higher risk of gastrointestinal bleeding (4% or higher) No prophylaxis Prophylaxis or Strong All or nearly all informed people would likely want this option. Benefits outweigh harms for almost everyone. Weak Most people would likely want this option. Benefits outweigh harms for the majority, but not for everyone. Weak Most people would likely want this option. Benefits outweigh harms for the majority, but not for everyone. Strong All or nearly all informed people would likely want this option. Benefits outweigh harms for almost everyone.

Calculating bleed risk

Mechanical ventilation without enteral nutrition 2 or more factors from 2-4% category Mechanical ventilation with enteral nutrition Critically ill patients without any risk factor Cancer Male gender Acute hepatic failure Use of steroids or immunosuppression Acute kidney injury Sepsis Highest risk High risk SUGGESTED CUT POINT FOR OFFERING PROPHYLAXIS 8-10% 4-8% Moderate risk Low risk 2-4% 1-2% For patients near this threshold, individual values and preferences become more important Concerning coagulopathy • Platelets below 50×10/L 1 or more of: • International normalized ratio (INR) above 1.5 • Prothrombin time (PT) above 20 seconds Shock • Continuous infusion with vasopressors or inotropes 1 or more of: • Systolic blood pressure below 90 mm Hg • Mean arterial blood pressure below 70 mm Hg • Plasma lactate level 4 mmol/L or above Use of anticoagulants • Vitamin K antagonists • Direct acting oral anticoagulants • Therapeutic doses of unfractionated or low molecular weight heparin • Intravenous direct thrombin (II) inhibitors • Adenosine diphosphate receptor inhibitor • Similar drugs Chronic liver disease • Portal hypertension • Cirrhosis proven by biopsy, computed tomography scan, or ultrasound • History of variceal bleeding • History of hepatic encephalopathy 1 or more of:

Evidence profile Proton pump inhibitors

Favours no prophylaxis Favours proton pump inhibitor Evidence quality Events per 1000 people No important difference The panel found that this difference was not important for most patients, because the intervention effects were negligible and/or very imprecise, for example confidence intervals that include both important benefit and harm

No important difference Important bleeding (1-2% risk) Moderate More 12 7

Risk of Bias No serious concerns Imprecision No serious concerns Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Proton pump inhibitors reduce the risk of gastrointestinal bleeding. For people with a 1-2% risk of clinically important bleeding however, the effect is probably small enough that most people would choose not to use them Moderate GRADE score, because of: Uncertainty in baseline risk for some risk factors GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Important bleeding (2-4% risk) Low More 30 19

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Proton pump inhibitors reduce the risk of gastrointestinal bleeding. For people with a 2-4% risk of clinically important bleeding however, the effect may be small enough that most people would choose not to use them Low GRADE score, because of: Uncertainty in baseline risk for some risk factors GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

23 fewer Important bleeding (4-8% risk) Moderate More 60 37

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Proton pump inhibitors reduce the risk of gastrointestinal bleeding. For people with a 4-8% risk of clinically important bleeding, the effect is probably large enough that most people would choose to use them Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

33 fewer Important bleeding (8-10% risk) Moderate More 90 57

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Proton pump inhibitors reduce the risk of gastrointestinal bleeding. For people with a 8-10% risk of clinically important bleeding, the effect is probably large enough that most people would choose to use them Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Mortality Moderate More 304 317

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Proton pump inhibitors probably do not have an important effect on mortality Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

50 fewer Pneumonia Low More 162 212

Risk of Bias Serious Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Proton pump inhibitors may increase the risk of pneumonia Low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Clostridium difficile infection Moderate More 15 12

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Proton pump inhibitors probably do not have an important effect on Clostridium difficile infection Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low
Evidence quality Mean days

No important difference Length of stay in intensive care Moderate More 7.7 7.4

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Proton pump inhibitors probably do not have an important effect on length of stay in intensive care Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low
See all outcomes

Evidence profile Histamine-2 receptor antagonist

Favours no prophylaxis Favours H2RA Evidence quality Events per 1000 people No important difference The panel found that this difference was not important for most patients, because the intervention effects were negligible and/or very imprecise, for example confidence intervals that include both important benefit and harm

No important difference Important bleeding (1-2% risk) Moderate More 12 6

Risk of Bias No serious concerns Imprecision No serious concerns Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Moderate GRADE score, because of: Uncertainty in baseline risk for some risk factors Histamine-2 receptor antagonists reduce the risk of gastrointestinal bleeding. For people with a 1-2% risk of clinically important bleeding however, the effect is probably small enough that most people would choose not to use them GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Important bleeding (2-4% risk) Low More 30 14

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Low GRADE score, because of: Uncertainty in baseline risk for some risk factors Histamine-2 receptor antagonists reduce the risk of gastrointestinal bleeding. For people with a 2-4% risk of clinically important bleeding however, the effect may be small enough that most people would choose not to use them GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

31 fewer Important bleeding (4-8% risk) Moderate More 60 29

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Moderate GRADE score, because of: Histamine-2 receptor antagonists reduce the risk of gastrointestinal bleeding. For people with a 4-8% risk of clinically important bleeding, the effect is probably large enough that most people would choose to use them GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

46 fewer Important bleeding (8-10% risk) Moderate More 90 44

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Moderate GRADE score, because of: Histamine-2 receptor antagonists reduce the risk of gastrointestinal bleeding. For people with a 8-10% risk of clinically important bleeding, the effect is probably large enough that most people would choose to use them GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Mortality Moderate More 304 295

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Histamine-2 receptor antagonists probably do not have an important effect on mortality Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

34 fewer Pneumonia Low More 162 196

We are skeptical of the result becauseit is in conflict with the largest trial Risk of Bias Serious Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Histamine-2 receptor antagonists may increase the risk of pneumonia Low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Clostridium difficile infection Very low More 15 15

Risk of Bias Serious Imprecision Very serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Whether Histamine-2 receptor antagonists increase the risk of Clostridium difficile infection or not is very uncertain Very low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low
Evidence quality Mean days

No important difference Length of stay in intensive care Moderate More 7.7 7.3

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Histamine-2 receptor antagonists probably do not have an important effect on length of stay in intensive care Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low
See all outcomes

Individual considerations

Key practical issues Duration of treatment Values and preferences Proton pump inhibitors Histamine-2 receptor antagonists No prophylaxis None Can be administered intravenously or enterally Typically administered once per day Typically administered two or three times per day A system should be in place to prevent inadvertent continuation of gastric acid suppression It may be challenging to implement shared decision making because there are often many other more important decisions. However, shared decision making should be pursued whenever possible.
Recommendation 2 PPI PPI Proton pump inhibitor H2RA H2RA Histamine-2receptor antagonist Sucralfate SAF S W S W S W S W S W S W S All or nearly all informed people would likely want this intervention. Benefits outweigh harms for almost everyone Strong recommendation W Most people would likely want this intervention. Benefits outweigh harms for the majority, but notfor everyone Weak recommendation In critically ill patients who are going to receive prophylaxis against gastrointestinal bleeding, we suggest a proton pump inhibitor. A histamine-2 receptor antagonist is also a reasonable choice. We recommend not using sucralfate

Evidence profile Proton pump inhibitor versus histamine-2 receptor antagonist

Favours proton pump inhibitor Favours H2RA Evidence quality Events per 1000 people No important difference The panel found that this difference was not important for most patients, because the intervention effects were negligible and/or very imprecise, for example confidence intervals that include both important benefit and harm

No important difference Important bleeding (1-2% risk) Low More 7 12

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Low GRADE score, because of: Uncertainty in baseline risk for some risk factors For people with 1 to 2 % risk of clinically important gastrointestinal bleeding, there may be no important difference between proton pump inhibitors and histamine-2 receptor antagonists GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

Important bleeding (2-4% risk) Low More 32 13 fewer 19

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 2 to 4 % risk of clinically important gastrointestinal bleeding, proton pump inhibitors may reduce the risk more than histamine-2 receptor antagonists Low GRADE score, because of: Uncertainty in baseline risk for some risk factors GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

Important bleeding (4-8% risk) Moderate More 62 25 fewer 37

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 4 to 8 % risk of clinically important gastrointestinal bleeding, proton pump inhibitors probably reduce the risk more than histamine-2 receptor antagonists Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

Important bleeding (8-10% risk) Moderate More 94 37 fewer 57

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 8 to 10 % risk of clinically important gastrointestinal bleeding, proton pump inhibitors probably reduce the risk more than histamine-2 receptor antagonists Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Mortality Very low More 317 295

Risk of Bias No serious concerns Imprecision Extremely serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Whether there is an important difference between proton pump inhibitors and hastamine-2 receptor antagonists on the risk of death or not is very uncertain Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Pneumonia Low More 212 196

Risk of Bias Serious Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns There may be no important difference between proton pump inhibitors and histamine-2 receptor antagonists on risk of pneumonia Low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Clostridium difficile infection Low More 12 15

Risk of Bias Serious Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns There may be no important difference between proton pump inhibitors and histamine-2 receptor antagonists on risk of Clostridium difficile infection Low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low
Evidence quality Mean days

No important difference Length of stay in intensive care High More 7.4 7.3

Risk of Bias No serious concerns Imprecision No serious concerns Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns There is no important difference between proton pump inhibitors and histamine-2 receptor antagonists on length of stay in intensive care Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low
See all outcomes

Evidence profile Proton pump inhibitor versus sucralfate

Favours proton pump inhibitor Favours sucralfate Evidence quality Events per 1000 people No important difference The panel found that this difference was not important for most patients, because the intervention effects were negligible and/or very imprecise, for example confidence intervals that include both important benefit and harm

Important bleeding (1-2% risk) Very low More 23 16 fewer 7

Risk of Bias No serious concerns Imprecision Very serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Very low GRADE score, because of: Uncertainty in baseline risk for some risk factors For people with 1 to 2% risk of clinically important gastrointestinal bleeding, whether there is an important difference between proton pump inhibitors and sucralfate on clinically important gastrointestinal bleeding or not is very uncertain GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

Important bleeding (2-4% risk) Very low More 61 42 fewer 19

Risk of Bias No serious concerns Imprecision Very serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 2 to 4% risk of clinically important gastrointestinal bleeding, whether there is an important difference between proton pump inhibitors and sucralfate on clinically important gastrointestinal bleeding or not is very uncertain Very low GRADE score, because of: Uncertainty in baseline risk for some risk factors GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

Important bleeding (4-8% risk) Low More 113 76 fewer 37

Risk of Bias No serious concerns Imprecision Very serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 4 to 8% risk of clinically important gastrointestinal bleeding, proton pump inhibitors may reduce the risk compared with sucralfate Low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

Important bleeding (8-10% risk) Low More 168 111 fewer 57

Risk of Bias No serious concerns Imprecision Very serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 8 to 10% risk of clinically important gastrointestinal bleeding, proton pump inhibitors may reduce the risk compared with sucralfate Low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Mortality Very low More 317 280

Risk of Bias No serious concerns Imprecision Extremely serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Whether there is an important difference between proton pump inhibitors and sucralfate on the risk of death or not is very uncertain Very low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

70 fewer Pneumonia Low More 212 142

We are skeptical of the result becauseit is in conflict with the largest trial Risk of Bias Serious Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Proton pump inhibitors may increase the risk of pneumonia compared with sucralfate Low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low
Evidence quality Mean days

No important difference Length of stay in intensive care Moderate More 7.4 7.1

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns There is probably no important difference between proton pump inhibitors and sucralfate on length of stay in intensive care Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low
See all outcomes

Evidence profile Histamine-2 receptor antagonist versus sucralfate

Favours H2RA Favours sucralfate Evidence quality Events per 1000 people No important difference The panel found that this difference was not important for most patients, because the intervention effects were negligible and/or very imprecise, for example confidence intervals that include both important benefit and harm

No important difference Important bleeding (1-2% risk) Low More 6 13

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 1 to 2 % risk of clinically important gastrointestinal bleeding, there may be no important difference between histamine-2 receptor antagonists and sucralfate on clinically important gastrointestinal bleeding Low GRADE score, because of: Uncertainty in baseline risk for some risk factors GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

Important bleeding (2-4% risk) Low More 30 16 fewer 14

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 2 to 4 % risk of clinically important gastrointestinal bleeding, histamine-2 receptor antagonists may reduce the risk compared with sucralfate Low GRADE score, because of: Uncertainty in baseline risk for some risk factors GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

32 fewer Important bleeding (4-8% risk) Moderate More 29 61

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 4 to 8 % risk of clinically important gastrointestinal bleeding, histamine-2 receptor antagonists probably reduce the risk compared with sucralfate Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

47 fewer Important bleeding (8-10% risk) Moderate More 44 91

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns For people with 8 to 10 % risk of clinically important gastrointestinal bleeding, histamine-2 receptor antagonists probably reduce the risk compared with sucralfate Low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

No important difference Mortality Moderate More 295 280

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns There is probably no important difference between histamine-2 receptor antagonists and sucralfate on the risk of death Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low

53 fewer Pneumonia Low More 196 142

Risk of Bias Serious Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns Histamine-2 receptor antagonists may increase the risk of pneumonia compared with sucralfate Low GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low
Evidence quality Mean days

No important difference Length of stay in intensive care Moderate More 7.3 7.1

Risk of Bias No serious concerns Imprecision Serious Indirectness No serious concerns Inconsistency No serious concerns Publication bias No serious concerns There is probably no important difference between histamine-2 receptor antagonists and sucralfate on length of stay in intensive care Moderate GRADE score, because of: GRADE certainty ratings The authors have a lot of confidence that the true effect is similar to the estimated effect The authors believe that the true effect is probably close to the estimated effect High The true effect might be markedly different from the estimated effect The true effect is probably markedly different from the estimated effect Moderate Low Very low