Confidential: for Review Only Efficacy and Safety of Gastrointestinal Bleeding Prophylaxis in Critically Ill Patients: Systematic Review and Network Meta-Analysis

BMJ

Confidential: For Review Only Efficacy and safety of gastrointestinal bleeding prophylaxis in critically ill patients: systematic review and network meta-analysis

Journal: BMJ

Manuscript ID BMJ-2019-052088

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the 08-Aug-2019 Author:

Complete List of Authors: Wang, Ying; Beijing Chao-Yang Hospital, Capital Medical University, Department of Pharmacy YE, ZHIKANG; McMaster University, Department of Health Research Methods, Evidence and Impact Ge, Long; School of Public Health, Lanzhou University, Evidence Based Social Science Research Center; School of Basic Medical Sciences, Lanzhou University, Evidence Based Medicine Center Siemieniuk, Reed; McMaster University, Health Research Methods, Evidence, and Impact Wang, Xin; Beijing Chaoyang Hospital, Capital Medical University, Beijing, China Wang, Yingkai; Beijing Chaoyang Hospital, Capital Medical University, Beijing, China Hou, Liangying; Evidence Based Social Science Research Center, School of Public Health, Lanzhou University, Lanzhou, China Ma, Zhuo; Beijing Chaoyang Hospital, Capital Medical University, Beijing, China Agoritsas, Thomas; McMaster University, Department of Clinical Epidemiology and Biostatistics; University Hospitals of Geneva, Division of General Internal Medicine & Division of Epidemiology Vandvik, Per; Innlandet Hospital Trust-divisjon Gjøvik, Department of Medicine Perner, Anders; Copenhagen University Hospital Rigshospitalet, Dept. of Intensive Care 4131 Møller, Morten; Copenhagen University Hospital Rigshospitalet, Dept. of Intensive Care 4131 Guyatt, Gordon; McMaster University, Liu, Li-Hong; Department of Pharmacy, Beijing Chao-Yang Hospital, Capital Medical University, Department of Pharmacy

gastrointestinal bleeding prophylaxis, systematic review and network Keywords: meta-analysis, proton pump inhibitors, histamine-2 receptor antagonists, sucralfate

https://mc.manuscriptcentral.com/bmj Page 1 of 109 BMJ

1 2 3 4 5 6 7 8 9 10 11 Confidential: For Review Only 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 2 of 109

1 2 3 4 Efficacy and safety of gastrointestinal bleeding prophylaxis in critically ill 5 6 patients: systematic review and network meta-analysis 7 8 9 1 1,2 3 2,4 1 10 Ying Wang, Zhikang Ye, Long Ge, Reed A Siemieniuk, Xin Wang, Yingkai 11 Wang,1 Liangying Hou,3 Zhuo Ma,1 Thomas Agoritsas,2,5 Per Olav Vandvik,6 Anders 12 Confidential: For Review Only 13 Perner,7 Morten H Møller,7 Gordon H Guyatt,2 Lihong Liu1 14 15 16 17 1Department of Pharmacy, Beijing Chaoyang Hospital, Capital Medical University, 18 19 Beijing, China 20 21 2Department of Health Research Methods, Evidence and Impact, McMaster 22 23 University, Hamilton, Canada 24 25 3Evidence Based Social Science Research Center, School of Public Health, Lanzhou 26 27 University, Lanzhou, China 28 29 4Department of Medicine, University of Toronto, Toronto, Canada 30 31 5Division General Internal Medicine, and Division of Clinical Epidemiology, 32 33 University Hospitals of Geneva, Geneva, Switzerland 34 35 6Institute of Health and Society, University of Oslo, Norway 36 37 7Department of Intensive Care, Copenhagen University Hospital Rigshospitalet, 38 39 Copenhagen, Denmark 40 41 Correspondence to: L Liu [email protected] 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 1

https://mc.manuscriptcentral.com/bmj Page 3 of 109 BMJ

1 2 3 4 ABSTRACT 5 6 OBJECTIVE 7 8 To address the relative impact of gastrointestinal bleeding prophylaxis (often referred 9 10 to as stress ulcer prophylaxis or SUP) with proton pump inhibitors (PPIs), histamine-2 11 receptor antagonists (H2RAs), sucralfate or no prophylaxis in critically ill patients. 12 Confidential: For Review Only 13 DESIGN 14 15 Systematic review and network meta-analysis. 16 17 DATA SOURCES 18 19 Medline, PubMed, Embase, Cochrane Central Register of Controlled Trials, trial 20 21 registers, and grey literature up to March 2019. 22 23 STUDY SELECTION 24 25 Randomised controlled trials comparing gastrointestinal bleeding prophylaxis with 26 27 PPIs, H2RAs, or sucralfate versus one another or placebo or no prophylaxis in adult 28 29 critically ill patients. 30 31 REVIEW METHODS 32 33 Two reviewers independently screened studies for eligibility, extracted data and 34 35 assessed risk of bias. A parallel guideline committee (BMJ Rapid Recommendation) 36 37 provided critical oversight of the systematic review, including identifying outcomes 38 39 important to patients. We performed random-effects pairwise and network 40 41 meta-analyses (NMA) and used GRADE to assess certainty of evidence for each 42 43 outcome. When results differed between low and high risk of bias studies we used the 44 45 former as best estimates. 46 47 RESULTS 48 49 Seventy-two trials including 12 660 patients proved eligible. For patients at highest 50 (above 8%) or high (4% to 8%) risk of bleeding, both PPIs (odds ratio 0.61, 95% 51 52 confidence interval 0.42 to 0.89, 3.3% fewer for highest and 2.3% fewer for high risk 53 54 patients, moderate certainty) and H2RAs (OR 0.46, 0.27 to 0.79, 4.6% fewer for 55 56 highest and 3.1% fewer for high risk patients, moderate certainty) probably reduce 57 58 clinically important gastrointestinal bleeding compared with placebo or no 59 60 2

https://mc.manuscriptcentral.com/bmj BMJ Page 4 of 109

1 2 3 4 prophylaxis. PPIs (OR 1.38, 0.97 to 2.08, 4.8% more, low certainty) and H2RAs (OR 5 6 1.26, 0.90 to 1.86, 3.4% more, low certainty) may increase pneumonia compared with 7 8 no prophylaxis. It is likely that neither PPIs nor H2RAs impact on mortality; 9 10 otherwise, results provided no support for any impacting on mortality, Clostroides 11 difficile infection, length of ICU stay, length of hospital stay, or duration of 12 Confidential: For Review Only 13 mechanical ventilation (varying certainty of evidence). 14 15 CONCLUSIONS 16 17 Evidence failed to support differences in a number of outcomes, including mortality. 18 19 For higher risk critically ill patients, PPIs and H2RAs compared with no prophylaxis 20 21 likely result in important reductions in gastrointestinal bleeding; for patients at low 22 23 risk, the reduction in bleeding may be unimportant. Both PPIs and H2RAs may 24 25 increase pneumonia. 26 27 SYSTEMATIC REVIEW REGISTRATION 28 29 PROSPERO CRD42019126656. 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 Introduction 59 60 3

https://mc.manuscriptcentral.com/bmj Page 5 of 109 BMJ

1 2 3 4 Critically ill patients in the intensive care unit (ICU) are at risk for gastrointestinal 5 1 6 bleeding from a number of sources, one of which is stress ulceration. Authorities 7 8 have suggested gastrointestinal bleeding prophylaxis (often referred to as stress ulcer 9 10 prophylaxis or SUP) as essential to optimize the care of critically ill patients, and most 11 patients considered high-risk receive acid suppression during an ICU stay.2 3 Proton 12 Confidential: For Review Only 13 pump inhibitors (PPIs) are the most common prophylactic agent, followed by 14 15 histamine-2 receptor antagonists (H2RAs); clinicians seldom use sucralfate and 16 17 antacids.2 4 18 19 Recently, however, the practice has generated controversy.5 Doubts include 20 21 questions of whether bleeding occurs frequently enough, or whether its consequences 22 23 are serious enough, to warrant prophylaxis.5 Moreover, observational studies have 24 25 reported increased rates of nosocomial pneumonia and Clostroides difficile infection 26 27 with the use of acid-suppressive drugs,6 7 raising concern that harms may outweigh the 28 29 benefits. 30 31 Many published systematic reviews and meta-analyses have summarized 32 33 randomised controlled trial (RCT) evidence addressing the efficacy and safety of 34 35 interventions for gastrointestinal bleeding prophylaxis,8-13 including a network 36 37 meta-analysis (NMA) conducted by members of our team.8 Since the publication of 38 39 the last NMA, several trials have been published,14-17 including a large, international, 40 17 41 multicenter RCT (the SUP-ICU trial). This trial compared pantoprazole with 42 43 placebo and concluded that pantoprazole did not reduce mortality or a composite 44 45 secondary outcome of ‘clinically important events’, and questioned the routine use of 46 47 PPIs in critically ill adults. 48 49 We therefore conducted an updated systematic review and NMA addressing the 50 potential benefits and harms of gastrointestinal bleeding prophylaxis with PPIs, 51 52 H2RAs, and sucralfate in critically ill patients. This review is part of the BMJ Rapid 53 54 Recommendations project, a collaborative effort from the MAGIC Evidence 55 56 Ecosystem Foundation (www.magicproject.org) and BMJ. The aim of the project is to 57 58 respond to new potentially practice changing evidence and provide a trustworthy 59 60 4

https://mc.manuscriptcentral.com/bmj BMJ Page 6 of 109

1 2 3 18 4 practice guideline in a timely manner. In this case, the stimulus was the SUP-ICU 5 17 6 trial. This systematic review informed the parallel guideline published in a 7 8 multilayered electronic format on bmj.com and MAGICapp 9 10 (https://app.magicapp.org/app#/guideline/3096). 11 Methods 12 Confidential: For Review Only 13 Protocol registration 14 15 We registered the protocol for this systematic review with PROSPERO 16 17 (CRD42019126656).19 18 19 Guideline panel and patient involvement 20 21 In accordance with the BMJ Rapid Recommendations process, a guideline panel 22 23 provided critical oversight during the review process, which included identifying 24 25 populations, subgroups, and outcomes of interest. The guideline panel consisted of 26 27 content experts, critical care clinicians, gastroenterologist, pharmacists, 28 29 methodologists, former patients and a patient caregiver. Patients received personal 30 31 training and support to optimize contributions throughout the guideline development 32 33 process. 34 35 Search strategy 36 37 Our literature search, developed in collaboration with a research librarian, added the 38 39 brand, generic and experimental names of PPIs, H2RAs and sucralfate to the original 40 41 search terms from our previous NMA (appendix 1). The search included Medline and 42 43 Embase, the Web of Science, Cochrane Central Register of Controlled Trials 44 45 (CENTRAL), International Clinical Trials Registry Platform (ICTRP) and 46 47 clinicaltrials.gov from January 2017 to March 2019 and a PubMed search for studies 48 49 not yet indexed or not found in Medline. Reviewers also searched the abstracts of the 50 past two years of proceedings for the following conferences: Digestive Disease Week, 51 52 United European Gastroenterology Week, European Society of Intensive Medicine, 53 54 Society of Critical Care Medicine, and scanned the reference lists of included studies 55 56 and relevant systematic reviews to identify potential eligible studies. 57 58 Study selection 59 60 5

https://mc.manuscriptcentral.com/bmj Page 7 of 109 BMJ

1 2 3 4 Two reviewers independently performed the study selection, including screening titles 5 6 and abstracts, and evaluating full-text eligibility of potentially eligible studies. 7 8 Reviewers resolved disagreements by discussion or by consultation with a third 9 10 reviewer. 11 We included RCTs that compared the efficacy and safety of gastrointestinal 12 Confidential: For Review Only 13 bleeding prophylaxis, PPIs, H2RAs, or sucralfate versus one another or placebo or no 14 15 prophylaxis in adult critically ill patients at risk of gastrointestinal bleeding. We 16 17 excluded studies that did not report sufficient information to pool data (e.g., uncertain 18 19 number of events or number of patients in each group). We applied no restriction 20 21 based on dose or route of administration, duration of prophylaxis, or language of 22 23 publication. 24 25 Outcomes included mortality at the longest follow up reported, clinically important 26 27 gastrointestinal bleeding (CIB), pneumonia, C. difficile infection, overt 28 29 gastrointestinal bleeding, length of ICU stay, length of hospital stay and duration of 30 31 mechanical ventilation. We accepted study definitions of CIB, which typically 32 33 included evidence of upper gastrointestinal bleeding with any of the following: 34 35 significant hemodynamic changes not explained by other causes, need for transfusion 36 37 of more than two units of blood, significant decrease in hemoglobin level, evidence of 38 39 bleeding on upper gastrointestinal endoscopy, or need for surgery to control bleeding. 40 41 We accepted study definitions of overt bleeding typically defined as hematemesis, 42 43 melena, hematochezia, or coffee-grounds emesis or aspirate. When extracting data on 44 45 overt bleeding, we included CIB. We accepted the definitions of pneumonia and C. 46 47 difficile infection used in each trial. 48 49 Data extraction 50 For each eligible study, two reviewers independently abstracted the following items 51 52 with adjudication by a third reviewer: study characteristics (year of publication and 53 54 country), population characteristics (sample size, age, proportion of males, ICU type, 55 56 risk factors for gastrointestinal bleeding and proportion of mechanical ventilation and 57 58 enteral nutrition), description of interventions and comparators (the name, dose, 59 60 6

https://mc.manuscriptcentral.com/bmj BMJ Page 8 of 109

1 2 3 4 administration route, frequency, and duration), outcomes and their definitions. 5 6 Risk of bias assessment 7 8 Two reviewers independently assessed risk of bias with adjudication by a third 9 20 10 reviewer, using a modified Cochrane Collaboration tool that includes sequence 11 generation, allocation sequence concealment, blinding, missing outcome data (we 12 Confidential: For Review Only 13 judged high risk of bias if the rate of missing data was more than 5%), and other bias 14 15 (in this case, early trial discontinuation due to benefit). Each criterion was judged as 16 17 definitely or probably satisfied (low risk of bias), or probably or definitely not 18 19 satisfied (high risk of bias). 20 21 Data synthesis and analysis 22 23 For each direct comparison for each outcome, we performed a Bayesian 24 25 random-effects pairwise meta-analysis assessing heterogeneity with visual inspection 26 27 of forest plots and the I2 statistic.21 We calculated odds ratios (ORs) and 28 29 corresponding 95% credible intervals for all dichotomous outcomes, and mean 30 31 differences (MDs) with corresponding 95% credible intervals for continuous 32 33 outcomes. We performed Egger test to assess the publication bias when 10 or more 34 35 studies were available for a comparison.22 36 37 We performed a Bayesian random-effects NMA using Markov-chain Monte-Carlo 38 39 simulation.23 24 We used three chains with 100 000 iterations after an initial burn-in of 40 41 10 000 and a thinning of 10. We assessed the convergence based on trace plots and 42 43 the Brooks-Gelman-Rubin statistic, with an acceptable threshold of <1.05 for all 44 45 nodes. We used non-informative priors for all parameters and assumed a common 46 47 heterogeneity parameter for all treatment comparisons. We calculated the ORs and 48 49 corresponding 95% credible intervals of network estimates with a consistency model. 50 For all dichotomous outcomes, we calculated the absolute treatment effects of the 51 52 network estimates based-on the ORs and the event rates in the placebo arm in 53 54 SUP-ICU trial using the modified Dias model, incorporating lines into the model.25 55 56 The guideline panel searched for evidence on risk factors for bleeding and proposed 57 58 four categories of risk of bleeding (linked BMJ recommendation shows details on 59 60 7

https://mc.manuscriptcentral.com/bmj Page 9 of 109 BMJ

1 2 3 4 calculating baseline risks) and calculated absolute effects for each category for both 5 6 CIB and overt bleeding. For continuous variables, when absent in reports, we 7 8 estimated standard deviations (SDs) from standard errors, P values, ranges, or 9 10 interquartile ranges or, if none of these options were available, from other studies 11 included in our NMA using a linear regression approach.26 12 Confidential: For Review Only 13 For each paired comparison we used conventional pairwise random-effects 14 15 meta-analysis to inform the direct estimates and, through a node-splitting method, 16 17 obtained the indirect estimates. We evaluated the ranking probabilities and calculated 18 19 surface under the cumulative ranking curves (SUCRA). We used the node-splitting 20 21 method to assess local incoherence (incoherence was assessed using both this test and 22 23 the difference in point estimates and overlap of the credible intervals).27 We assessed 24 25 the incoherence of the entire network using global I2 statistics. A comparison-adjusted 26 27 funnel plot of treatment estimates addressed small-study effects.28 28 29 We conducted meta-regression to explore the impact of studies with high versus 30 31 low risk of bias for each risk of bias criterion. For the network meta-regression, the no 32 33 treatment or placebo group was the referent, and we assumed that effect modification 34 35 differed between comparisons. If risk of bias influenced results, we included only low 36 37 risk of bias studies in generating best estimates. To explore the impact of enteral 38 39 nutrition and mechanical ventilation, we conducted meta-regression using the 40 41 proportion of patients with enteral nutrition or mechanical ventilation as the 42 43 independent variable. We performed subgroup analysis comparing results in studies 44 45 that specified inclusion of critically ill patients with risk factors for gastrointestinal 46 47 bleeding versus studies that did not mention risk factors in their inclusion criteria. 48 49 When networks are sparse, random-effects models may generate implausibly wide 50 credible intervals from NMA estimates, even when the direct and indirect estimates 51 52 are coherent. When this occurred, we either conducted a fixed-effect NMA or used the 53 54 direct estimates as our best estimates of the treatment effects.29 30 All NMAs were 55 56 performed using the gemtc package of R version 3.4.3 (R Core Team, Vienna, 57 58 Austria), absolute effects in networks were calculated using WinBUGS version 1.4.3 59 60 8

https://mc.manuscriptcentral.com/bmj BMJ Page 10 of 109

1 2 3 4 (MRC Biostatistics Unit, Cambridge, UK), and networkplot command of Stata 5 6 version 15.1 (StataCorp, College Station, Texas, USA) was used to draw the network 7 31 8 plots. 9 10 In networks with incoherence in many comparisons, we compared network 11 estimates to direct and indirect estimates. When we found direct versus network 12 Confidential: For Review Only 13 estimates inconsistent and counterintuitive, we examined ORs and relative risks in 14 15 both random-effects and fixed-effect models, seeking an approach that provided 16 17 plausible network estimates. When none of these approaches were satisfactory, we did 18 19 not consider the NMA further for those outcomes, and instead relied on direct paired 20 21 comparisons as the best estimates of effect. 22 23 Assessment of certainty of evidence 24 25 The GRADE (Grading of Recommendations, Assessment, Development, and 26 27 Evaluation) approach informed the assessment of certainty of evidence for each 28 29 outcome,32-34 including certainty ratings from the direct comparisons in our pairwise 30 31 meta-analyses,35-40 and certainty of evidence in NMAs.32 34 Certainty ratings of 32 33 indirect estimates started at the lowest ratings of the direct comparisons that 34 35 contributed to the most-dominant first order loop with further rating down, when 36 37 necessary, for intransitivity. Ratings of the certainty of estimates for direct and 38 39 indirect estimates to inform the rating of network estimates included risk of bias, 40 41 inconsistency, indirectness and publication bias, while imprecision was assessed at the 42 43 network level. We judged imprecision by comparing the confidence intervals or 44 36 45 credible intervals to decision thresholds chosen by the guideline panel. For the 46 47 certainty of network estimates, we started with the estimate – direct or indirect – that 48 49 dominated the network estimate, or used the higher of the direct and indirect estimates 50 if they both contributed importantly to the network estimate. If incoherence was 51 52 present, we rated down the certainty of the network estimates and used the estimate – 53 54 direct or indirect – with the higher certainty evidence as the best estimate of treatment 55 56 effect. We used the MAGICapp platform to develop the GRADE summary of finding 57 58 tables. 59 60 9

https://mc.manuscriptcentral.com/bmj Page 11 of 109 BMJ

1 2 3 4 Results 5 6 Description of included studies 7 8 Figure 1 shows the details of study selection process. We identified 421 records from 9 10 our updated literature search and 13 potentially eligible studies from relevant reviews. 11 After title and abstract screening, we assessed 70 full text articles, of which 18 proved 12 Confidential: For Review Only 13 eligible. Including 54 trials from our previous NMA (populations from 3 trials 14 15 identified from previous NMA did not meet our eligibility criteria), we ultimately 16 17 included 72 trials. 18 19 These 72 RCTs, with sample sizes from 22 to 3 298, enrolled a total of 12 660 20 21 patients. Figure 2 presents the network plot including all outcomes and demonstrates 22 23 that the most common comparisons were between H2RAs and placebo or no 24 25 prophylaxis followed by H2RAs versus sucralfate and PPIs versus H2RAs. 26 27 Fifty trials included critically ill patients with risk factors for gastrointestinal 28 29 bleeding (appendix 2 summarizes risk factors listed), and 22 trials only specified 30 31 inclusion of critically ill patients. Sixty-five trials included ICU patients, 2 trials 32 33 neurosurgery patients, 1 trial patients with fulminant hepatic failure, 1 trial patients 34 35 from three wards (department of intensive care, neurosurgery and plastic surgery) and 36 37 3 trials critically ill patients without further specification (appendix 3). 38 39 Risk of bias 40 41 Appendix 4 summarizes risk of bias assessment that identified the main limitations as 42 43 possible lack of allocation sequence concealment (54.2%), lack of blinding (62.5%), 44 45 and excessive loss to follow-up (23.6%). 46 47 Outcomes 48 49 Appendix 5 presents network plots for each outcome. Tests of incoherence proved 50 unproblematic for mortality, pneumonia, C. difficile infection, length of ICU stay, 51 52 length of hospital stay and duration of mechanical ventilation. For CIB and overt 53 54 bleeding, we found incoherence for many comparisons and found counterintuitive 55 56 results in each of the models we explored (appendix 6). We therefore focused on 57 58 results from conventional pairwise meta-analysis rather than from NMA for CIB and 59 60 10

https://mc.manuscriptcentral.com/bmj BMJ Page 12 of 109

1 2 3 4 overt bleeding. Appendix 7 presents SUCRA ranking of interventions for each 5 6 outcome. 7 8 Mortality 9 14 15 17 41-89 10 NMA estimates from 51 trials including 10 325 patients demonstrated odds 11 ratios for all comparisons close to 1.0 with moderate or very low certainty (table 1 12 Confidential: For Review Only 13 appendix 8). Network meta-regression failed to demonstrate an impact of risk of bias 14 15 on results (appendix 9). The SUP-ICU trial reported a possible subgroup effect 16 17 suggesting that PPIs relative to placebo may increase mortality in the sickest 18 19 patients,17 but the credibility of this subgroup effect is low (most important reasons 20 21 for low credibility: the subgroup hypothesis and the direction of the subgroup effect 22 23 are not specified a priori).90 24 25 CIB 26 27 Forty-three trials including 10 458 patients14-17 43 46-49 51 54-60 62 63 65-67 69 71 75 76 78 80-85 28 29 87-89 91-98 reported CIB. Subgroup analysis based on each risk of bias criterion 30 31 (concealment and blinding were highly correlated, so they were combined into one 32 33 variable) showed that studies that failed to conceal/blind reported larger effect for 34 35 PPIs versus H2RAs and smaller effect for H2RAs versus sucralfate (table 1 and 36 37 appendix 10); for these comparisons, we focused on low risk of bias results. 38 39 We grouped patients into 4 categories according to risk of bleeding: low risk 40 41 (below 2%), moderate risk (2% to 4%), high risk (4% to 8%), and highest risk (above 42 43 8%) (appendix 11). Results demonstrated that for patients at highest or high risk of 44 45 bleeding, both PPIs (odds ratio 0.61, 95% confidence interval 0.42 to 0.89, 33 fewer 46 47 per 1000 for highest and 23 fewer for high risk patients, moderate certainty; table 2 48 49 and figure 3) and H2RAs (0.46, 0.27 to 0.79, 46 fewer per 1000 for highest and 31 50 fewer for high risk patients, moderate certainty; table 2 and figure 4) probably reduce 51 52 the risk of CIB compared with placebo or no prophylaxis. For patients at lower risk 53 54 the absolute effects were considerably smaller (table 2). PPIs possibly reduce CIB 55 56 relative to H2RAs but the confidence intervals are wide and include a benefit of 57 58 H2RAs (0.58, 0.29 to 1.16; table 2 and figure 5). H2RAs compared to sucralfate 59 60 11

https://mc.manuscriptcentral.com/bmj Page 13 of 109 BMJ

1 2 3 4 probably reduce CIB (0.46, 0.23 to 0.91, moderate certainty for higher risk groups; 5 6 table 2). Evidence regarding sucralfate versus PPIs and placebo was essentially 7 8 uninformative with very wide confidence intervals (table 2). Both subgroup analysis 9 10 within SUP-ICU trial and meta-regression between trials suggested enteral nutrition 11 did not influence the relative impact of interventions (appendix 9 presents results for 12 Confidential: For Review Only 13 meta-regression and subgroup analysis). 14 15 Pneumonia 16 17 Forty trials including 9 663 patients14 17 41-43 46-50 55 56 58 59 61 63 65-69 71 73 75 77 78 81-85 87 92-97 18 19 99-101 reported pneumonia. PPIs (odds ratio 1.38, 95% credible interval 0.97 to 2.08, 20 21 48 more per 1000 patients, low certainty; table 3) and H2RAs (1.26, 0.90 to 1.86, 34 22 23 more per 1000 patients, low certainty; table 3) may increase the risk of pneumonia 24 25 compared with placebo or no prophylaxis. PPIs (1.58, 1.09 to 2.39, 65 more per 1000 26 27 patients, low certainty; table 3) and H2RAs (1.44, 1.09 to 2.00, 51 more per 1000 28 29 patients, low certainty; table 3) may increase the risk of pneumonia compared with 30 31 sucralfate. Network meta-regression failed to demonstrate an impact of risk of bias on 32 33 results for pneumonia (appendix 9). 34 35 C. difficile infection 36 37 Five trials including 3 856 patients14 15 17 82 98 reported C. difficile infection, of which 38 39 four compared PPI with placebo or no prophylaxis and one compared PPI with H2RA. 40 41 C. difficile infection was rare (baseline risk of 1.5%) and the absolute effect of any 42 43 intervention would therefore be very small (e.g. for PPIs versus placebo, odds ratio 44 45 0.83, 95% credible interval 0.31 to 2.48, 3 fewer per 1000 patients, moderate 46 47 certainty; table 2 appendix 8). Credible intervals around odds ratios were very wide 48 49 and results therefore essentially uninformative (table 2 appendix 8). Network 50 meta-regression failed to demonstrate an impact of risk of bias on results for C. 51 52 difficile infection (appendix 9). 53 54 Overt gastrointestinal bleeding 55 56 Sixty-five trials including 12 022 patients14-17 41 43-52 54-60 62-71 73-85 87-89 91-98 100-109 57 58 reported on overt bleeding. Subgroup showed that concealment/blinding influenced 59 60 12

https://mc.manuscriptcentral.com/bmj BMJ Page 14 of 109

1 2 3 4 results for the comparisons of PPIs versus placebo and H2RAs versus sucralfate, and 5 6 missing outcome data influenced results for PPIs versus placebo (table 4 and appendix 7 8 10). We therefore used low risk of bias results for these two comparisons. PPIs reduce 9 10 overt bleeding relative to placebo (odds ratio 0.59, 95% confidence interval 0.45 to 11 0.77, high certainty; table 3 appendix 8). This is probably also true for H2RAs versus 12 Confidential: For Review Only 13 placebo (0.38, 0.24 to 0.59, moderate certainty; table 3 appendix 8). Sucralfate 14 15 possibly reduces overt bleeding relative to placebo but the confidence interval 16 17 includes increased bleeding (0.58, 0.30 to 1.11, moderate certainty; table 3 appendix 18 19 8). PPIs reduce overt bleeding relative to H2RAs (0.44, 0.28 to 0.69, high certainty; 20 21 table 3 appendix 8). We did not find compelling evidence for a subgroup effect of 22 23 receipt versus non-receipt of enteral nutrition, or of RCTs that restricted to enrollment 24 25 to RCTs including patients with risk factors for bleeding (appendix 9).90 26 27 Length of ICU stay and length of hospital stay 28 29 Seventeen trials including 3 877 patients14 15 48 49 53 58 62 65 66 73 75 79 82-84 89 92 98 reported 30 31 length of ICU stay and 7 trials including 842 patients14 15 42 61 75 82 97 reported length of 32 33 hospital stay. Results suggested no important difference between any of interventions 34 35 (table 4 and table 5 appendix 8). Network meta-regression failed to demonstrate an 36 37 impact of risk of bias on results (appendix 9). 38 39 Duration of mechanical ventilation 40 14 15 41-43 45 48 55 56 58 61 63 66 67 73 75 80 81 83 84 87 41 Twenty-three trials including 3 644 patients 42 93 97 98 43 reported duration of mechanical ventilation or duration of intubation. Results 44 45 suggested no significant difference between any of interventions in terms of duration 46 47 of mechanical ventilation or duration of intubation (table 6 appendix 8). Network 48 49 meta-regression failed to demonstrate an impact of risk of bias on results for duration 50 of mechanical ventilation (appendix 9). 51 52 Discussion 53 54 This review provides moderate certainty evidence that neither PPIs not H2RAs 55 56 relative to no prophylaxis impact on mortality and lower certainty evidence of no 57 58 impact on mortality for other comparisons (table 1 appendix 8). The review provides 59 60 13

https://mc.manuscriptcentral.com/bmj Page 15 of 109 BMJ

1 2 3 4 moderate or low certainty evidence that PPIs and H2RAs probably reduce CIB 5 6 relative to no prophylaxis (table 2) and provides high or moderate certainty evidence 7 8 that both drugs reduce overt gastrointestinal bleeding relative to no prophylaxis (table 9 10 3 appendix 8). For higher risk patients, the impact on CIB may be important (PPIs 11 absolute difference for highest risk patients 3.3%, for high risk patients 2.3%; H2RAs 12 Confidential: For Review Only 13 for highest risk patients 4.6%, for high risk patients 3.1%; table 2); this may not be 14 15 true for lower risk patients (table 2). PPIs relative to H2RAs possibly reduce CIB but 16 17 the confidence interval includes superiority of H2RAs (odds ratio 0.58, 95% 18 19 confidence interval 0.29 to 1.16; table 2 and figure 5). Sucralfate may not reduce CIB 20 21 (table 2). 22 23 PPIs (absolute increase 4.8%) and H2RAs (absolute increase 3.4%) may increase 24 25 the risk of pneumonia (table 3). C. difficile infection occurred infrequently resulting in 26 27 very wide credible intervals around odds ratios that were essentially uninformative. 28 29 However, because C. difficile infection was rare, even the largest plausible increase in 30 31 risk will be small. Few trials reported on other adverse effects. Prophylaxis may have 32 33 no impact on length of ICU stay, length of hospital stay or duration of mechanical 34 35 ventilation. 36 37 Strength and limitations 38 39 Strengths of this review include a comprehensive search to identify eligible trials; 40 41 independent study selection, data extraction, and risk of bias assessment by two 42 43 reviewers; focus on low risk of bias studies when low and high risk of bias studies 44 45 yielded differing results; and application of the GRADE approach to rate the certainty 46 47 of evidence. We also presented absolute effects for all comparisons and outcomes and 48 49 for patients at different risk of gastrointestinal bleeding. 50 Finally, rather than mechanically reporting NMA results, we also considered their 51 52 credibility. It is probably inadvisable to conduct NMA when direct and indirect 53 54 estimates differ considerably throughout the network (i.e. incoherence).110 We found 55 56 this problem for both CIB and overt bleeding and, probably as a result, found NMA 57 58 estimates that were inconsistent with direct estimates and counterintuitive. Testing 59 60 14

https://mc.manuscriptcentral.com/bmj BMJ Page 16 of 109

1 2 3 4 different measures of effect or model assumptions did not resolve the issue. 5 6 Therefore, for these two outcomes, we relied on direct comparisons only. 7 8 Limitations include clinical heterogeneity across eligible trials that enrolled 9 10 different populations, and often did not clearly identify the risk factors for bleeding in 11 the participating individuals. However, most patients in most trials received 12 Confidential: For Review Only 13 mechanical ventilation, an indication for prophylaxis in most ICUs worldwide. 14 15 Definitions of CIB and overt bleeding differed; most, however, adhered to standard 16 17 criteria. Most trials did not report the duration of follow-up. 18 19 Some results proved logically inconsistent. For instance, in comparisons against 20 21 placebo or no prophylaxis, results showed larger effects with H2RAs than with PPIs 22 23 on reduction in CIB. Direct comparisons of the two agents, however, raise the 24 25 possibility that PPIs result in larger reductions in CIB than do H2RAs. 26 27 Enteral nutrition may provide protection against gastrointestinal bleeding and the 28 29 impact of prophylaxis in patients receiving enteral nutrition may therefore be very 30 31 small, at least in patients at otherwise lower risk. It is also possible that enteral 32 33 nutrition decreases the relative effect of prophylaxis on bleeding. Although 34 35 meta-regression results suggested H2RAs had smaller relative effect on overt bleeding 36 37 when patients received enteral nutrition, this subgroup effect proved of low credibility 38 39 (appendix 9).90 We also identified a possible subgroup effect for PPIs versus placebo: 40 41 studies in which investigators specified that patients had risk factors had apparently 42 43 smaller effects on overt bleeding than those in which there was no mention or risk 44 45 factors. Once again, however, the apparent subgroup effect is of low credibility 46 90 47 (appendix 9). 48 49 Relation to prior work 50 Compared with previous conventional pairwise meta-analysis, the use of indirect 51 52 comparisons within this NMA adds additional information to the evidence regarding 53 54 gastrointestinal bleeding prophylaxis. Our review included more trials and 55 56 substantially more patients: most importantly the SUP-ICU trial,17 which is the largest 57 58 trial to date addressing prophylaxis. In this review, we found that H2RAs likely 59 60 15

https://mc.manuscriptcentral.com/bmj Page 17 of 109 BMJ

1 2 3 4 reduce risk of CIB compared with no prophylaxis and sucralfate; our previous NMA 5 6 suggested no convincing impact of H2RAs. One consequence of this review’s focus 7 8 on low risk of bias studies when results differed in high and low risk of bias studies 9 10 was a confidence interval for the comparison of PPIs versus H2RAs on CIB wider 11 than those in previous studies and including the possibility of benefit with H2RAs. 12 Confidential: For Review Only 13 We found moderate certainty evidence that neither PPIs nor H2RAs reduce 14 15 mortality relative to no prophylaxis. The SUP-ICU trial reported a possible subgroup 16 17 effect suggesting PPIs may increase mortality in the sickest patients. Applying 18 19 suggested criteria, however, demonstrated this possible subgroup effect is of low 20 21 credibility (appendix 9).90 22 23 Implications 24 25 This systematic review provides important information for weighing the potential 26 27 benefits and harms of alternative interventions for gastrointestinal bleeding 28 29 prophylaxis in adult critically ill patients. Key messages include the likelihood that no 30 31 intervention influences mortality; that PPIs and H2RAs likely reduce bleeding; that 32 33 reduction in bleeding is higher in patients with higher baseline risk of bleeding; and 34 35 that resolving a number of issues, including the possibility that prophylactic agents 36 37 increase the risk of pneumonia will require additional RCTs. This review provides key 38 39 evidence for the associated BMJ Rapid Recommendation. 40 41 42 43 44 45 46 47 48 49 50 We thank members of the Rapid Recommendations panel for critical feedback on 51 52 outcome selection, GRADE judgments, and manuscript feedback. We thank Karin 53 54 Dearness for helping with developing search strategy. We thank Davide Papola, 55 56 Stefan Schandelmaier, Suzana Alves da Silva for helping with data extraction of some 57 58 non-English studies. We thank Melanie Chiarot for helping us finding full texts for 59 60 16

https://mc.manuscriptcentral.com/bmj BMJ Page 18 of 109

1 2 3 4 some studies. 5 6 Contributors: GHG, ZY, YW and LL conceived the study. KD and ZY designed the 7 8 search strategy and KD performed the literature search. YW and XW screened studies 9 10 for eligibility. XW, YW, Yingkai Wang, and ZM performed data extraction. YW, ZY 11 and GHG assessed the risk of bias. LG, LH and YW performed data analysis. YW, 12 Confidential: For Review Only 13 RAS, LG, ZY and GHG interpreted the data analysis and assessed the certainty of 14 15 evidence. YW, LG and ZY wrote the first draft of the manuscript and all other authors 16 17 revised the manuscript. GHG and LL are the guarantors. 18 19 Funding: This systematic review is linked to a BMJ Rapid Recommendation, which 20 21 funded by the Digestive Medical Coordinated Development Center of Beijing 22 23 Hospitals Authority. The funding did not play any role in the preparation and 24 25 production of the systematic review. 26 27 Competing interests: All authors have completed the ICMJE uniform disclosure 28 29 form at www.icmje.org/coi_disclosure.pdf and declare: AP and MHM are the 30 31 co-authors of SUP-ICU trial (doi: 10.1056/NEJMoa1714919). GHG is one of the 32 33 investigators of REVISE trial. There are no other relationships or activities that could 34 35 appear to have influenced the submitted work. 36 37 Ethical approval: Not required. 38 39 Data sharing: No additional data available. 40 41 Transparency: The manuscript’s guarantors (GHG and LL) affirm that the 42 43 manuscript is an honest, accurate, and transparent account of the recommendation 44 45 being reported; that no important aspects of the recommendation have been omitted; 46 47 and that any discrepancies from the recommendation as planned have been explained. 48 49 This is an Open Access article distributed in accordance with the Creative Commons 50 Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to 51 52 distribute, remix, adapt, build upon this work non-commercially, and license their 53 54 derivative works on different terms, provided the original work is properly cited and 55 56 the use is noncommercial. See: http://creativecommons.org/licenses/by-nc/4.0/ 57 58 59 60 17

https://mc.manuscriptcentral.com/bmj Page 19 of 109 BMJ

1 2 3 4 Appendix 1: Search terms and strategies 5 6 Appendix 2: List of risk factors identified from included trials 7 8 Appendix 3: Study characteristics and outcome definition 9 10 Appendix 4: Risk of bias of included studies 11 Appendix 5: Network plots for each outcome 12 Confidential: For Review Only 13 Appendix 6: Network meta-analysis results for clinically important gastrointestinal 14 15 bleeding and overt gastrointestinal bleeding using different models 16 17 Appendix 7: Cumulative ranking of interventions for different outcomes 18 19 Appendix 8: GRADE summary of findings for other outcomes 20 21 Appendix 9: Meta-regression and subgroup analysis 22 23 Appendix 10: Subgroup analysis for risk of bias for clinically important 24 25 gastrointestinal bleeding and overt gastrointestinal bleeding 26 27 Appendix 11: Baseline risk of clinically important gastrointestinal bleeding and overt 28 29 gastrointestinal bleeding for each risk group 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 18

https://mc.manuscriptcentral.com/bmj BMJ Page 20 of 109

1 2 3 4 5 1. Cook DJ, Griffith LE, Walter SD, et al. The attributable mortality and length of intensive care unit 6 7 stay of clinically important gastrointestinal bleeding in critically ill patients. Critical care 8 (London, England) 2001;5(6):368-75. 9 2. Barletta JF, Kanji S, MacLaren R, et al. Pharmacoepidemiology of stress ulcer prophylaxis in the 10 United States and Canada. J Crit Care 2014;29(6):955-60. doi:10.1016/j.jcrc.2014.06.025 11 12 3. KragConfidential: M, Perner A, Wetterslev J, et al. Prevalence For andReview outcome of gastrointestinal Only bleeding and use 13 of acid suppressants in acutely ill adult intensive care patients. Intensive Care Med 14 2015;41(5):833-45. doi:10.1007/s00134-015-3725-1 15 16 4. Krag M, Perner A, Wetterslev J, et al. Stress ulcer prophylaxis in the intensive care unit: an 17 international survey of 97 units in 11 countries. Acta Anaesthesiol Scand 2015;59(5):576-85. 18 doi:10.1111/aas.12508 19 20 5. Cook D, Guyatt G. Prophylaxis against Upper Gastrointestinal Bleeding in Hospitalized Patients. 21 The New England journal of medicine 2018;378(26):2506-16. doi:10.1056/NEJMra1605507 22 6. Buendgens L, Bruensing J, Matthes M, et al. Administration of proton pump inhibitors in critically 23 ill medical patients is associated with increased risk of developing Clostridium 24 25 difficile-associated diarrhea. J Crit Care 2014;29(4):696 e11-5. 26 doi:10.1016/j.jcrc.2014.03.002 27 7. Herzig SJ, Howell MD, Ngo LH, et al. Acid-suppressive medication use and the risk for 28 29 hospital-acquired pneumonia. Jama 2009;301(20):2120-8. doi:10.1001/jama.2009.722 30 8. Alhazzani W, Alshamsi F, Belley-Cote E, et al. Efficacy and safety of stress ulcer prophylaxis in 31 critically ill patients: a network meta-analysis of randomized trials. Intensive Care Med 32 33 2018;44(1):1-11. doi:10.1007/s00134-017-5005-8 34 9. Alshamsi F, Belley-Cote E, Cook D, et al. Efficacy and safety of proton pump inhibitors for stress 35 ulcer prophylaxis in critically ill patients: a systematic review and meta-analysis of 36 randomized trials. Critical care (London, England) 2016;20(1):120. 37 38 doi:10.1186/s13054-016-1305-6 39 10. Barbateskovic M, Marker S, Granholm A, et al. Stress ulcer prophylaxis with proton pump 40 inhibitors or histamin-2 receptor antagonists in adult intensive care patients: a systematic 41 42 review with meta-analysis and trial sequential analysis. Intensive Care Med 43 2019;45(2):143-58. doi:10.1007/s00134-019-05526-z 44 11. Huang HB, Jiang W, Wang CY, et al. Stress ulcer prophylaxis in intensive care unit patients 45 46 receiving enteral nutrition: a systematic review and meta-analysis. Critical care (London, 47 England) 2018;22(1):20. doi:10.1186/s13054-017-1937-1 48 12. Huang J, Cao Y, Liao C, et al. Effect of histamine-2-receptor antagonists versus sucralfate on stress 49 ulcer prophylaxis in mechanically ventilated patients: a meta-analysis of 10 randomized 50 51 controlled trials. Critical care (London, England) 2010;14(5):R194. doi:10.1186/cc9312 52 13. Toews I, George AT, Peter JV, et al. Interventions for preventing upper gastrointestinal bleeding in 53 people admitted to intensive care units. Cochrane Database Syst Rev 2018;6:CD008687. 54 55 doi:10.1002/14651858.CD008687.pub2 56 14. Alhazzani W, Guyatt G, Alshahrani M, et al. Withholding Pantoprazole for Stress Ulcer 57 Prophylaxis in Critically Ill Patients: A Pilot Randomized Clinical Trial and Meta-Analysis. 58 59 Critical care medicine 2017;45(7):1121-29. 60 19

https://mc.manuscriptcentral.com/bmj Page 21 of 109 BMJ

1 2 3 15. El-Kersh K, Jalil B, McClave SA, et al. Enteral nutrition as stress ulcer prophylaxis in critically ill 4 5 patients: A randomized controlled exploratory study. Journal of Critical Care 6 2018;43:108-13. 7 16. Gundogan K, Karakoc E, Teke T, et al. Effects of enteral nutrition on stress ulcer hemorrhage in 8 critically ill patients: Multicenter randomized controlled trial. Intensive Care Medicine 9 10 Experimental 2017;Conference:30th Annual Congress of the European Society of Intensive 11 Care Medicine, ESICM 2017. Austria. 5 (2 Supplement 1) (no pagination). 12 Confidential: For Review Only 17. Krag M, Marker S, Perner A, et al. Pantoprazole in Patients at Risk for Gastrointestinal Bleeding in 13 14 the ICU. New England Journal of Medicine 2018;379(23):2199-208. 15 18. Siemieniuk RA, Agoritsas T, Macdonald H, et al. Introduction to BMJ Rapid Recommendations. 16 BMJ 2016;354:i5191. doi:10.1136/bmj.i5191 17 18 19. Wang Y, Ye Z, Ge L, et al. PROTOCOL: Efficacy and safety of stress ulcer prophylaxis in 19 critically ill patients: a systematic review with network meta-analysis. 20 20. Guyatt GH, Busse JW. Modification of Cochrane Tool to assess risk of bias in randomized trials. 21 21. Deeks J, Higgins J, Altman D, et al. Cochrane handbook for systematic reviews of interventions 22 23 version 5.1. 0 (updated March 2011). The Cochrane Collaboration 2011. 24 22. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical 25 test. BMJ 1997;315(7109):629-34. 26 27 23. Ades AE, Sculpher M, Sutton A, et al. Bayesian methods for evidence synthesis in 28 cost-effectiveness analysis. PharmacoEconomics 2006;24(1):1-19. 29 doi:10.2165/00019053-200624010-00001 30 31 24. Lumley T. Network meta-analysis for indirect treatment comparisons. Statistics in medicine 32 2002;21(16):2313-24. doi:10.1002/sim.1201 33 25. Dias S, Sutton AJ, Ades AE, et al. Evidence synthesis for decision making 2: a generalized linear 34 modeling framework for pairwise and network meta-analysis of randomized controlled trials. 35 36 Medical decision making : an international journal of the Society for Medical Decision 37 Making 2013;33(5):607-17. doi:10.1177/0272989x12458724 38 26. Marinho VC, Worthington HV, Walsh T, et al. Fluoride gels for preventing dental caries in children 39 40 and adolescents. Cochrane Database Syst Rev 2015(6):CD002280. 41 doi:10.1002/14651858.CD002280.pub2 42 27. van Valkenhoef G, Dias S, Ades AE, et al. Automated generation of node-splitting models for 43 44 assessment of inconsistency in network meta-analysis. Research synthesis methods 45 2016;7(1):80-93. doi:10.1002/jrsm.1167 46 28. Chaimani A, Salanti G. Using network meta-analysis to evaluate the existence of small-study 47 effects in a network of interventions. Research synthesis methods 2012;3(2):161-76. 48 49 doi:10.1002/jrsm.57 50 29. Brignardello-Petersen R, Murad MH, Walter SD, et al. GRADE approach to rate the certainty from 51 a network meta-analysis: avoiding spurious judgments of imprecision in sparse networks. J 52 53 Clin Epidemiol 2019;105:60-67. doi:10.1016/j.jclinepi.2018.08.022 54 30. Brignardello-Petersen R, Mustafa RA, Siemieniuk RAC, et al. GRADE approach to rate the 55 certainty from a network meta-analysis: Addressing Incoherence. J Clin Epidemiol 2018. 56 57 doi:10.1016/j.jclinepi.2018.11.025 58 31. Chaimani A, Higgins JP, Mavridis D, et al. Graphical tools for network meta-analysis in STATA. 59 PLoS One 2013;8(10):e76654. doi:10.1371/journal.pone.0076654 60 20

https://mc.manuscriptcentral.com/bmj BMJ Page 22 of 109

1 2 3 32. Brignardello-Petersen R, Bonner A, Alexander PE, et al. Advances in the GRADE approach to rate 4 5 the certainty in estimates from a network meta-analysis. J Clin Epidemiol 2018;93:36-44. 6 doi:10.1016/j.jclinepi.2017.10.005 7 33. Guyatt GH, Oxman AD, Vist GE, et al. GRADE: an emerging consensus on rating quality of 8 evidence and strength of recommendations. BMJ 2008;336(7650):924-6. 9 10 doi:10.1136/bmj.39489.470347.AD 11 34. Puhan MA, Schunemann HJ, Murad MH, et al. A GRADE Working Group approach for rating the 12 Confidential: For Review Only quality of treatment effect estimates from network meta-analysis. BMJ 2014;349:g5630. 13 14 doi:10.1136/bmj.g5630 15 35. Balshem H, Helfand M, Schunemann HJ, et al. GRADE guidelines: 3. Rating the quality of 16 evidence. J Clin Epidemiol 2011;64(4):401-6. doi:10.1016/j.jclinepi.2010.07.015 17 18 36. Guyatt GH, Oxman AD, Kunz R, et al. GRADE guidelines 6. Rating the quality of 19 evidence--imprecision. J Clin Epidemiol 2011;64(12):1283-93. 20 doi:10.1016/j.jclinepi.2011.01.012 21 37. Guyatt GH, Oxman AD, Kunz R, et al. GRADE guidelines: 8. Rating the quality of 22 23 evidence--indirectness. J Clin Epidemiol 2011;64(12):1303-10. 24 doi:10.1016/j.jclinepi.2011.04.014 25 38. Guyatt GH, Oxman AD, Kunz R, et al. GRADE guidelines: 7. Rating the quality of 26 27 evidence--inconsistency. J Clin Epidemiol 2011;64(12):1294-302. 28 doi:10.1016/j.jclinepi.2011.03.017 29 39. Guyatt GH, Oxman AD, Montori V, et al. GRADE guidelines: 5. Rating the quality of 30 31 evidence--publication bias. J Clin Epidemiol 2011;64(12):1277-82. 32 doi:10.1016/j.jclinepi.2011.01.011 33 40. Guyatt GH, Oxman AD, Vist G, et al. GRADE guidelines: 4. Rating the quality of evidence--study 34 limitations (risk of bias). J Clin Epidemiol 2011;64(4):407-15. 35 36 doi:10.1016/j.jclinepi.2010.07.017 37 41. Apte NM, Karnad DR, Medhekar TP, et al. Gastric colonization and pneumonia in intubated 38 critically ill patients receiving stress ulcer prophylaxis: a randomized, controlled trial. Crit 39 40 Care Med 1992;20(5):590-3. 41 42. Bashar FR, Manuchehrian N, Mahmoudabadi M, et al. Effects of ranitidine and pantoprazole on 42 ventilator-associated pneumonia: a randomized double-blind clinical trial. Tanaffos 43 44 2013;12(2):16-21. 45 43. Ben-Menachem T, Fogel R, Patel RV, et al. Prophylaxis for stress-related gastric hemorrhage in the 46 medical intensive care unit. A randomized, controlled, single-blind study. Annals of internal 47 medicine 1994;121(8):568-75. 48 49 44. Burgess P, Larson GM, Davidson P, et al. Effect of ranitidine on intragastric pH and stress-related 50 upper gastrointestinal bleeding in patients with severe head injury. Dig Dis Sci 51 1995;40(3):645-50. 52 53 45. Cannon LA, Heiselman D, Gardner W, et al. Prophylaxis of upper gastrointestinal tract bleeding in 54 mechanically ventilated patients. A randomized study comparing the efficacy of sucralfate, 55 cimetidine, and antacids. Arch Intern Med 1987;147(12):2101-6. 56 57 46. Chan KH, Lai EC, Tuen H, et al. Prospective double-blind placebo-controlled randomized trial on 58 the use of ranitidine in preventing postoperative gastroduodenal complications in high-risk 59 neurosurgical patients. Journal of neurosurgery 1995;82(3):413-7. 60 21

https://mc.manuscriptcentral.com/bmj Page 23 of 109 BMJ

1 2 3 doi:10.3171/jns.1995.82.3.0413 4 5 47. Conrad SA, Gabrielli A, Margolis B, et al. Randomized, double-blind comparison of 6 immediate-release omeprazole oral suspension versus intravenous cimetidine for the 7 prevention of upper gastrointestinal bleeding in critically ill patients. Crit Care Med 8 2005;33(4):760-5. 9 10 48. Cook D, Guyatt G, Marshall J, et al. A comparison of sucralfate and ranitidine for the prevention of 11 upper gastrointestinal bleeding in patients requiring mechanical ventilation. Canadian Critical 12 Confidential: For Review Only Care Trials Group. The New England journal of medicine 1998;338(12):791-7. 13 14 doi:10.1056/nejm199803193381203 15 49. De Azevedo JRA, Soares MDGA, Silva GAE, et al. Prevention of stress ulcer bleeding in high risk 16 patients. Comparison of three drugs. GED - Gastrenterologia Endoscopia Digestiva 17 18 2000;19(6):239-44. 19 50. Eddleston JM, Pearson RC, Holland J, et al. Prospective endoscopic study of stress erosions and 20 ulcers in critically ill adult patients treated with either sucralfate or placebo. Crit Care Med 21 1994;22(12):1949-54. 22 23 51. Fink M, Karlstadt RG, Maroko RT, et al. Intravenous pantoprazole (IVP) and continuous infusion 24 cimetidine (C) prevent upper gastrointestinal bleeding (UGIB) regardless of APSII score 25 (APACHE II) in high risk intensive care unit (ICU) patients. Gastroenterology 26 27 2003;124(4):A625-A26. doi:Doi 10.1016/S0016-5085(03)83170-0 28 52. Groll A, Simon JB, Wigle RD, et al. Cimetidine prophylaxis for gastrointestinal bleeding in an 29 intensive care unit. Gut 1986;27(2):135-40. 30 31 53. Gursoy O, Memis D, Sut N. Effect of proton pump inhibitors on gastric juice volume, gastric pH 32 and gastric intramucosal pH in critically ill patients : a randomized, double-blind, 33 placebo-controlled study. Clin Drug Investig 2008;28(12):777-82. 34 doi:10.2165/0044011-200828120-00005 35 36 10.2165/0044011-200828120-00005. 37 54. Halloran LG, Zfass AM, Gayle WE, et al. Prevention of acute gastrointestinal complications after 38 severe head injury: a controlled trial of cimetidine prophylaxis. Am J Surg 1980;139(1):44-8. 39 40 55. Hanisch EW, Encke A, Naujoks F, et al. A randomized, double-blind trial for stress ulcer 41 prophylaxis shows no evidence of increased pneumonia. Am J Surg 1998;176(5):453-7. 42 56. Harlaftis N, Basdanis G, Papapolychroniadis C, et al. Nosocomial pneumonia in mechanically 43 44 ventilated patients during stress ulcer prophylaxis with sucralfate and ranitidine. Hell J 45 Gastroenterol 1997;10:230-35. 46 57. Jakob SM, Parviainen I, Ruokonen E, et al. Lack of effect of ranitidine on gastric luminal pH and 47 mucosal PCO2 during the first day in the ICU. Acta Anaesthesiol Scand 2005;49(3):390-6. 48 49 doi:10.1111/j.1399-6576.2005.00651.x 50 58. Kantorova I, Svoboda P, Scheer P, et al. Stress ulcer prophylaxis in critically ill patients: a 51 randomized controlled trial. Hepatogastroenterology 2004;51(57):757-61. 52 53 59. Karlstadt RG, Iberti TJ, Silverstein J, et al. Comparison of cimetidine and placebo for the 54 prophylaxis of upper gastrointestinal bleeding due to stress-related gastric mucosal damage in 55 the intensive care unit. Journal of Intensive Care Medicine 1990;5(1):26-32. 56 57 60. Kaushal S, Midha V, Sood A, et al. A comparative study of the effects of famotidine and sucralfate 58 in prevention of upper gastro-intestinal bleeding in patients of head injury. Indian Journal of 59 Pharmacology 2000;32(3):246-49. 60 22

https://mc.manuscriptcentral.com/bmj BMJ Page 24 of 109

1 2 3 61. Khorvash F, Abbasi S, Meidani M, et al. The comparison between proton pump inhibitors and 4 5 sucralfate in incidence of ventilator associated pneumonia in critically ill patients. Advanced 6 biomedical research 2014;3:52. doi:10.4103/2277-9175.125789 7 62. Labattut AG, Santolalla PM, De Andres AP, et al. Efficacy of sucralfate in the prevention of upper 8 gastrointestinal stress bleeding in intensive care patients: comparison vs a control group. Clin 9 10 Intensive Care 1992;3(5):S19-S25. 11 63. Laggner AN, Lenz K, Base W, et al. Prevention of upper gastrointestinal bleeding in long-term 12 Confidential: For Review Only ventilated patients. Sucralfate versus ranitidine. The American journal of medicine 13 14 1989;86(6A):81-4. 15 64. Laggner AN, Lenz K, Graninger W, et al. Stress ulcer prophylaxis in a general intensive care unit: 16 sucralfate versus ranitidine. Der Anaesthesist 1988;37:704-10. 17 18 65. Lee TH, Hung FM, Yang LH. Comparison of the efficacy of esomeprazole and famotidine against 19 stress ulcers in a neurosurgical intensive care unit. Advances in Digestive Medicine 20 2014;1(2):50-53. doi:10.1016/j.aidm.2013.06.001 21 66. Levy MJ, Seelig CB, Robinson NJ, et al. Comparison of omeprazole and ranitidine for stress ulcer 22 23 prophylaxis. Dig Dis Sci 1997;42(6):1255-9. 24 67. Lin CC, Hsu YL, Chung CS, et al. Stress ulcer prophylaxis in patients being weaned from the 25 ventilator in a respiratory care center: A randomized control trial. Journal of the Formosan 26 27 Medical Association 2016;115(1):19-24. 28 68. Liu BL, Li B, Zhang X, et al. A randomized controlled study comparing omeprazole and cimetidine 29 for the prophylaxis of stress-related upper gastrointestinal bleeding in patients with 30 31 intracerebral hemorrhage. Journal of neurosurgery 2013;118(1):115-20. 32 doi:10.3171/2012.9.jns12170 33 69. Lou W, Xia Y, Xiang P, et al. Prevention of upper gastrointestinal bleeding in critically ill Chinese 34 patients: a randomized, double-blind study evaluating esomeprazole and cimetidine. Current 35 36 medical research and opinion 2018:1‐7. doi:10.1080/03007995.2018.1464132 37 70. Macdougall BR, Bailey RJ, Williams R. H2-receptor antagonists and antacids in the prevention of 38 acute gastrointestinal haemorrhage in fulminant hepatic failure. Two controlled trials. Lancet 39 40 (London, England) 1977;1:617-19. 41 71. Martin LF, Booth FV, Karlstadt RG, et al. Continuous intravenous cimetidine decreases 42 stress-related upper gastrointestinal hemorrhage without promoting pneumonia. Crit Care 43 44 Med 1993;21(1):19-30. 45 72. Nielsen HJ, Witt K, Moesgaard F, et al. Ranitidine for improvement of delayed hypersensitivity 46 response in patients with sepsis. Acta Chir Scand 1989;155(9):445-49. 47 73. Nourian A, Mohammadi M, Beigmohammadi MT, et al. Comparing efficacy of enteral nutrition 48 49 plus ranitidine and enteral nutrition alone as stress ulcer prophylaxis. Journal of Comparative 50 Effectiveness Research 2018;7(5):493-501. doi:https://dx.doi.org/10.2217/cer-2017-0098 51 74. Peura DA, Johnson LF. Cimetidine for prevention and treatment of gastroduodenal mucosal lesions 52 53 in patients in an intensive care unit. Annals of internal medicine 1985;103(2):173-7. 54 75. Pickworth KK, Falcone RE, Hoogeboom JE, et al. Occurrence of nosocomial pneumonia in 55 mechanically ventilated trauma patients: a comparison of sucralfate and ranitidine. Crit Care 56 57 Med 1993;21(12):1856-62. 58 76. Powell H, Morgan M, Li SK, et al. Inhibition of gastric acid secretion in the intensive care unit after 59 coronary artery bypass graft. A pilot control study of intravenous omeprazole by bolus and 60 23

https://mc.manuscriptcentral.com/bmj Page 25 of 109 BMJ

1 2 3 infusion, ranitidine and placebo. Theoretical Surgery 1993;8(3):125-30. 4 5 77. Prakash S, Rai A, Gogia AR, et al. Nosocomial pneumonia in mechanically ventilated patients 6 receiving ranitidine or sucralfate as stress ulcer prophylaxis. Indian Journal of Anaesthesia 7 2008;52(2):179-84. 8 78. Prod'hom G, Leuenberger P, Koerfer J, et al. Nosocomial pneumonia in mechanically ventilated 9 10 patients receiving antacid, ranitidine, or sucralfate as prophylaxis for stress ulcer. A 11 randomized controlled trial. Annals of internal medicine 1994;120(8):653-62. 12 Confidential: For Review Only 79. Reusser P, Gyr K, Scheidegger D, et al. Prospective endoscopic study of stress erosions and ulcers 13 14 in critically ill neurosurgical patients: current incidence and effect of acid-reducing 15 prophylaxis. Crit Care Med 1990;18(3):270-4. 16 80. Ruiz-Santana S, Ortiz E, Gonzalez B, et al. Stress-induced gastroduodenal lesions and total 17 18 parenteral nutrition in critically ill patients: frequency, complications, and the value of 19 prophylactic treatment. A prospective, randomized study. Crit Care Med 1991;19(7):887-91. 20 81. Ryan P, Dawson J, Teres D, et al. Nosocomial pneumonia during stress ulcer prophylaxis with 21 cimetidine and sucralfate. Archives of surgery (Chicago, Ill. : 1960) 1993;128(12):1353-7. 22 23 82. Selvanderan SP, Summers MJ, Finnis ME, et al. Pantoprazole or Placebo for Stress Ulcer 24 Prophylaxis (POP-UP): Randomized Double-Blind Exploratory Study. Critical care medicine 25 2016;44(10):1842-50. 26 27 83. Solouki M, Mar'ashian SM, Koochak M, et al. Ventilator-associated pneumonia among ICU 28 patients receiving mechanical ventilation and prophylaxis of gastrointestinal bleeding. Iranian 29 Journal of Clinical Infectious Diseases 2009;4(3):177-80. 30 31 84. Solouki M, Marashian SM, Kouchak M, et al. Comparison between the preventive effects of 32 ranitidine and omeprazole on upper gastrointestinal bleeding among ICU patients. Tanaffos 33 2009;8(4):37-42. 34 85. Somberg L, Morris J, Fantus R, et al. Intermittent intravenous pantoprazole and continuous 35 36 cimetidine infusion: effect on gastric pH control in critically ill patients at risk of developing 37 stress-related mucosal disease. The Journal of trauma, 2008:1202-10. 38 86. Spapen H, Diltoer M, Nguyen DN, et al. One week treatment with cimetidine does not attenuate the 39 40 cortisol response to a short corticotropin test in stable intensive care patients: a prospective, 41 randomized, and controlled study. Acta Anaesthesiol Belg 1995;46(3-4):133-40. 42 87. Thomason MH, Payseur ES, Hakenewerth AM, et al. Nosocomial pneumonia in ventilated trauma 43 44 patients during stress ulcer prophylaxis with sucralfate, antacid, and ranitidine. The Journal of 45 trauma 1996;41(3):503-8. 46 88. Tryba M, Zevounou F, Torok M, et al. Prevention of acute stress bleeding with sucralfate, antacids, 47 or cimetidine. A controlled study with pirenzepine as a basic medication. The American 48 49 journal of medicine 1985;79(2C):55-61. 50 89. Zinner MJ, Zuidema GD, Smith P, et al. The prevention of upper gastrointestinal tract bleeding in 51 patients in an intensive care unit. Surg Gynecol Obstet 1981;153:214-20. 52 53 90. Sun X, Briel M, Walter SD, et al. Is a subgroup effect believable? Updating criteria to evaluate the 54 credibility of subgroup analyses. BMJ 2010;340:c117. doi:10.1136/bmj.c117 55 91. Darlong V, Jayalakhsmi TS, Kaul HL, et al. Stress ulcer prophylaxis in patients on ventilator. 56 57 Tropical gastroenterology : official journal of the Digestive Diseases Foundation 58 2003;24(3):124-8. 59 92. Fabian TC, Boucher BA, Croce MA, et al. Pneumonia and stress ulceration in severely injured 60 24

https://mc.manuscriptcentral.com/bmj BMJ Page 26 of 109

1 2 3 patients. A prospective evaluation of the effects of stress ulcer prophylaxis. Archives of 4 5 surgery (Chicago, Ill. : 1960) 1993;128(2):185-91; discussion 91-2. 6 93. Fogas JF, Kiss KK, Gyura FG, et al. Effects of proton pump inhibitor versus H2-receptor antagonist 7 stress ulcer prophylaxis on ventilator-associated pneumonia: A pilot study. Critical Care 8 2013;17((Fogas J.F.; Kiss K.K.; Gyura F.G.; Tóbiás Z.T.; Molnár Z.M.) University of Szeged, 9 10 Hungary):S150-S51. doi:10.1186/cc12340 11 94. Hummer-Sigiel M, Jacquier A, Girard A, et al. Ranitidine pour la prophylaxie de l'ulcére de stress 12 Confidential: For Review Only chexz les traumatisés crâniens graves. Ann Med Nancy l'Est 1986;25:101-03. 13 14 95. Kotlyanskaya A, Luka B, Mukherji R, et al. A comparison trial of lansoprazole disintegrating tablet, 15 lansoprazole suspension or ranitidine for stress ulcer prophylaxis in critically ill patients. Crit 16 Care Med 2007;35(12):A194-A94. 17 18 96. Phillips J, Metzler M, Huckfeldt R, et al. A multicenter, prospective, randomized clinical trial of 19 continuous infusion i.v. ranitidine vs. omeprazole suspension in the prophylaxis of stress 20 ulcers. Crit Care Med 1998;26(1):101A. 21 97. Simms HH, DeMaria E, McDonald L, et al. Role of gastric colonization in the development of 22 23 pneumonia in critically ill trauma patients: results of a prospective randomized trial. The 24 Journal of trauma 1991;31(4):531-6; discussion 36-7. 25 98. Wee B, Liu CH, Cohen H, et al. IV famotidine vs. IV pantoprazole for stress ulcer prevention in the 26 27 ICU: A prospective study. Crit Care Med 2013;41(12):A181. 28 doi:10.1097/01.ccm.0000439969.36301.c9 29 99. Bhanot RD. Nosocomial pneumonia in mechanically ventilated patients receiving ranitidine, 30 31 omeprazole or sucralfate as stress ulcer prophylaxis. American Journal of Respiratory and 32 Critical Care Medicine 2010;181(1). 33 100. Metz CA, Livingston DH, Smith JS, et al. Impact of multiple risk factors and ranitidine 34 prophylaxis on the development of stress-related upper gastrointestinal bleeding: a prospective, 35 36 multicenter, double-blind, randomized trial. The Ranitidine Head Injury Study Group. Crit 37 Care Med 1993;21(12):1844-9. 38 101. Mustafa NA, Akturk G, Ozen I, et al. Acute stress bleeding prophylaxis with sucralfate versus 39 40 ranitidine and incidence of secondary pneumonia in intensive care unit patients. Intensive 41 Care Med 1995;21(3):287. 42 102. Basso N, Bagarani M, Materia A, et al. Cimetidine and antacid prophylaxis of acute upper 43 44 gastrointestinal bleeding in high risk patients. Controlled, randomized trial. Am J Surg 45 1981;141(3):339-41. 46 103. Brophy GM, Brackbill ML, Bidwell KL, et al. Prospective, randomized comparison of 47 lansoprazole suspension, and intermittent intravenous famotidine on gastric pH and acid 48 49 production in critically ill neurosurgical patients. Neurocrit Care 2010;13(2):176-81. 50 doi:10.1007/s12028-010-9397-3 51 104. Cartier F, Gauthier-Lafaye P, Lareng L, et al. Cimetideine in patients at risk of stress ulcers: a 52 53 multi-centre controlled trial. Intensive Care Medicine 1980;6:54. 54 105. Friedman CJ, Oblinger MJ, Suratt PM, et al. Prophylaxis of upper gastrointestinal hemorrhage in 55 patients requiring mechanical ventilation. Crit Care Med 1982;10(5):316-9. 56 57 106. Larson GM, Davidson P, Brown J, et al. Comparison of ranitidine versus placebo on 24-hour 58 gastric Ph and upper gastrointestinal (UGI) bleeding in head injury patients. The American 59 journal of gastroenterology 1989;84:1165. 60 25

https://mc.manuscriptcentral.com/bmj Page 27 of 109 BMJ

1 2 3 107. Luk GD, Summer WR, Messersmith JF. Cimetidine and antacid in prophylaxis of acute 4 5 gastrointestinal bleeding: a randomized, doubleblind, controlled study. Gastroenterology 6 1982;82:1121. 7 108. Pan X, Zhang W, Li Z, et al. The preventive effects of rabeprazole on upper gastrointestinal tract 8 hemorrhage in patients with severe acute pancreatitis. [Chinese]. Chinese Journal of 9 10 Gastroenterology, 2004:30-2. 11 109. van den Berg B, van Blankenstein M. Prevention of stress-induced upper gastrointestinal bleeding 12 Confidential: For Review Only by cimetidine in patients on assisted ventilation. Digestion 1985;31(1):1-8. 13 14 doi:10.1159/000199170 15 110. Brignardello-Petersen R, Mustafa RA, Siemieniuk RAC, et al. GRADE approach to rate the 16 certainty from a network meta-analysis: addressing incoherence. J Clin Epidemiol 17 18 2019;108:77-85. doi:10.1016/j.jclinepi.2018.11.025 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 26

https://mc.manuscriptcentral.com/bmj BMJ Page 28 of 109

1 2 3 4 5 Table 1 Low versus high risk of bias results for clinically important gastrointestinal bleeding (CIB) 6 7 Comparison Low risk of bias result High risk of bias result* Subgroup difference (p value) 8 PPIs versus placeboConfidential:0.62 (0.42 to 0.90) For0.33 (0.01 Review to 8.21) Only0.70 9 H2RAs versus placebo 0.43 (0.18 to 1.01) 0.54 (0.26 to 1.10) 0.69 10 11 Sucralfate versus placebo 3.36 (0.34 to 33.13) 0.63 (0.28 to 1.41) 0.18 12 PPIs versus H2RAs 0.58 (0.29 to 1.16) 0.20 (0.07 to 0.54) 0.09† 13 PPIs versus sucralfate 0.31 (0.03 to 3.05) 0.27 (0.01 to 6.93) 0.95 14 H2RAs versus sucralfate 0.46 (0.23 to 0.91) 1.15 (0.64 to 2.08) 0.05† 15 16 *Studies were at high risk of bias if either allocation sequence concealment or blinding were at high risk of bias. 17 †Important difference between low and high risk of bias result, so we used low risk of bias result as best estimate. 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 27 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 29 of 109 BMJ

1 2 3 4 5Table 2 GRADE summary of findings for clinically important gastrointestinal bleeding (CIB) for different comparisons 6 7Comparison Study results (95% CI) Absolute effect estimates (per Absolute Certainty Plain text summary 8 and measurementsConfidential:1000) For differenceReview (95% in effectOnly 9 CI) (per 1000) estimates 10PPIs versus placebo OR 0.61 (0.42 to 0.89). Low risk Placebo: PPIs: 7 -5 (-7 to -1) Moderate* PPIs probably reduce CIB by less than the 11 12 4324 patients in 8 12 amount most people would need to choose a 13 studies. PPI 14 Moderate risk 30 19 -11 (-17 to -3) Low*† PPIs may reduce CIB by less than the 15 16 amount most people would need to choose a 17 PPI 18 High risk 60 37 -23 (-34 to -6) Moderate† PPIs probably reduce CIB 19 20 Highest risk 90 57 -33 (-50 to -9) Moderate† PPIs probably reduce CIB 21H2RAs versus OR 0.46 (0.27 to 0.79). Low risk Placebo: H2RAs: -6 (-9 to -2) Moderate* H2RAs probably reduce CIB by less than 22placebo 1248 patients in 14 12 6 the amount most people would need to 23 24 studies. choose a H2RA 25 Moderate risk 30 14 -16 (-22 to -6) Low*† H2RAs may reduce CIB by less than the 26 amount most people would need to choose a 27 28 H2RA 29 High risk 60 29 -31 (-43 to -12) Moderate† H2RAs probably reduce CIB 30 Highest risk 90 44 -46 (-64 to -18) Moderate† H2RAs probably reduce CIB 31 32Sucralfate versus OR 0.76 (0.36 to 1.62). Low risk Placebo: Suc: 9 -3 (-8 to 7) Moderate* Sucralfate probably does not have an 33placebo 874 patients in 6 studies. 12 important effect 34 Moderate risk 30 23 -7 (-19 to 18) Low*† Sucralfate may not have an important effect 35 36 High risk 60 46 -14 (-38 to 34) Low‡ Sucralfate may not have an important effect 37 Highest risk 90 70 -20 (-56 to 48) Low‡ Sucralfate may not have an important effect 38 39 40 28 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 30 of 109

1 2 3 4 5PPIs versus H2RAs OR 0.58 (0.29 to 1.16). Low risk H2RAs: PPIs: 7§ -5 (-17 to 1) Low*† There may be no important difference 6 1021 patients in 5 studies 12 7 with low risk of bias. Moderate risk 32 19§ -13 (-44 to 3) Low*† PPIs may reduce CIB more than H2RAs 8 Confidential: For Review Only 9 High risk 62 37§ -25 (-80 to 5) Moderate† PPIs probably reduce CIB more than H2RAs 10 Highest risk 94 57§ -37 (-116 to 8) Moderate† PPIs probably reduce CIB more than H2RAs 11 12PPIs versus OR 0.30 (0.05 to 1.92). Low risk Suc: 23 PPIs: 7§ -16 (-117 to 3) Very low*‡ Whether there is an important difference or 13sucralfate 211 patients in 2 studies. not is very uncertain 14 Moderate risk 61 19§ -42 (-260 to 9) Very low*‡ Whether there is an important difference or 15 16 not is very uncertain 17 High risk 113 37§ -76 (-398 to 17) Low‡ PPIs may reduce CIB compared with 18 sucralfate 19 Highest risk 168 57§ -111 (-490 to 27) Low‡ PPIs may reduce CIB compared with 20 21 sucralfate 22H2RAs versus OR 0.46 (0.23 to 0.91). Low risk Suc: 13 H2RAs: -7 (-20 to -1) Low*† There may be no important difference 23sucralfate 1340 patients in 2 studies 6¶ 24 25 with low risk of bias. Moderate risk 30 14¶ -16 (-44 to -1) Low*† H2RAs may reduce CIB compared with 26 sucralfate 27 High risk 61 29¶ -32 (-86 to -3) Moderate† H2RAs probably reduce CIB compared with 28 29 sucralfate 30 Highest risk 91 44¶ -47 (-123 to -4) Moderate† H2RAs probably reduce CIB compared with 31 sucralfate 32 33CI=confidence interval; GI= gastrointestinal; PPIs=proton pump inhibitors; OR=odds ratio; H2RAs=histamine-2 receptor antagonists. 34*Rated down due to uncertainty in baseline risk for some risk factors. 35†Rated down for imprecision. 36 37‡Rated down 2 levels for imprecision. 38§We used the point estimate of absolute effect for PPIs, obtained from PPIs versus placebo, to calculate absolute effect for PPIs versus H2RAs and PPIs versus sucralfate. 39 40 29 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 31 of 109 BMJ

1 2 3 4 5¶We used the point estimate of absolute effect for H2RAs, obtained from H2RAs versus placebo, to calculate absolute effect for H2RAs versus sucralfate. 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38Table 3 GRADE summary of findings for pneumonia for different comparisons 39 40 30 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 32 of 109

1 2 3 4 5Comparison Direct estimate (95% Indirect estimate (95% Network estimate (95% Absolute effect estimates† (95% CrI) (per 1000) 6 CrI); Certainty of CrI); Certainty of CrI); Certainty of 7 evidence evidence evidence* 8 Confidential: For Review Only 9PPIs versus placebo 1.09 (0.77 to 1.76); 1.74 (0.96 to 3.19); 1.38 (0.97 to 2.08); Placebo: 162 PPIs: 210 (159 to 287) 10 Moderate‡ Moderate‡ Low‡§¶ Difference: 48 (-3 to 125) 11 12H2RAs versus placebo 1.32 (0.87 to 2.00); 0.93 (0.50 to 2.00); 1.26 (0.90 to 1.86); Placebo: 162 H2RAs: 196 (148 to 263) 13 Moderate‡ Moderate‡ Low‡§¶ Difference: 34 (-14 to 101) 14Sucralfate versus placebo 2.11 (0.66 to 6.49); 0.71 (0.43 to 1.14); 0.87 (0.57 to 1.35); Placebo: 162 Sucralfate: 144 (100 to 15 16 Moderate‡ Moderate‡ Low‡§ 208) 17 Difference: -18 (-62 to 46) 18PPIs versus H2RAs 1.09 (0.79 to 1.49); 1.01 (0.57 to 1.90); 1.10 (0.81 to 1.49); H2RAs: 196 (148 to 263) PPIs: 210 (159 to 287) 19 Moderate‡ Moderate‡ Low‡§ 20 Difference: 15 (-35 to 67) 21PPIs versus sucralfate 2.67 (1.29 to 5.56); 1.24 (0.82 to 2.01); 1.58 (1.09 to 2.39); Sucralfate: 144 (100 to PPIs: 210 (159 to 287) 22 Moderate‡ Moderate‡ Low‡§ 208) 23 24 Difference: 65 (12 to 129) 25H2RAs versus sucralfate 1.32 (1.02 to 1.80); 2.48 (0.80 to 7.62); 1.44 (1.09 to 2.00); Sucralfate: 144 (100 to H2RAs: 196 (148 to 263) 26 Moderate‡ Moderate‡ Low‡§ 208) 27 28 Difference: 51 (12 to 96) 29CrI=credible interval; PPIs=proton pump inhibitors; H2RAs=histamine-2 receptor antagonists. 30*Higher of direct or indirect confidence (without consider imprecision), and then considered imprecision and incoherence. 31†We used as baseline risk in the placebo group of the SUP-ICU trial. 32 33‡Rated down for risk of bias. 34§Rated down for imprecision. 35¶We are skeptical of the result because the pooled result including smaller studies conflicts with the evidence from the largest study (SUP-ICU). 36 37 38 39 40 31 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 33 of 109 BMJ

1 2 3 4 5 Table 4 Low versus high risk of bias results for overt gastrointestinal bleeding 6 7 Comparison Low risk of bias result High risk of bias result* Subgroup difference (p value) 8 PPIs versus placeboConfidential:0.59 (0.45 to 0.77) For0.20 (0.09 Review to 0.48) Only0.02† 9 H2RAs versus placebo 0.34 (0.19 to 0.61) 0.45 (0.23 to 0.88) 0.54 10 11 Sucralfate versus placebo 3.36 (0.34 to 33.13) 0.52 (0.28 to 0.95) 0.12 12 PPIs versus H2RAs 0.54 (0.30 to 0.99) 0.32 (0.16 to 0.65) 0.27 13 PPIs versus sucralfate 0.31 (0.03 to 3.05) 0.11 (0.01 to 2.20) 0.59 14 H2RAs versus sucralfate 0.46 (0.23 to 0.91) 1.44 (0.99 to 2.10) 0.004† 15 16 *For PPIs versus placebo, studies were at high risk of bias if either allocation sequence concealment or blinding or missing outcome data 17 were at high risk of bias. For other comparisons, studies were at high risk of bias if either allocation sequence concealment or blinding were at 18 high risk of bias. 19 20 †Important difference between low and high risk of bias result, so we used low risk of bias result as best estimate. 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 32 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 34 of 109

1 2 3 4 5 Randomised controlled Records identified from Potentially eligible 6 7 trials identified from updated literature search studies identified from 8 previous NMA (n=57) (n=421) relevant reviews (n=13) 9 10 11 12 Confidential: For Review Only 13 Excluded due to Records after duplicates removed (n=378) 14 inappropriate 15 population (n=3) 16 Records excluded (n=308) 17 18 19 Full text articles assessed for eligibility (n=70) 20 Trials included from 21 previous NMA (n=54) 22 Full text articles excluded (n=52): 23 24 Not randomised controlled trial (n=28) 25 Inappropriate population (n=2) 26 Inappropriate comparisons (n=5) 27 28 No data provided (n=13) 29 Not adult (n=4) 30 31 32 Trials included from the new search (n=18) 33 34 35 36 Total randomised controlled trials included (n=72) 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 35 of 109 BMJ

1 2 3 PPI 4 5 Confidential: For Review3564 Only 6 7 8 9 10 11 12 10 trials 20 trials 13 14 15 4 trials 16 17 18 31 trials 19 Placebo or no 20 H2RA 21 prophylaxis 22 3669 23 3531 24 25 26 27 28 29 7 trials 21 trials 30 31 32 33 34 Sucralfate 35 36 1896 37 38 39 https://mc.manuscriptcentral.com/bmj 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 36 of 109

1 2 3 4 5 6 7 8 9 10 Confidential: For Review Only 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 37 of 109 BMJ

1 2 3 4 Appendix 1: Search terms and strategies 5 6 7 8 MEDLINE, EMBASE and PUBMED epublications 9 10 1 Proton Pump Inhibitors/ or (proton adj5 pump adj5 inhibitor*).mp. or ppi*.tw. 11 or exp Omeprazole/ or Benzimidazoles/ or Omeprazole.mp. or losec.mp. or omez.mp. 12 Confidential: For Review Only 13 or prilosec.mp. or rapinex.mp. or zegerid.mp. or esomeprazole.mp. or axagon.mp. or 14 15 esporal.mp. or lucen.mp. or nexiam.mp. or nexium.mp. or sompraz.mp. or soleri.mp. 16 17 or Lansoprazole.mp. or agopton.mp. or bamalite.mp. or lansoprazole.mp. or lanzor.mp. 18 19 or monolitum.mp. or ogast.mp. or ogastro.mp. or ogastro.mp. or opiren.mp. or 20 21 prevacid.mp. or prezal.mp. or pro ulco.mp. or promeco.mp. or takepron.mp. or 22 23 ulpax.mp. or zoton.mp. or 2-Pyridinylmethylsulfinylbenzimidazoles/ or 24 25 dexlansoprazole.mp. or kapidex.mp. or dexilant.mp. or pantoprazole.mp. or 26 27 pantoloc.mp. or pantecta.mp. or protium.mp. or protonix.mp. or rabeprazole.mp. or 28 29 AcipHex.mp. or Pariet.mp. or dexrabeprazole.mp. or rablet.mp. or Lansoprazol.mp. or 30 31 Esomeprazol.mp. or Prevacid.mp. or Dexilant.mp. or Prilosec.mp. or Zegerid.mp. or 32 33 Omeprazol.mp. or Pantroprazol.mp. or "TAK 390MR".mp. or "AG 1749".mp. or "BY 34 35 1023".mp. 36 37 2 exp Histamine H2 Antagonists/ or ((H2 or histamine) adj3 (blocker* or 38 39 blockader*)).mp. or ((H2 or histamine) adj3 (receptor* or antagonist*)).mp. or (H2 adj1 40 41 antihistaminic*).mp. or H2RA*.mp. or exp Cimetidine/ or Guanidines/ or Imidazoles/ 42 43 or cimetidine.mp. or acitak.mp. or altramet.mp. or biomet.mp. or dyspamet.mp. or 44 45 eureceptor.mp. or galenamet.mp. or histodil.mp. or peptimax.mp. or phimetin.mp. or 46 47 tagamet.mp. or ultec.mp. or zita.mp. or FAMOTIDINE/ or famotidine.mp. or fluxid.mp. 48 49 or leader acid reducer.mp. or mylanta.mp. or pepcid.mp. or Thiazoles/ or roxatidine.mp. 50 or Roxit.mp. or NIZATIDINE/ or Nizatidine.mp. or axid.mp. or tazac.mp. or zinga.mp. 51 52 or RANITIDINE/ or ranitidine.mp. or azanplus.mp. or biotidin.mp. or pylorid.mp. or 53 54 raciran.mp. or raniberl.mp. or ranisen.mp. or ranitidin.mp. or rantec.mp. or sostril.mp. 55 56 or taladine.mp. or tritec.mp. or wal-zan.mp. or zantac.mp. or Furans/ or Burimamide/ 57 58 or Metiamide/ or Zantic.mp. 59 60 3 exp Peptic Ulcer/ or Ulcer*.mp. or lesion*.mp.

https://mc.manuscriptcentral.com/bmj BMJ Page 40 of 109

1 2 3 4 4 (bleed* or re-bleed* or rebleed* or h?emorrhag*).mp. 5 6 5 3 and 4 7 8 6 (stress adj3 ulcer*).mp. 9 10 7 5 or 6 11 8 exp Gastrointestinal Hemorrhage/ 12 Confidential: For Review Only 13 9 ((gastrointestinal or gastro-intestinal) adj5 (bleed* or re-bleed* or rebleed* or 14 15 h?emorrhag*)).mp. 16 17 10 (h?ematochezia* or h?ematemese* or mel?ena*).mp. 18 19 11 exp injury/ and exp gastric mucosa/ 20 21 12 (mucos* adj5 injur*).tw. 22 23 13 (stomach or antrum or antral or pyloric or pylorus or gastri* or gastropathy or 24 25 epigastr* or duodenal or duodenum or gastro-duodenal or gastroduodenal or oeso*ag* 26 27 or esp*ag* or "upper GI" or UGI or "upper gastrointestinal" or "upper 28 29 gastrointestinal").mp. 30 31 14 4 and 13 32 33 15 exp Gastritis/ 34 35 16 4 and 15 36 37 17 or/7-12,14,16 38 39 18 exp Critical Care/ 40 41 19 exp intensive care 42 43 20 exp Critical Illness/ 44 45 21 exp Intensive Care Units/ 46 47 22 ICU*.tw. 48 49 23 ((critical* or intensive) adj3 (care or ill*)).tw. 50 24 exp Intubation, Gastrointestinal/ 51 52 25 exp Monitoring, Physiologic/ 53 54 26 exp Multiple Organ Failure/ 55 56 27 exp Acid-Base Equilibrium/ 57 58 28 exp Multiple Trauma/ 59 60 29 (serious* adj injur*).tw.

https://mc.manuscriptcentral.com/bmj Page 41 of 109 BMJ

1 2 3 4 30 (severe adj (traum* or shock)).tw. 5 6 31 exp Perioperative Care/ 7 8 32 ((preoperative or intraoperative or perioperative) adj (care or procedure* or 9 10 period)).tw. 11 33 exp Resuscitation/ 12 Confidential: For Review Only 13 34 exp Shock/ 14 15 35 exp sepsis/ 16 17 36 exp Ventilator Weaning/ 18 19 37 exp Ventilators, Mechanical/ 20 21 38 exp Ventilators, Negative-Pressure/ 22 23 39 (protocol* adj weaning).mp. 24 25 40 (ventilat* adj weaning).mp. 26 27 41 ((artificial or mechanical) adj ventilat*).mp. 28 29 42 ventilat*.tw. 30 31 43 or/18-42 32 33 44 randomized controlled trial.pt. or Randomi?ed Controlled trial*.mp. or 34 35 Randomi?ed clinical trial*.mp. or Random Allocation/ or Random allocation.mp. or 36 37 random*.tw. or clinical trial/ or controlled clinical trial/ or single-blind method/ or 38 39 double-blind method/ or ((singl* or doubl* or trip* or trebl*) adj25 (blind* or 40 41 mask*)).mp. or Placebos/ or placebo*.tw. or drug therapy.fs. or trial.ab. or groups.ab. 42 43 45 1 or 2 44 45 46 17 and 43 and 44 and 45 46 47 47 limit 46 to (ed=20170101-20190304 or ez=20170101-20190304) [Limit not 48 49 valid in Embase; records were retained] 50 48 limit 47 to dd=20170101-20190304 [Limit not valid in Ovid 51 52 MEDLINE(R),Ovid MEDLINE(R) Daily Update,Ovid MEDLINE(R) In-Process,Ovid 53 54 MEDLINE(R) Publisher; records were retained] 55 56 49 remove duplicates from 48 57 58 50 placebos.mp. or Placebos/ 59 60 51 control.mp. or Control Groups/

https://mc.manuscriptcentral.com/bmj BMJ Page 42 of 109

1 2 3 4 52 sucralfate.mp. or Sucralfate/ or antacids.mp. or Antacids/ or Anti-Ulcer 5 6 Agents/ or antiulcer agents.mp. or 16,16-dimethylprostaglandin e2/ or arbaprostil/ or 7 8 burimamide/ or carbenoxolone/ or cisapride/ or enprostil/ or gefarnate/ or metiamide/ 9 10 or misoprostol/ or pirenzepine/ or proglumide/ or propantheline/ or rioprostil/ or 11 vitamin u/ or Carafate.mp. or Antepsin.mp. 12 Confidential: For Review Only 13 53 1 or 2 or 50 or 51 or 52 14 15 54 17 and 43 and 44 and 53 16 17 55 limit 54 to (ed=20170101-20190304 or ez=20170101-20190304) [Limit not 18 19 valid in Embase; records were retained] 20 21 56 limit 55 to dd=20170101-20190304 [Limit not valid in Ovid 22 23 MEDLINE(R),Ovid MEDLINE(R) Daily Update,Ovid MEDLINE(R) In-Process,Ovid 24 25 MEDLINE(R) Publisher; records were retained] 26 27 57 remove duplicates from 56 28 29 58 49 or 57 30 31 32 33 CENTRAL (The Cochrane Library Issue 1, 2019) 34 35 #1 MeSH descriptor: [Proton Pump Inhibitors] explode all trees 36 37 #2 proton near/5 pump near/5 inhibitor* 38 39 #3 ppi 40 41 #4 MeSH descriptor: [Omeprazole] this term only 42 43 #5 MeSH descriptor: [Benzimidazoles] this term only 44 45 #6 Omeprazole or losec or omez or prilosec or rapinex or zegerid or esomeprazole or 46 47 axagon or esporal or lucen or nexiam or nexium or sompraz or soleri or Lansoprazole 48 49 or agopton or bamalite or lansoprazole or lanzor or monolitum or ogast or ogastro or 50 ogastro or opiren or prevacid or prezal or pro ulco or promeco or takepron or ulpax or 51 52 zoton 53 54 #7 MeSH descriptor: [2-Pyridinylmethylsulfinylbenzimidazoles] this term only 55 56 #8 losec.mp. or omez.mp. or prilosec.mp. or rapinex.mp. or zegerid.mp. or 57 58 esomeprazole.mp. or axagon.mp. or esporal.mp. or lucen.mp. or nexiam.mp. or 59 60 nexium.mp. or sompraz.mp. or soleri.mp. or Lansoprazole.mp. or agopton.mp. or

https://mc.manuscriptcentral.com/bmj Page 43 of 109 BMJ

1 2 3 4 bamalite.mp. or lansoprazole.mp. or lanzor.mp. or monolitum.mp. or ogast.mp. or 5 6 ogastro.mp. or ogastro.mp. or opiren.mp. or prevacid.mp. or prezal.mp. or pro ulco.mp. 7 8 or promeco.mp. or takepron.mp. or ulpax.mp. or zoton.mp. or dexlansoprazole.mp. or 9 10 kapidex.mp. or dexilant.mp. or pantoprazole.mp. or pantoloc.mp. or pantecta.mp. or 11 protium.mp. or protonix.mp. or rabeprazole.mp. or AcipHex.mp. or Pariet.mp. or 12 Confidential: For Review Only 13 dexrabeprazole.mp. or rablet.mp. or Lansoprazol.mp. or Esomeprazol.mp. or 14 15 Prevacid.mp. or Dexilant.mp. or Prilosec.mp. or Zegerid.mp. or Omeprazol.mp. or 16 17 Pantroprazol.mp. or "TAK 390MR".mp. or "AG 1749".mp. or "BY 1023".mp. 14390 18 19 #9 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 20 21 #10 MeSH descriptor: [Histamine H2 Antagonists] explode all trees 22 23 #11 MeSH descriptor: [Cimetidine] explode all trees 24 25 #12 MeSH descriptor: [Guanidines] this term only 26 27 #13 MeSH descriptor: [Imidazoles] this term only 28 29 #14 MeSH descriptor: [Famotidine] this term only 30 31 #15 MeSH descriptor: [Nizatidine] explode all trees 32 33 #16 MeSH descriptor: [Metiamide] explode all trees 34 35 #17 MeSH descriptor: [Burimamide] explode all trees 36 37 #18 MeSH descriptor: [Thiazoles] explode all trees 38 39 #19 MeSH descriptor: [Furans] explode all trees 40 41 #20 MeSH descriptor: [Ranitidine] explode all trees 42 43 #21 cimetidine or acitak or altramet or biomet or dyspamet or eureceptor or galenamet 44 45 or histodil cimetidine or acitak or altramet or biomet or dyspamet or eureceptor or 46 47 galenamet or histodil or peptimax or phimetin or tagamet or ultec or zita or famotidine 48 49 or fluxid or leader acid reducer or mylanta or pepcid or roxatidine or Roxit or Nizatidine 50 or axid or tazac or zinga or ranitidine or azanplus or biotidin or pylorid or raciran or 51 52 raniberl or ranisen or ranitidin or rantec or sostril or taladine or tritec or wal-zan or 53 54 zantac or Zantic or ((H2 or histamine) near/3 (blocker* or blockader*)) or ((H2 or 55 56 histamine) near/3 (receptor* or antagonist*)) or (H2 near antihistaminic*) or H2RA* 57 58 #22 #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 59 60 #23 MeSH descriptor: [Peptic Ulcer] explode all trees

https://mc.manuscriptcentral.com/bmj BMJ Page 44 of 109

1 2 3 4 #24 ulcer* or Lesion* 5 6 #25 #23 or #24 7 8 #26 (bleed* or re-bleed* or rebleed* or h?emorrhag*) 9 10 #27 #25 and #26 11 #28 (stress near/3 ulcer*) 12 Confidential: For Review Only 13 #29 #27 or #28 14 15 #30 MeSH descriptor: [Gastrointestinal Hemorrhage] explode all trees 16 17 #31 (gastrointestinal or gastro-intestinal) near/5 (bleed* or re-bleed* or rebleed* or 18 19 h*emorrhag*) 20 21 #32 (h?ematochezia* or h?ematemese* or mel?ena*) 22 23 #33 MeSH descriptor: [Wounds and Injuries] explode all trees 24 25 #34 MeSH descriptor: [Gastric Mucosa] explode all trees 26 27 #35 (mucos* near/5 injur*) 28 29 #36 (stomach or antrum or antral or pyloric or pylorus or gastri* or gastropathy or 30 31 epigastr* or duodenal or duodenum or gastro-duodenal or gastroduodenal or oeso*ag* 32 33 or esp*ag* or "upper GI" or UGI or "upper gastrointestinal" or "upper gastrointestinal") 34 35 #37 #26 and #36 36 37 #38 MeSH descriptor: [Gastritis] explode all trees 38 39 #39 #26 and #38 40 41 #40 #29 OR #30 OR #31 OR #32 OR #33 OR #34 OR #35 OR #36 OR #37 OR #39 42 43 #41 #29 or #30 or #31 44 45 #42 MeSH descriptor: [Critical Care] explode all trees 46 47 #43 MeSH descriptor: [Critical Care] explode all trees 48 49 #44 MeSH descriptor: [Critical Illness] explode all trees 50 #45 MeSH descriptor: [Intensive Care Units] explode all trees 51 52 #46 ICU* 53 54 #47 ((critical or intensive) near/3 (care or illness)) 55 56 #48 MeSH descriptor: [Intubation, Gastrointestinal] explode all trees 57 58 #49 MeSH descriptor: [Monitoring, Physiologic] explode all trees 59 60 #50 MeSH descriptor: [Multiple Organ Failure] explode all trees

https://mc.manuscriptcentral.com/bmj Page 45 of 109 BMJ

1 2 3 4 #51 MeSH descriptor: [Acid-Base Equilibrium] explode all trees 5 6 #52 MeSH descriptor: [Multiple Trauma] explode all trees 7 8 #53 (serious* near injur*) 9 10 #54 (severe near (traum* or shock)) 11 #55 MeSH descriptor: [Perioperative Care] explode all trees 12 Confidential: For Review Only 13 #56 ((preoperative or intraoperative or perioperative) near (care or procedure* or 14 15 period)) 16 17 #57 MeSH descriptor: [Resuscitation] explode all trees 18 19 #58 MeSH descriptor: [Shock] explode all trees 20 21 #59 MeSH descriptor: [Sepsis] explode all trees 22 23 #60 MeSH descriptor: [Ventilator Weaning] explode all trees 24 25 #61 MeSH descriptor: [Ventilators, Mechanical] explode all trees 26 27 #62 Ventilators, Negative-Pressure 28 29 #63 (protocol* near weaning) 30 31 #64 (ventilat* near weaning) 32 33 #65 ((artificial or mechanical) near ventilat*) 34 35 #66 ventilat* 36 37 #67 #42 or #43 or #44 or #45 or #46 or #47 or #48 or #49 or #50 or #51 or #52 or #53 38 39 or #54 or #56 or #57 or #58 or #59 or #60 or #61 or #62 or #63 or #64 or #65 or #66 40 41 #68 #9 or #22 42 43 #69 #41 and #67 and #68 with Cochrane Library publication date Between Jan 2017 44 45 and Mar 2019 46 47 #70 placebos 48 49 #71 MeSH descriptor: [Placebos] explode all trees 50 #72 #70 or #71 51 52 #73 control 53 54 #74 MeSH descriptor: [Control Groups] explode all trees 55 56 #75 #73 or #74 57 58 #76 MeSH descriptor: [Sucralfate] explode all trees 59 60 #77 MeSH descriptor: [Antacids] explode all trees

https://mc.manuscriptcentral.com/bmj BMJ Page 46 of 109

1 2 3 4 #78 MeSH descriptor: [16,16-Dimethylprostaglandin E2] explode all trees 5 6 #79 MeSH descriptor: [Arbaprostil] explode all trees 7 8 #80 MeSH descriptor: [Burimamide] explode all trees 9 10 #81 MeSH descriptor: [Carbenoxolone] explode all trees 11 #82 MeSH descriptor: [Cisapride] explode all trees 12 Confidential: For Review Only 13 #83 MeSH descriptor: [Enprostil] explode all trees 14 15 #84 MeSH descriptor: [Gefarnate] explode all trees 16 17 #85 MeSH descriptor: [Metiamide] explode all trees 18 19 #86 MeSH descriptor: [Misoprostol] explode all trees 20 21 #87 MeSH descriptor: [Proglumide] explode all trees 22 23 #88 MeSH descriptor: [Propantheline] explode all trees 24 25 #89 MeSH descriptor: [Rioprostil] explode all trees 26 27 #90 MeSH descriptor: [Vitamin U] explode all trees 28 29 #91 MeSH descriptor: [Anti-Ulcer Agents] explode all trees 30 31 #92 sucralfate or antacids or antiulcer agents or Carafate or Antepsin 32 33 #93 #76 or #77 or #78 or #79 or #80 or #81 or #82 or #83 or #84 or #85 or #86 or #87 34 35 or #88 or #89 or #90 or #91 or #92 36 37 #94 #9 or #22 or #72 or #75 or #93 38 39 #95 #41 and #67 and #94 with Cochrane Library publication date Between Jan 2017 40 41 and Mar 2019 42 43 #96 #69 OR #95 with Cochrane Library publication date Between Jan 2017 and Mar 44 45 2019, in Trials 46 47 48 49 WEB OF SCIENCE 50 TITLE: (TITLE: (Proton Pump Inhibitors or ppi* or Omeprazole or Benzimidazoles 51 52 or Omeprazole or losec or omez or prilosec or rapinex or zegerid or esomeprazole or 53 54 axagon or esporal or lucen or nexiam or nexium or sompraz or soleri or Lansoprazole 55 56 or agopton or bamalite or lansoprazole or lanzor or monolitum or ogast or ogastro or 57 58 ogastro or opiren or prevacid or prezal or pro ulco or promeco or takepron or ulpax or 59 60 zoton or 2-Pyridinylmethylsulfinylbenzimidazoles or dexlansoprazole or kapidex or

https://mc.manuscriptcentral.com/bmj Page 47 of 109 BMJ

1 2 3 4 dexilant or pantoprazole or pantoloc or pantecta or protium or protonix or rabeprazole 5 6 or AcipHex or Pariet or dexrabeprazole or rablet or Lansoprazol or Esomeprazol or 7 8 Prevacid or Dexilant or Prilosec or Zegerid or Omeprazol or Pantroprazol or "TAK 9 10 390MR" or "AG 1749" or "BY 1023" or Histamine H2 Antagonists or H2RA* or 11 Cimetidine or Guanidines or Imidazoles or cimetidine or acitak or altramet or biomet 12 Confidential: For Review Only 13 or dyspamet or eureceptor or galenamet or histodil or peptimax or phimetin or tagamet 14 15 or ultec or zita or FAMOTIDINE or famotidine or fluxid or leader acid reducer or 16 17 mylanta or pepcid or Thiazoles or roxatidine or Roxit or NIZATIDINE or Nizatidine 18 19 or axid or tazac or zinga or RANITIDINE or ranitidine or azanplus or biotidin or pylorid 20 21 or raciran or raniberl or ranisen or ranitidin or rantec or sostril or taladine or tritec or 22 23 wal-zan or zantac or Furans or Burimamide or Metiamide or Zantic or sucralfate or 24 25 antacids or Anti-Ulcer Agents)) AND ALL FIELDS: (("critical care unit" OR "shock" 26 27 OR "critical illness" OR "intubation" OR "resuscitation" OR trauma OR ventilator OR 28 29 sepsis)) 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 48 of 109

1 2 3 4 Appendix 2: List of risk factors identified from included trials 5 6 7 8 Mechanical ventilation 9 10 Neurosurgery 11 Head injury 12 Confidential: For Review Only 13 Respiratory failure 14 15 Sepsis 16 17 Burns 18 19 Hypotension 20 21 Postoperative complications 22 23 Acute renal failure 24 25 Multiple trauma 26 27 Acid-base disorder 28 29 Coagulopathy 30 31 Multiple surgical procedures 32 33 Circulatory failure 34 35 Coma 36 37 Preoperative coma GCS <9 38 39 Inappropriate secretion of antidiuretic hormone 40 41 Major postoperative complications requiring reoperation 42 43 ≥60 years old 44 45 Pyogenic central nervous system infection 46 47 Marked jaundice 48 49 Shock 50 Stroke or head trauma with Glasgow <10 51 52 Corticotherapy 53 54 Treatment with anticoagulants 55 56 Hepatic failure 57 58 Acute pancreatitis 59 60 Major thoracic or abdominal surgery

https://mc.manuscriptcentral.com/bmj Page 49 of 109 BMJ

1 2 3 4 Renal replacement therapy 5 6 History of or ongoing coagulopathy 7 8 History of chronic liver disease 9 10 Major neurologic insult 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 50 of 109

1 2 3 4 5 Appendix 3: Study characteristics and outcome definition 6 7 Table 1 Characteristics of included studies 8 Confidential: For Review Only 9Study Population Mean Age Males APAC Enteral Comparisons Outcomes 10 (years) (%) HE II Nutrition 11 Alhazzani 2017 Adults ICU patients with MV ≥48 h 59.4 57 21 89.0% Pantoprazole 40 mg in 0.9% NaCl 50 mL (0.8 Hospital mortality 12 13Multicenter (n=91) mg/mL) IV once daily (n=49) CIB 14 Placebo 0.9% NaCl, 50 mL IV once daily (n=42) VAP 15 Overt bleeding 16 17 Clostroides difficile infection 18 Length of ICU stay 19 Length of hospital stay 20 21Apte 1992 Tracheotomized medical ICU patients 26.5 67.6 NR 100% Ranitidine 50mg IV q6h (n=16) Mortality 22India (n=34) with tetanus (MV: 26%) No prophylaxis (n=18) Pneumonia 23 Overt bleeding 24 Bashar 2013 Traumatic ICU patients with MV 40.2 71.7 15.2 NR Pantoprazole 40 mg IV once daily then 40 mg PO Hospital mortality 25 26Iran (n=120) once daily (n=60) VAP 27 Ranitidine 50 mg IV three times daily, then 150 mg Length of hospital stay 28 PO twice daily (n=60) 29 30Basso 1981 Department of Intensive Care, NR NR NR NR Cimetidine 200 mg IV or PO q6h (n=60) Overt bleeding 31Italy (n=116) Neurosurgery and Plastic Surgery patients Antacids (Maalox) PO or NG 10 ml/h (n=52) 32 with risk factor for gastrointestinal No prophylaxis (n=56) 33 34 bleeding (neurosurgery, head injury, 35 respiratory failure, sepsis, burns, 36 hypotension, postoperative 37 complications, acute renal failure, 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 51 of 109 BMJ

1 2 3 4 5 multiple trauma) (respiratory 6 failure :43%) 7 Ben-Menachem Medical ICU patients (MV: 71%) 59.6 51 17.1 66% Cimetidine 300 mg loading dose IV followed by Hospital mortality 8 Confidential: For Review Only 91994 continuous infusion (n=100) CIB 10USA (n=300) Sucralfate 1 g PO or NG q6h (n=100) Pneumonia 11 No prohylaxis (n=100) Overt bleeding 12 13 Mental statue changed (mild 14 confusion, somnolence) 15 Thrombocytopenia 16 17Bhanot 2010 ICU patients with MV NR NR NR NR Omeprazole 40 mg NG daily (n=50) Pneumonia 18India (n=150) Ranitidine 50 mg IV q6h (n=50) 19Abstract Sucralfate 1 g NG q6h (n=50) 20 21Brophy 2010 Neurosurgical ICU patients with MV or 40.3 72.5 NR 39% Lansoprazole suspension 30 mg NG daily (n=28) Overt bleeding 22USA (n=51) one of the following risk factor for Famotidine 20 mg IV q12h (n=23) Thrombocytopenia 23 gastrointestinal bleeding (head injury 24 with altered mental status, acid-base 25 26 disorder, multiple trauma, coagulopathy, 27 multiple surgical procedures, hypotension 28 >1 h, or sepsis) (MV: 100%) 29 30Burgess 1995 Surgical ICU adults with severe head 36.5 73.5 NR 0 Ranitidine infusion 6.25 mg/h continuous IV (n=16) Mortality 31USA (n=34) injury and a GCS score ≤10 (MV: 79%) Placebo (n=18) Overt bleeding 32Cannon 1987 Mixed ICU patients with acute respiratory 61.2 NR NR 37% Cimetidine 300 mg IV q6h (n=21) Mortality 33 34USA (n=40) failure with MV Sucralfate 1 g NG q6h (n=19) Overt bleeding 35 Antacids 30 mL NG (n=19) 36Cartier 1980 Critically ill patients with risk factor for NR NR NR NR Cimetidine 1.2 g IV once daily (n=58) Overt bleeding 37 38 gastrointestinal bleeding (respiratory, Placebo (n=63) 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 52 of 109

1 2 3 4 5France (n=121) renal or circulatory failure, coma, 6 septicaemia) 7 Chan 1995 Neurosurgery patients (nontraumatic 61 53.5 NR NR Ranitidine 50 mg IV q6h, then 150 mg PO twice Mortality 8 Confidential: For Review Only 9Hong Kong, China cerebral disease) with two or more risk daily after EN be given (n=49) CIB 10(n=101) factors for gastrointestinal bleeding Placebo: normal saline (n=52) Pneumonia 11 (preoperative coma GCS <9, Adverse drug effect 12 13 inappropriate secretion of antidiuretic 14 hormone, major postoperative 15 complications requiring reoperation, ≥60 16 17 years old, pyogenic central nervous 18 system infection) 19Conrad 2005 ICU patients with MV >48 h, APACHE II 55.7 58.5 23.7 0 Omeprazole suspension 40 mg NG twice daily, then Mortality 20 21USA (n=359) >11, and one more risk factor for 40 mg once daily (n=178) CIB 22 gastrointestinal bleeding Cimetidine 300 mg IV, then infusion at 50 mg/h Pneumonia 23 (n=181) Overt bleeding 24 Cook 1998 ICU patients with MV >48 h 58.7 60.3 24.7 71.1% Ranitidine 50 mg IV q8h and sucralfate placebo 1 g Mortality 25 26Canada (n=1200) q6h (n=596) CIB 27 Sucralfate 1 g q6h and ranitidine placebo 50 mg IV Pneumonia 28 q8h (n=604) Length of ICU stay 29 30Darlong 2003 Mixed ICU patients with MV >24 h 41.5 53.8 NR 0 Ranitidine 50 mg IV q8h (n=24) CIB 31India (n=52) Sucralfate 1 g NG q8h (n=21) Overt bleeding 32 No prohylaxis (n=7) 33 34De Azevedo 2000 ICU patients with risk factor for 56.7 52 55.3 NR Omeprazole 40 mg IV q12h (n=38) Hospital mortality 35Brazil (n=108) gastrointestinal bleeding (stroke or head Ranitidine 150 mg/day IV infusion (n=38) CIB 36 trauma with Glasgow <10, coagulopathy, Sucralfate 1 g NG q6h (n=32) VAP 37 38 corticotherapy, treatment with Overt bleeding 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 53 of 109 BMJ

1 2 3 4 5 anticoagulants; hepatic failure, acute Length of ICU stay 6 renal failure, invasive MV ≥48 h, acute 7 pancreatitis, burning, sepsis) 8 Confidential: For Review Only 9Eddleston 1994 ICU patients with MV ≥4 d and one more 55.3 65.4 13.6 NR Sucralfate 2 g NG q8h (n=14) Mortality 10UK (n=26) risk factor for gastrointestinal bleeding No prophylaxis NG q8h (n=12) Pneumonia 11 (hypotension, sepsis, respiratory failure, Overt bleeding 12 13 renal dysfunction, hepatic dysfunction) 14El-Kersh 2018 Medical ICU patients with MV >48 h 60 56.9 NR 100% Pantoprazole 40 mg IV once daily and early EN Hospital mortality 15USA (n=124) (n=62) CIB 16 17 Placebo and early EN (n=62) Overt bleeding 18 Clostroides difficile infection 19 Length of ICU stay 20 21 Length of hospital stay 22Fabian 1993 Trauma ICU patients with MV 35 76 13 NR Cimetidine (n=410) CIB 23Canada (n=616) 300 mg IV q8h (n=204), or 300 mg as a loading dose, Pneumonia 24 then 40 mg/h (n=206) Overt bleeding 25 26 Sucralfate 1 g NG q6h (n=206) Length of ICU stay 27 Confusion 28 Thrombocytopenia 29 30Fink 2003 High risk ICU patients NR NR 15.2 NR Pantoprazole (n=158) Mortality 31USA (n=189) 40 mg IV once daily (n=32), 40 mg IV twice daily CIB 32Abstract (n=35), 80 mg IV once daily (n=20), 80 mg IV twice Overt bleeding 33 34 daily (n=37), or 80 mg IV three times daily (n=34) 35 Cimetidine IV 300 mg bolus, then 50 mg/h infusion 36 (n=31) 37 38Fogas 2013 Patients with MV >48 h 69.5 61 27 NR PPI (n=38) CIB 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 54 of 109

1 2 3 4 5Hungary (n=79) H2RA (n=41) VAP 6Abstract Overt bleeding 7 Friedman 1982 Patients with MV >12 h NR NR NR NR Cimetidine 300 mg IV q6h and placebo antacid, 30 Overt bleeding 8 Confidential: For Review Only 9USA (n=25) mL NG q1h (n=11) Diarrhea 10 Antacids : Mylanta Ⅱ 30 ml NG q1h and placebo Hypomagnesemia 11 cimetidine IV q6h (n=11) Hypophosphatemia 12 13 Placebo cimetidine IV q6h and placebo antacid, 30 14 ml NG q1h (n=14) 15Groll 1986 Mixed ICU patients (respiratory failure: 57.5 64.7 NR NR Cimetidine 300 mg in 20 ml normal saline q6h Mortality 16 17Canada (n=221) 47%) (n=114) Overt bleeding 18 Placebo (n=107) 19Gundogan 2017 Critically ill patients 61.8 NR 19.2 100% Pantoprazole and oral/ enteral nutrition (n=152) CIB 20 21Turkey (n=300) Oral/enteral nutrition (n=148) Overt bleeding 22Abstract 23Gursoy 2008 Trauma, mixed ICU patients with MV 58 49.3 14 0 PPI NG daily (n=60) Mortality 24 Turkey (n=75) Omeprazole 20 mg capsule (n=15); Pantoprazole 40 Length of ICU stay 25 26 mg tablet (n=15); Esomeprazole 20 mg tablet 27 (n=15); Rabeprazole 20 mg tablet (n=15) 28 Saline 100 mL NG daily (n=15) 29 30Halloran 1980 ICU patients with severe head injury 30.1 82 NR NR Cimetidine 300 mg IV q4h ,then PO when tube Mortality 31USA (n=50) feeding or a diet was started (n=26) CIB 32 Placebo IV q4h (n=24) Overt bleeding 33 34Hanisch 1998 Surgical ICU patients with MV ≥48 h 53.3 NR 19.3 NR Ranitidine 50 mg IV q8h (n=57) Mortality 35Germany (n=114) Pirenzepine 10 mg IV q8h (n=44) CIB 36 Placebo (n=57) Pneumonia 37 38Harlaftis 1997 ICU patients with MV 45.5 76.3 17.1 0 Ranitidine IV 25 mg/h for two hours, three times Mortality 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 55 of 109 BMJ

1 2 3 4 5Greece (n=80) daily (n=40) CIB 6 Sucralfate 1 g NG q6h (n=40) Pneumonia 7 Overt bleeding 8 Confidential: For Review Only 9Hummer 1986 Neurosurgical ICU patients with severe 24 77 NR NR Ranitidine continuous IV infusion 0.2 mg/kg/h CIB 10France (n=22) cerebral injuries (n=11) Overt bleeding 11 Placebo infusion at the same debit 0.021/kg/h Pneumonia 12 13 (n=11) 14Jakob 2005 ICU patients with MV >24 h with acute 61 60 21 0 Ranitidine 50 mg q8h (n=24) Mortality 15Switzerland (n=46) circulatory or respiratory failure Placebo 50 mg q8h (n=22) CIB 16 17 Overt bleeding 18Kantorova 2004 ICU patients with polytrauma or major 47 67 18.4 38.3% Omeprazole 40 mg IV once daily (n=72) Hospital Mortality 19Czech Republic intraabdominal or intrathoracic surgery, Famotidine 40 mg IV twice daily (n=71) CIB 20 21(n=287) with MV ≥48 h or coagulopathy Sucralfate suspension 1 g q6h (n=69) Nosocomial pneumonia 22 Placebo (n=75) Serious adverse drug reaction 23 Length of ICU stay 24 Karlstadt 1990 ICU patients with risk factor for 59.2 54 NR NR Cimetidine 300 mg continuous infusion 50 mg/h Mortality 25 26USA (n=87) gastrointestinal bleeding (major thoracic (n=54) CIB 27 or abdominal surgery, major multiple Placebo (n=33) Pneumonia 28 trauma, hypotension hypovolemic shock, Overt bleeding 29 30 sepsis, acute respiratory failure) (acute Nausea/vomiting 31 respiratory failure: 45%) Mental statue changed 32Kaushal 2000 Neurosurgery Unit patients with head 34.0 96 NR NR Famotidine 20 mg twice daily (n=25) ICU Mortality 33 34India (n=75) injury of < 24 hours’ duration, GCS score Sucralfate 2 g twice daily (n=25) CIB 35 <10, need for nasogastric intubation No prophylaxis (n=25) 36Khorvash 2014 ICU patients with MV 50.6 79.6 NR NR Pantoprazole (n=74) Mortality 37 38Iran (n=148) Sucralfate (n=74) Pneumonia 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 56 of 109

1 2 3 4 5 Length of hospital stay 6Kotlyanskaya 2007 Medical ICU patients with MV with at 71.2 NR 27.6 NR PPI (n=45) CIB 7 USA (n=66) least one additional risk factor for Lansoprazole suspension (n=22); Lansoprazole Pneumonia 8 Confidential: For Review Only 9Abstract gastrointestinal bleeding tablet (n=23) Overt bleeding 10 Ranitidine (n=21) Non-C-difficile diarrhea 11 Thrombocytopenia 12 13 Elevated AST/ALT 14Krag 2018 ICU patients with an acute condition with 67 63.9 NR 57.3% Pantoprazole 40 mg suspended in 10 ml of 0.9% 90-day mortality 15European, risk factor for gastrointestinal bleeding sodium chloride IV once daily until ICU discharge CIB 16 17multicenter (shock, renal replacement therapy, or death, for a maximum of 90 days (n=1645) Pneumonia 18(n=3298) MV >24 h, anticoagulant drugs, history of Placebo (suspended in 10 ml of 0.9% sodium Clostroides difficile infection 19 or ongoing coagulopathy, history of chloride) IV once daily (n=1653) Overt bleeding 20 21 chronic liver disease) (invasive MV: Myocardial ischemia 22 79%) Serious adverse reactions 23Labattut 1992 Mixed ICU patients (MV: 16%) 61.7 73.6 NR NR Sucralfate 1 g q4h enterally (n=197) Mortality 24 Spain (n=398) No prophylaxis (n=201) CIB 25 26 Overt bleeding 27 Length of ICU stay 28 Laggner 1989 Acute respiratory failure patients with 53.5 56.3 NR 0 Ranitidine 50-100 mg IV q6h (n=16) Mortality 29 30Austria (n=32) MV >48 h Sucralfate 1 g NG q6h (n=16) CIB 31 Pneumonia 32 Overt bleeding 33 34Laggner 1988 ICU patients with risk factor for 54.3 64.9 NR NR Ranitidine 6x50 to 6x100 mg daily (n=43) Mortality 35Austria (n=84) gastrointestinal bleeding (MV: 65%) Sucralfate 6x1 g NG or PO (n=41) Overt bleeding 36 Diarrhea 37 38Larson 1989 Severe head injury (GCS ≤10)/comatose NR NR NR NR Ranitidine IV 6.25 mg/h, 150 mg/d (n=13) Overt bleeding 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 57 of 109 BMJ

1 2 3 4 5USA (n=31) ICU patients with MV Placebo (n=18) 6Abstract 7 Lee 2014 Neurosurgical ICU patients with MV, 57.7 60 17.1 100% Esomeprazole 40 mg NG daily for 7 days (n=30) 30-day mortality 8 Confidential: For Review Only 9Taiwan (n=60) most patients received surgical treatment Famotidine 20 mg IV q12h for 7 days (n=30) CIB 10 for ICH VAP 11 Length of ICU stay 12 13Levy 1997 ICU patients with risk factor for 57.1 55.1 18.9 NR Omeprazole 40 mg PO or NG daily (n=32) Mortality 14USA (n=67) gastrointestinal bleeding (burns, Ranitidine infusion 50 mg, then 150 mg IV daily or CIB 15 coagulopathy, acute hepatic failure, major 50 mg IV q8h (n=35) Nosocomial pneumonia 16 17 neurologic insult, acute renal failure, Length of ICU stay 18 respiratory failure, sepsis, shock, trauma) 19 (respiratory: 73%) 20 21Lin 2016 Patients being weaned from MV 65.8 65.25 20.6 100% Lansoprazole 30 mg NG once daily for 14 days Mortality 22Taiwan (n=120) (difficulties being weaned off ventilators (n=60) CIB 23 in the medical or surgical ICU) No prophylaxis (n=60) Pneumonia 24 Overt bleeding 25 26Liu 2013 Neurosurgical ICU patients with ICH > 18 58 NR 0 Omeprazole 40 mg IV q12h (n=58) Mortality 27China (n=165) (MV >48 h: 22%) Cimetidine 300 mg IV q6h (n=54) Pneumonia 28 Placebo q12h (n=53) Overt bleeding 29 30Lou 2018 Critically ill patients with MV >48 h, with 50.4 73.0 NR 0 Esomeprazole 30 min intermittent IV infusions Mortality 31China (n=311) at least one additional risk factor (major given for maximum 14 days (n=155) CIB 32 surgical procedure or major trauma, Cimetidine 30 min bolus IV infusion, then Pneumonia 33 34 severe burns, shock, multiple organ continuous IV infusion given for maximum 14 days Overt bleeding 35 failure, sepsis, acidosis, coagulopathy, (n=156) Pyrexia 36 coma, and needing high dose Hypokalemia 37 38 corticosteroids) Nervous system disorders 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 58 of 109

1 2 3 4 5 Hypercapnia 6 Septic shock 7 Luk 1982 ICU patients NR NR NR NR Cimetidine 300 mg IV q6h and placebo antacid 30 Overt bleeding 8 Confidential: For Review Only 9USA (n=123) mL PO q3h (n=62) 10Abstract Placebo IV q6h and true antacid 30 mL PO q3h 11 (n=59) 12 13 Placebo IV q6h and placebo antacid (n=61) 14Maasoumi 2016 ICU patients after CABG NR NR NR NR Pantoprazole (n=45) Abdominal distension and 15Iran (n=90) Ranitidine (n=45) vomiting 16 17Macdougall 1977 Patients with fulminant hepatic failure in NR NR NR NR H2RA (n=26) Hospital mortality 18UK (n=50) liver failure unit Metiamide then change Cimetidine (n=10); Overt bleeding 19 Cimetidine (n=16) 20 21 Metiamide or Cimetidine 150 mg IV infused 100 22 mg/h to maintain an pH >5 measured two-hourly 23 No prophylaxis (n=24) 24 Martin 1993 ICU patients with risk factor for 59.5 67.9 16 3.82% Cimetidine 300 mg in a 50 ml loading dose and 50 Mortality 25 26USA (n=131) gastrointestinal bleeding (major surgery, mg/h thereafter (n=65) CIB 27 multiple trauma, hypotension, Placebo (n=66) Pneumonia 28 hypovolemic shock, sepsis, acute Overt bleeding 29 30 respiratory failure) (respiratory failure: Mental statue changed 31 20%) 32Metz 1993 ICU patients with severe head injury 33.9 73.7 NR 0 Ranitidine 6.25 mg/h continuous infusion (n=86) Pneumonia 33 34USA (n=167) (GCS ≤10) (MV: 86%) Placebo (n=81) Overt bleeding 35Mustafa 1995 ICU patients NR NR NR NR Ranitidine 100 mg IV q8h (n=16) Pneumonia 36Turkey (n=31) Sucralfate 2 g NG q6h (n=15) Overt bleeding 37 38Abstract 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 59 of 109 BMJ

1 2 3 4 5Nielsen 1989 Mixed ICU patients with septicemia or 57.56 36 13.5 NR Ranitidine 50 mg IV q6h (n=12) Mortality 6Denmark (n=25) intra-abdominal sepsis (ventilation: 24%) No ranitidine (n=13) 7 Nourian 2018 Critically ill patients with risk factor for 64.3 60 17.7 100% Ranitidine 50 mg IV injection q8h and EN for 7 days 28-day ICU Mortality 8 Confidential: For Review Only 9Iran (n=58) gastrointestinal bleeding with EN (n=29) Pneumonia 10 EN without ranitidine for 7 days (n=29) Overt bleeding 11 Length of ICU stay 12 13 Sepsis 14Pan 2004 Gastroenterology ICU patients with 47.5 53.3 12.3 0 Rabeprazole 20 mg PO daily for one week (n=20) Overt bleeding 15China (n=30) severe acute pancreatitis Famotidine 40 mg IV twice daily for one week Rash 16 17 (n=10) Granulopenia 18 Abnormal liver function 19Peura 1985 Medical ICU >5 d patients 56.2 71.8 NR NR Cimetidine 300 mg IV q6h (n=21) Mortality 20 21USA (n=39) Placebo IV q6h (n=18) Overt bleeding 22 Mental confusion 23 Pancytopenia 24 Phillips 1998 Critically ill patients with respiratory NR NR NR NR Omeprazole 40 mg NG, then 20 mg NG daily (n=33) CIB 25 26USA (n=58) failure/MV >48 h and one or more other Ranitidine 50 mg IV, then 150-200 mg infusion daily Pneumonia 27Abstract risk factors for gastrointestinal bleeding (n=25) Serious adverse drug reaction 28 Pickworth 1993 Surgical ICU trauma patients with MV 27 71.8 18.1 NR Ranitidine 50 mg IV q6h (n=44) Mortality 29 30Canada (n=83) Sucralfate 1 g NG q6h (n=39) CIB 31 Pneumonia 32 Overt bleeding 33 34 Length of ICU stay 35 Length of hospital stay 36Powell 1993 ICU patients after CABG 56.5 86 NR NR PPI (n=20) Mortality 37 38UK (n=41) Omeprazole 80 mg IV then 40 mg IV influsion q8h CIB 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 60 of 109

1 2 3 4 5 (n=10); Omeprazole 80 mg IV then 40 mg IV bolus Overt bleeding 6 q8h (n=10) 7 Ranitidine 50 mg IV q8h (n=11) 8 Confidential: For Review Only 9 Placebo q8h (n=10) 10Prakash 2008 ICU patients with MV ≥24 h 31.3 64 13.8 NR Ranitidine 50 mg IV q6h (n=25) Mortality 11 India (n=50) Sucralfate 1 g NG q6h (n=25) Pneumonia 12 13 Overt bleeding 14Prod’hom 1994 Mixed ICU patients with MV 48.2 67.6 17.1 26.4% Ranitidine 150 mg/d IV infusion (n=80) Mortality 15Switzerland Sucralfate 1 g NG q4h (n=83) CIB 16 17(n=163) Antacids 20 mL q2h, if the gastric pH＜ 4.0, 40 mL Pneumonia 18 NG (n=81) Overt bleeding 19 Leukopenia 20 21 Rash 22Reusser 1990 Surgical ICU patients with severe acute NR 77.5 NR NR Ranitidine 50 mg IV q8h (n=19) Hospital mortality 23Switzerland (n=40) intracranial lesion, and respiratory failure No prophylaxis (n=21) Overt bleeding 24 with MV >48 h Length of ICU stay 25 26Ruiz-Santana 1991 Mixed ICU patients with MV ≥6 d 38.3 68.8 15.7 0 Ranitidine 50 mg IV q6h and total PN (n=19) Mortality 27Spain (n=73) Sucralfate 1 g NG q4h and total PN (n=24) CIB 28 No prophylaxis and total PN (n=30) Overt bleeding 29 30Ryan 1993 Mixed ICU patients with MV 63.5 62.3 NR 45.6% Cimetidine continuous infusion 300 mg by 37.5 Mortality 31USA (n=114) mg/h (n=56) CIB 32 Sucralfate 1 g NG q6h (n=58) Pneumonia 33 34 Overt bleeding 35Selvanderan 2016 ICU patients with invasive MV >24 h and 52 65.5 27 85.0% Pantoprazole 40 mg in 10 mL of 0.9% saline IV once Mortality 36Australia (n=214) receive EN within 48 h of admission daily until the patient was no longer mechanically CIB 37 38 ventilated or for a maximum of 14 days (n=106) Pneumonia 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 61 of 109 BMJ

1 2 3 4 5 Placebo IV once daily, until the patient was no Clostroides difficile infection 6 longer mechanically ventilated or for a maximum of Overt bleeding 7 14 days (n=108) Length of ICU stay 8 Confidential: For Review Only 9 Length of hospital stay 10Simms 1991 Surgical ICU trauma patients 33.9 NR NR 62.9% Cimetidine 900-1200 mg IV q2-4h (n=32) CIB 11 USA (n=62) Sucralfate 1g NG four times daily (n=30) Pneumonia 12 13 Antacids 30- 60 ml IV q2-4h (n=27) Length of hospital stay 14Solouki 2009 ICU patients with MV >48 h 50.7 51.9 NR 98.4% Omeprazole suspension 20mg NG and placebo IV Mortality 15Iran (n=129) twice daily (n=61) CIB 16 17 Ranitidine 50 mg IV and placebo NG twice daily VAP 18 (n=68) Overt bleeding 19 Length of ICU stay 20 21Somberg 2008 ICU patients with risk factor for 40.8 71.8 15.2 NR PPI (n=167) Mortality 22USA (n=202) gastrointestinal bleeding (postoperative Pantoprazole 40 mg IV q24h (n=32); 40 mg IV q12h CIB 23 major surgery, major trauma, (n=38); 80 mg IV q24h (n=23); 80 mg IV q12h Pneumonia 24 hypovolemic shock, sepsis, acute (n=39); 80 mg IV q8h (n=35) Overt bleeding 25 26 respiratory failure, burns, coagulopathy) Cimetidine 300 mg bolus then 50 mg/h infusion Pyrexia 27 (respiratory failure: 15%) (n=35) Thrombocytopenia 28 Cholestatic jaundice 29 30 Abnormal liver function 31 Pruritis 32 Phlebitis 33 34Spapen 1995 Hemodynamically stable ICU patients 65.9 55 14 NR H2RA (n=20) Mortality 35Belgium (n=30) (MV: 55%) Cimetidine 1200 mg continuous infusion (n=10); 36 Ranitidine 200 mg continuous infusion (n=10) 37 Control (n=10) 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 62 of 109

1 2 3 4 5Thomason 1996 Trauma, surgical, or neurosurgical ICU 30.5 65.2 18 NR Ranitidine 150 mg/250 mL D5W continuous Mortality 6USA (n=160) patients with MV infusion at 10 mL/h (6.25 mg/hour). If I < 4.0, CIB 7 increase 10 mL q2h until the gastric pH ≥ 4.0 (n=80) Pneumonia 8 Confidential: For Review Only 9 Sucralfate 1 g NG q6h (n=80) Overt bleeding 10 Antacids 1350 mg of aluminum hydroxide and 1200 11 mg of magnesium hydroxide, q4h if gastric pH<4.0 12 13 q2h until pH ≥ 4.0 (n=82) 14Tryba 1985 Anesthesiological and neurosurgical ICU 56.2 36 NR NR Cimetidine 2 g q24h (n=33) Mortality 15Germany (n=67) patients with high risk for stress ulcer Sucralfate 1 g q4h (n=34) CIB 16 17 Antacids 10 ml q2h (n=33) Overt bleeding 18 Dryness of the mouth 19 Confusion 20 Nausea and vomiting 21 22 23 24 25 26Van den Berg 1985 Mixed ICU patients with MV ≥3 d 46.2 64.3 NR 60.7% Cimetidine 20 mg/kg q24h, continuous IV infusion Overt bleeding 27Netherlands (n=28) (no more than 400 mg q24h) (n=14) 28 Placebo (n=14) 29 30Wee 2013 ICU/CCU patients (MV: 77%) 72 NR 22 NR Pantoprazole 40 mg IV qam (n=68) CIB 31USA (n=129) Famotidine 20 mg IV q12h (n=61) Overt bleeding 32Abstract Clostroides difficile infection 33 34 Hypomagnesemia 35 Nausea/vomiting 36 Length of ICU stay 37 38Zinner 1981 Surgical ICU patients 57.0 61.67 NR NR Cimetidine 300 mg IV q6h (n=100) Mortality 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 63 of 109 BMJ

1 2 3 4 5USA (n=200) No prophylaxis (n=100) CIB 6 Aluminum hydroxide antacid titration NG to Overt bleeding 7 maintain pH >4.0 (n=100) Length of ICU stay 8 Confidential: For Review Only 9 Mental confusion 10 Diarrhea 11 Creatinine increase 12 13 Regurgitation 14 APACHE=Acute Physiology and Chronic Health Evaluation; ICU=intensive care unit; MV=mechanical ventilation; IV=intravenous; CIB=clinically important gastrointestinal 15 bleeding; VAP=ventilator-associated pneumonia; NR=not reported; PO=per os; NG=nasogastric; GCS=Glasgow coma scale; EN=enteral nutrition; H2RA=histamine-2 16 17 receptor antagonist; PPI=proton pump inhibitor; AST=aspartate transaminase; ALT=alanine aminotransferase; CABG=Coronary artery bypass graft; pH= power of hydrogen; 18 ICH=intracerebral hemorrhage; PN=parenteral nutrition; CCU=cardiac care unit. 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 64 of 109

1 2 3 4 5 Table 2 Outcome definition of included studies 6 Study Definition of gastrointestinal bleeding Definition of Pneumonia 7 Alhazzani 2017 Overt bleeding: Hematemesis, overt nasogastric bleeding, melena or hematochezia VAP: Any 2 of the following symptoms or signs: 8 Confidential: For Review Only 9 Multicenter (n=91) CIB: Overt bleeding plus at least one of the following: (1) fever (temperature >38°C) or hypothermia (temperature <36°C) 10 (1) a spontaneous drop of SBP or DBP of 20 mmHg or more within 24 hours of upper (2) relative neutropenia (<3.0 x 106/L) or leukocytosis (>10 x 106/L) 11 gastrointestinal bleeding (3) purulent sputum 12 13 (2) an orthostatic increase in pulse rate of 20 beats/min or more and a decrease in SBP of 14 greater than or equal to 10 mm Hg 15 (3) a decrease in hemoglobin of 2 g/dL or more in 24 hours or transfusion of two units or 16 17 more of packed RBCs within 24 hours of bleeding 18 Apte 1992 Overt bleeding: Gastric aspirate for bright red altered blood and once in morning for occult New infiltrate or bronchial breath sounds 19 India (n=34) blood Positive sputum culture 20 21 Fever > 38C 22 WBC > 13,000 23 Purulent sputum (>25 WBC/LPF) 24 Bashar 2013 NR Clinical Pulmonary Infection Score 25 26 Iran (n=120) 27 Basso 1981 Overt bleeding: Hematemesis, melena, blood in the nasogastric tube or the stool or changes NA 28 Italy (n=116) in hematocrit 29 30 31 Ben-Menachem 1994 Overt bleeding: Any of the following: CXR finding >=72 hours after ICU admission 32 USA (n=300) (1) persistent hematemesis (red blood or guaiac-positive “coffee grounds” that did not clear Fever 33 with 1.5L saline lavage Leukocytosis 34 35 (2) 3-point decrease in hematocrit during 24h accompanied by red blood or guaiac-positive Purulent tracheal aspirate 36 “coffee grounds” material that cleared with lavage, or melena, or three guaiac-positive stools Sputum gram stain >25 WBC/LPF 37 without evidence of lower gastrointestinal bleed “accepted” nosocomial pathogen culture 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 65 of 109 BMJ

1 2 3 4 5 (3) any unexplained 6-point decrease in hematocrit during a 48h period 6 CIB: Overt bleeding plus stress-related lesions found on endoscopy 7 Bhanot 2010 NR NA 8 Confidential: For Review Only 9 India (n=150) 10 Abstract 11 Brophy 2010 NR 12 Overt bleeding：Hematemesis, bloody gastric aspirate, melena, or hematochezia 13 USA (n=51) 14 Significant bleeding: At least one of the following: (but without reporting data) 15 (1) endoscopic evidence of stress related mucosal bleeding 16 17 (2) bright red blood per NG tube that did not clear after lavage 18 (3) overt bleeding plus either a decrease in blood pressure of 20 mm Hg, or a decrease of 2 19 g/dL in hemoglobin and 2 units of blood transfused within 24 h 20 21 Burgess 1995 Upper gastrointestinal bleeding: Hematemesis, hematochezia or bright red blood per NG tube NA 22 USA (n=34) Coffee ground NG tube aspirate; plus a 5% decrease from baseline in hematocrit occurring at 23 least 8 hr after study drug initiation 24 Cannon 1987 Overt bleeding: Coffee-ground heme-positive nasogastric aspirate NA 25 26 USA (n=40) 27 Cartier 1980 NA NR 28 France (n=121) 29 30 Chan 1995 CIB: Endoscopic stigmata of recent hemorrhage, requiring blood transfusion and/or surgery HAP: Radiographic evidence of pulmonary changes with or without a 31 Hong Kong, China to stop bleeding, or treatment of peritonitis as a result of perforation of acute gastroduodenal temperature rise to higher than 39°C or positive cultures from tracheal 32 (n=101) tract ulcers aspirates or sputum 33 34 Conrad 2005 Overt bleeding USFDA guidance document 35 USA (n=359) Clinically significant upper gastrointestinal bleeding: 36 (1) bright red blood not clearing after tube adjustment and 5–10 mins of lavage with saline 37 38 (2) 8 hrs of persistent Gastroccult-positive coffee-grounds material with aspirates (performed 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 66 of 109

1 2 3 4 5 every 2 hrs) not clearing with ≥100 mL of saline lavage on days 1–2 6 (3) persistent Gastroccult-positive coffee-grounds material over 2–4 hrs on days 3–14 in three 7 consecutive aspirates not clearing with ≥100 mL of saline lavage 8 Confidential: For Review Only 9 Cook 1998 Overt bleeding: Hematemesis, nasogastric aspirate containing blood or coffee-grounds CDC criteria 10 Canada (n=1200) material, melena, or hematochezia 11 CIB: Overt + one of four of: 12 13 (1) SBP or DBP drop of 20 mm Hg within 24 hours of upper gastrointestinal bleed 14 (2) increase HR of 20 bpm and decrease DBP of 10 mm Hg in upright position 15 (3) Hb drop of 2 gm/dL in 24 hours and transfusion of 2 units of blood or failure of the 16 17 hemoglobin concentration to increase by at least the number of units transfused 18 Darlong 2003 Overt bleeding: Hematemesis, melena, frank or coffee ground NG aspirate NR 19 India (n=52) CIB: None required blood transfusion 20 21 22 De Azevedo 2000 CIB CDC criteria 23 Brazil (n=108) Upper gastrointestinal hemorrhage: Hematemesis, unaltered blood or "coffee-ground" 24 through nasogastric tube or melena 25 Eddleston 1994 Overt bleeding: Blood noted in nasogastric aspirate New or progressive infiltrate on chest x-ray 26 27 UK (n=26) Unexplained reduction on partial pressure of arterial oxygen 28 Positive culture from tracheal aspirate with either fever (>38 C) or 29 increase in white cell count of >3 x 109 cells/L 30 31 El-Kersh 2018 Overt bleeding: Coffee-ground aspirate in nasogastric tube or coffee-ground emesis, bloody NA 32 USA (n=124) secretions in nasogastric tube or hematemesis, melena or hematochezia 33 Significant gastrointestinal bleeding: A 3-point decrease in hematocrit within a 24-hour 34 35 period with clinical signs of overt bleeding, or an unexplained 6-point decrease in hematocrit 36 in a 48-hour period 37 Fabian 1993 Overt bleeding: Frank red blood or coffee ground clearing with irrigation New and persistent infiltrate 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 67 of 109 BMJ

1 2 3 4 5 Canada (n=616) Clinically significant bleeding: Overt bleeding + blood transfusion Purulent sputum 6 Positive sputum culture 7 Fever >38 C 8 Confidential: For Review Only 9 WBC > 15,000 10 Fink 2003 NA NR 11 USA (n=189) 12 13 Abstract 14 Fogas 2013 NA Leukocytosis 15 Hungary (n=79) Elevated Procalcitonin 16 17 Abstract Fever 18 Purulent ETT secretion 19 Positive ETT microbiology 20 21 New/increase infiltrate 22 Friedman 1982 Occult bleeding: A drop in the hematocrit of 5 or more points, associated with positive tests NA 23 USA (n=25) for stool occult blood for 3 consecutive days without obvious non-upper gastrointestinal 24 bleeding. 25 26 Overt bleeding: Fresh or old blood in the nasogastric aspirate which failed to clear with saline 27 lavage in 15 min, or as melena. 28 Groll 1986 Bleeding did not lead to hemodynamic instability in any of the patients NA 29 30 Canada (n=221) 31 Gundogan 2017 NA NA 32 Turkey (n=300) 33 34 Abstract 35 Gursoy 2008 NR NR 36 Turkey (n=75) 37 38 Halloran 1980 Overt bleeding: Bright red blood or 4+ stool guaiac for 3 consecutive 8 hours period NA 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 68 of 109

1 2 3 4 5 USA (n=50) CIB: 2 or more units transfusions in 24 hours 6 Hanisch 1998 Clinically significant: Bright red blood in NG or melena with hemodynamic changes (systolic New infiltrate 7 Germany (n=114) blood pressure,100 mm Hg, tachycardia .100 beats per minute) and requirement of blood Purulent tracheal aspirate 8 Confidential: For Review Only 9 transfusion (HB drop >2 gm/dL in 24 hours) and endoscopic identification of bleeding site Positive sputum culture 10 and activity Temperature >38 C 11 WBC >10,000 12 13 Harlaftis 1997 Overt bleeding: Gastric aspirate for coffee ground or fresh blood New and persistent infiltrate and three of: 14 Greece (n=80) (1) Fever > 38 C 15 (2) WBC >10,000 16 17 (3) Positive sputum Gram stain 18 (4) Positive sputum culture 19 Hummer 1986 CIB: Endoscopic examination NA 20 21 France (n=22) 22 Jakob 2005 NA NR 23 Switzerland (n=46) 24 25 Kantorova 2004 Overt bleeding: Hematemesis, melena, positive nasogastric aspirate A new or progressive infiltrate on chest radiograph plus 3 of the 26 Czech Republic Clinically important hemorrhage: Overt bleeding plus one of the following: following 27 (n=287) (1) decrease in SBP > 20 mmHg or increase in pulse rate > 20 beats/min within 24 h (1) purulent tracheal aspirate with > 25 leukocutes/LPF 28 (2) decrease in Hb >2 g/dL (2) peripheral leukocytosis > 11 × 109/L or >10 % bands 29 30 (3) temperature >38.5 °C 31 (4) pathogen from aspirate, bronchiolar lavage ≥ 104 CFU/mL or 32 protected brush sampling ≥ 103 CFU/mL 33 34 (5) positive hemoculture or pleural cultures 35 Karlstadt 1990 Clinically significant upper gastrointestinal bleeding: One of the following: NA 36 USA (n=87) (1) hematemesis or >10 ml of bright red blood in a single aspirate 37 38 (2) melena or hematochezia 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 69 of 109 BMJ

1 2 3 4 5 (3) coffee ground and positive for hemoglobin in 3 consecutive NG aspirates 6 hours apart 6 and a HB drop by 1 gm/dL over 24 hours 7 (4) positive occult blood in NG aspirate not clearing with lavage 8 Confidential: For Review Only 9 Kaushal 2000 NR NA 10 India (n=75) 11 Khorvash 2014 NR CDC criteria 12 13 Iran (n=148) 14 15 Kotlyanskaya 2007 CIB: Overt bleeding associated with hemodynamic changes or Hb drop NR 16 USA (n=66) 17 18 Abstract 19 Krag 2018 Overt bleeding: Hematemesis, coffee ground emesis, melena, hematochezia or bloody Modified CDC criteria 20 European, multicenter nasogastric aspirate 21 22 (n=3298) CIB: Overt bleeding and at least one of the following: 23 (1) A spontaneous decrease in SBP, mean arterial pressure, or DBP of 20 mmHg or more 24 (2) initiation of treatment with a vasopressor or a 20% increase in vasopressor dose 25 (3) a decrease in Hb of at least 2 g per deciliter 26 27 (4) transfusion of two or more units of packed red cells 28 Labattut 1992 Overt bleeding: Macroscopically visible blood or coffee ground from gastric tube or melena NR 29 Spain (n=398) Further classification of severity: 30 31 Mild: HR <100/min, SBP > 100 + slight cutaneous vasoconstriction 32 Moderate: HR 100-120, SBP 90-100 + oliguria, skin pallor & seating 33 Severe: HR >120, SBP < 80 +extreme pallor, anuria, stupor 34 35 Laggner 1989 Overt bleeding: Macroscopically visible blood in gastric aspirates and/or hematemesis New infiltrate 36 Austria (n=32) CIB: With packed red blood cell Bronchial colonization 37 WBC > 15,000 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 70 of 109

1 2 3 4 5 Fever >38.5 C 6 Laggner 1988 Overt bleeding: Macroscopically visible blood in gastric aspirated or hematemesis NA 7 Austria (n=84) 8 Confidential: For Review Only 9 10 Larson 1989 Upper gastrointestinal bleeding: Blood or coffee ground drainage per NG tube accompanied NR 11 USA (n=31) by a ≥5% decrease in hematocrit 12 Abstract 13 14 Lee 2014 Upper gastrointestinal bleeding: Tarry stool, hematemesis, drainage of more than 60 mL > 48 hrs of ventilation and 3 or more of: 15 Taiwan (n=60) coffee ground substance from nasogastric tube, or decreased hemoglobin level more than 2 (1)Persistent (>48 hours) or new infiltrate 16 g/dL with proved lesions by endoscopic examination (2)Positive sputum smear 17 18 (3)Fever >38.3°C 19 (4)WBC >12×109/L 20 21 22 Levy 1997 Clinically important bleeding: Hemodynamic instability, aspiration of coffee ground material, Not defined 23 USA (n=67) or a decrease in Hb>2 g/dl, or melena 24 Lin 2016 Apparent gastrointestinal bleeding: Clinical Pulmonary Infection Score >6: 25 26 Taiwan (n=120) (1) a coffee ground substance from the NG aspirate 60 mL (1)body temperature 27 (2) fresh blood from the NG tube (2)white blood cell count 28 (3) passage of tarry stool (3)purulent secretions, 29 Clinically significant gastrointestinal bleeding: gastrointestinal bleeding with hemoglobin (4)diffuse or localized pulmonary infiltrate in Chest x-ray 30 31 level decrease 2 gm/dL or in need of a blood transfusion of >2 units (5)progression of infiltrate, 32 (6)culture of secretions 33 Liu 2013 Upper gastrointestinal bleeding: hematemesis, aspiration of coffee ground material from the NA 34 35 China (n=165) nasogastric tube, or melena, with or without hemodynamic instability resulting from gross 36 bleeding that needed transfusion 37 Lou 2018 Any overt bleeding: gastric tube, hematemesis, melena, or hematochezia NA 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 71 of 109 BMJ

1 2 3 4 5 China (n=311) Significant upper gastrointestinal bleeding: 6 (1) bright red blood in the gastric tube that did not clear after 5–10 minutes of at least 100 mL 7 lavage 8 Confidential: For Review Only 9 (2) persistent Gastroccult-positive ‘coffee grounds’ material for 8 hours on days 1–2 of the 10 given study drug treatment that did not clear with at least 100 mL of lavage 11 (3) gastroccult-positive ‘coffee grounds’ material persisting in at least three consecutive 12 13 gastric aspirates collected within 2–4 hours of each other on days 3–14 that did not clear with 14 at least 100 mL lavage 15 Luk 1982 Gastrointestinal bleeding: Both a decrease in hematocrit of 5% or requirement for transfusion NR 16 17 USA (n=123) and hemoccult positive stool or gastric aspirate 18 Abstract 19 Maasoumi 2016 NA NA 20 21 Iran (n=90) 22 Macdougall 1977 Bleeding: Aspiration of fresh blood via the nasogastric tube and when this occurred these NA 23 UK (n=50) patients were endoscoped to identify the site of bleeding 24 25 Martin 1993 Bleeding: one of the following: New or persistent infiltrate 26 27 USA (n=131) (1) hematemesis or bright red blood not clearing with flushing >25 WBC/LPF 28 (2) coffee ground persistent for 8 hours and not clearing with 10 ml lavage and Hct drop by Positive sputum culture 29 5% Numerous bacteria on oil immersion 30 31 CIB: bleeding plus more than 2 units of blood transfusion 32 Metz 1993 Bleeding: Gastroccult positive NG aspirate, bright red blood in NG aspirate, hematemesis, NA 33 USA (n=167) hemoccult positive stool, melena, hematochezia 34 35 Mustafa 1995 NA NR 36 Turkey (n=31) 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 72 of 109

1 2 3 4 5 Abstract 6 7 Nielsen 1989 NR NR 8 Denmark (n=25) Confidential: For Review Only 9 10 Nourian 2018 Overt bleeding: coffee ground gastric secretions, hematemesis or melena New or progressive lung infiltration on chest imaging and at least two of 11 Iran (n=58) the following; purulent secretions, temperature above 38.5℃, positive 12 tracheal or plural fluid culture, leukocytosis or leukopenia 13 14 Pan 2004 NA NR 15 China (n=30) 16 Peura 1985 Overt bleeding by endoscopic findings NA 17 18 USA (n=39) 19 Phillips 1998 NA NA 20 USA (n=58) 21 22 Abstract 23 Pickworth 1993 Overt bleeding: Gross blood not clearing with lavage, hematemesis, melena New infiltrate and three of: 24 Canada (n=83) (1)Fever >38.5 C 25 26 (2)WBC > 10,000 27 (3)Positive sputum culture 28 (4)Sputum with >25 WBC/LPF and numerous bacteria 29 30 31 32 Powell 1993 Bloody nasogastric tube aspirate NR 33 UK (n=41) 34 35 Prakash 2008 Overt bleeding: hematemesis, gross blood NG not clearing with lavage, melena, Infiltrate and three of: 36 India (n=50) hematochezia (1)WBC >10,000 37 (2)Pathogenic bacteria on a tracheal or blood culture 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 73 of 109 BMJ

1 2 3 4 5 (3)Gram’s stain of tracheal aspirate showing moderate to heavy 6 bacteria or polymorphs-neutrophils >25/HPF 7 (4)Fever >38 C 8 Confidential: For Review Only 9 Prod’hom 1994 Overt bleeding: Fresh blood NG or coffee-ground material New infiltrate and either: 10 Switzerland (n=163) Clinically significant: Overt + any of: (1) pleural fluid or blood culture positive for an organism also isolated 11 (1) Physical signs in the tracheal aspirate, radiographic cavitation, or histopathologic 12 13 (2) Transfusion evidence of pneumonia 14 (2) ETT aspirate > 25 WBC /LPF, new WBC >10,000 and increase of 15 25% or more, fever >= 38.5 C 16 17 Reusser 1990 Overt bleeding: NR 18 Switzerland (n=40) (1) bright red blood NG, melena 19 (2) or Hb drop >2g/dL within 24 h with a positive guaiac test or with gastric drainage > 100 20 21 ml coffee ground material 22 Ruiz-Santana 1991 Overt bleeding: Hematemesis, bloody NG aspirate, melena, coffee-ground material NA 23 Spain (n=73) CIB: Overt bleeding plus hemodynamic instability 24 25 Ryan 1993 Overt bleeding: Fresh blood or coffee grounds in the nasogastric aspirate, which tested CDC criteria 26 27 USA (n=114) positive with the Hemoccult test and persisted after 100 mL of saline lavage. 28 CIB: Bleeding required transfusion or treated surgically. 29 Selvanderan 2016 Overt bleeding: Hematemesis, bloody gastric aspirate, melena, or hematochezia. CDC criteria 30 31 Australia (n=214) CIB: Overt bleeding with one of: 32 (1) a reduction in mean arterial blood pressure of ≥20 mm Hg within 24 hours in the 33 absence of another cause 34 35 (2) reduction in hemoglobin of ≥20 g/L within 24 hours 36 (3) need for endoscopy or surgery to achieve hemostasis 37 Simms 1991 Bleeding: Upper gastrointestinal bleeding requiring transfusion Fever 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 74 of 109

1 2 3 4 5 USA (n=62) Leukocytosis 6 New infiltrate 7 Presence of sputum pathogens 8 Confidential: For Review Only 9 Solouki 2009 Overt bleeding: Hematemesis, coffee ground in NG tube, melena or hematochezia New infiltration in CXR along with 2 of the following criteria: fever≥ 10 Iran (n=129) CIB: Overt bleeding plus one of the following: 38.3 °C, leukocytosis ≥ 10000, and pus in tracheal tube suction 11 (1) 20 mmHg decrease in BP within 24 h 12 13 (2) 20 bpm increase in heart rate or 10 mmHg in SBP 14 (3) 2 g/dl decrease of Hb or 6% HCT within 24 h 15 (4) lack of increase in Hb after infusing 2 units of packed cell 16 17 Somberg 2008 Clinically significant upper gastrointestinal bleeding: CXR, fever, an elevated white blood cell count (>15%), and one of the 18 USA (n=202) (1) hematemesis or bright red blood in gastric aspirate that did not clear after adjustment of following: 19 nasogastric or orogastric tube and a 5- to 10-min lavage with iced water or saline (1) Cough 20 21 (2) persistent coffee ground material for 8 consecutive hours that did not clear with a 100 mL (2) new onset of purulent sputum production 22 lavage, or was accompanied by a 5% decrease in hematocrit (3) auscultatory findings on pulmonary examination or pulmonary 23 (3) a decrease in hematocrit requiring one or more transfusions that occurred in the absence consolidation 24 of any obvious source and required further diagnostic studies (4) respiratory rate≥20 breaths/min 25 26 (4) melena or frank bloody stools from an upper gastrointestinal source (5) PaO2<60mmHg,SO2<90%,orrespiratory failure requiring MV 27 (6) Tachycardia 28 (7) pleuritic chest pain 29 30 (8) new or worsened confusion 31 Spapen 1995 NR NR 32 Belgium (n=30) 33 34 Thomason 1996 Overt bleeding: Hematemesis, melena, hematochezia, or red blood per NG tube that did not CDC criteria 35 USA (n=160) clear with 500 mL of saline lavage 36 CIB: Overt bleeding plus hemodynamic instability or the need for surgical intervention 37 38 Tryba 1985 Macroscopically visible bleeding (bloody aspirate, melena, or hematemesis) NR 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 75 of 109 BMJ

1 2 3 4 5 Germany (n=67) One patient died due to bleeding 6 7 Van den Berg 1985 NA NA 8 Netherlands (n=28) Confidential: For Review Only 9 10 Wee 2013 NA NA 11 USA (n=129) 12 Abstract 13 14 Zinner 1981 Upper gastrointestinal tract bleeding: guaiac 4+ positive nasogastric aspirate continues for NR 15 USA (n=200) greater than 16 h -two consecutive nursing shifts -even after nasogastric lavage, guaiac 16 positive stools and a documented fall in hematocrit value, bright red blood per nasogastric 17 18 tube or by emesis 19 CIB: Bleeding required transfusions 20 CIB=clinically important gastrointestinal bleeding; SBP=systolic blood pressure; DBP=diastolic blood pressure; RBC=red blood cell; VAP=ventilator-associated pneumonia; 21 22 WBC=white blood cell; NR=not reported; NA= not applied; CXR=chest X-ray; ICU=intensive care unit; NG=nasogastric; HAP=hospital-associated pneumonia; USFDA= 23 U.S. Food and Drug Administration; HR=heart rate; Hb= hemoglobin; CDC= Center for Disease Control and Prevention; ETT=endotracheal tube; CFU=colony forming unit; 24 BP=blood pressure. 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 76 of 109

1 2 3 Appendix 5: Network plots for each outcome 4 5 6 7 Figure 1 Network plot for mortality 8 9 10 11 PPI 12 Confidential: For Review Only 13 14 15 16 17 18 Placebo 19 20 H2RA 21 22 23 24 25 26 27 28 29 30 Sucralfate 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 78 of 109

1 2 3 Figure 2 Network plot for clinically important gastrointestinal bleeding for all 4 5 studies 6 7 8 9 10 PPI 11 12 Confidential: For Review Only 13 14 15 16

17 Placebo 18 19 H2RA 20 21 22 23 24 25 26 27 28 Sucralfate 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 79 of 109 BMJ

1 2 3 Figure 3 Network plot for clinically important gastrointestinal bleeding for low 4 5 risk of bias studies 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 80 of 109

1 2 3 Figure 4 Network plot for pneumonia 4 5 6 7

8 PPI 9 10 11 12 Confidential: For Review Only 13 14 15 Placebo 16

17 H2RA 18 19 20 21 22 23 24 25 26 Sucralfate 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 81 of 109 BMJ

1 2 3 Figure 5 Network plot for Clostroides difficile infection 4 5 6 7 8 PPI 9 10 11 12 Confidential: For Review Only 13 14 15 16 H2RA 17 18 19 20 21 22 23

24 Placebo 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 82 of 109

1 2 3 Figure 6 Network plot for overt gastrointestinal bleeding for all studies 4 5 6 7 8 PPI 9 10 11 12 Confidential: For Review Only 13 14 15 Placebo 16 17 H2RA 18 19 20 21 22 23 24 25 26 Sucralfate 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 83 of 109 BMJ

1 2 3 Figure 7 Network plot for overt gastrointestinal bleeding for low risk of bias 4 5 studies 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 84 of 109

1 2 3 Figure 8 Network plot for length of ICU stay 4 5 6 7 8 PPI 9 10 11 12 Confidential: For Review Only 13 14

15 Placebo 16 17 H2RA 18 19 20 21 22 23 24 25 26 27 Sucralfate 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 85 of 109 BMJ

1 2 3 Figure 9 Network plot for length of hospital stay 4 5 6 7 8 PPI 9 10 11 12 Confidential: For Review Only 13 14 15 16 Placebo 17 H2RA 18 19 20 21 22 23 24 25 26 Sucralfate 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 86 of 109

1 2 3 Figure 10 Network plot for duration of mechanical ventilation 4 5 6 7 8 PPI 9 10 11 12 Confidential: For Review Only 13 14 15 Placebo 16 17 H2RA 18 19 20 21 22 23 24 25 26 Sucralfate 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 87 of 109 BMJ

1 2 3 4 5 Appendix 6: Network meta-analysis results for clinically important gastrointestinal bleeding and overt gastrointestinal bleeding using different models 6 7 8 9 Table 1 Clinically important gastrointestinal bleeding with low risk of bias (random-effects model) 10 11 OR RR 12 Conventional Conventional Direct Indirect Network Incoherence Direct Indirect Network Incoherence 13 meta-analysis meta-analysis 14 estimate estimate estimate (P-value) estimate estimate estimate (P-value) 15 (random) (random) 16 0.67 0.17 0.48 0.62 0.59 0.26 0.48 0.63 PPIs versus placebo 0.056 0.19 17 (0.30, 2.01)* (0.05, 0.60) (0.21, 0.99) (0.42, 0.90) (0.25, 1.29) (0.09, 0.74) (0.24, 0.85) (0.44, 0.91) 18 0.34 1.22Confidential:0.53 For0.43 Review0.58 1.12Only 0.73 0.49 19 H2RAs versus placebo 0.11 0.38 20 (0.16, 0.77) (0.35, 6.39) (0.27, 1.08)† (0.18, 1.01) (0.33, 1.25) (0.34, 4.10) (0.45, 1.40) (0.24, 0.99) 21 5.02 1.16 1.36 3.36 4.86 1.46 1.49 3.26 22 Sucralfate versus placebo 0.41 0.86 23 (0.12, 203.00)* (0.18, 7.47) (0.4, 5.09) (0.34, 33.13) (0.09, 251.60)* (0.37, 7.14) (0.54, 4.64) (0.35, 30.61) 24 0.54 1.84 0.90 0.58 0.45 1.04 0.65 0.59 PPIs versus H2RAs 0.08 0.16 25 (0.19, 1.32) (0.58, 7.43) (0.39, 1.83) (0.29, 1.16) (0.17, 1.00) (0.35, 2.49) (0.28, 1.14)† (0.30, 1.16) 26 0.21 0.38 0.35 0.31 0.21 0.30 0.32 0.32 27 PPIs versus sucralfate 0.76 0.59 28 (0.005, 8,14)* (0.04, 2.60) (0.09, 1.19) (0.03, 3.05) (0.004, 11.07)* (0.05, 1.25) (0.09, 0.90)‡ (0.03, 3.00) 29 0.44 0.39 0.46 0.54 0.49 0.47 30 H2RAs versus sucralfate NA NA NA NA 31 (0.14, 1.44)* (0.12, 1.19) (0.23, 0.91) (0.20, 1.61)* (0.19, 1.33) (0.24, 0.92) 32 NA=not applicable. 33 *The direct estimate from the network is wildly wider than the direct estimate from the conventional direct comparison. 34 †The indirect widens the credible interval from the network in comparison to the direct more than it should. 35 36 ‡The point estimate is not between the direct and indirect and adding the indirect to the direct narrows the credible interval more than seems intuitively reasonable given the very wide credible interval 37 around the indirect. 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 88 of 109

1 2 3 4 5 Table 2 Clinically important gastrointestinal bleeding with low risk of bias (fixed-effect model) 6 7 OR RR 8 Conventional Conventional 9 Direct Indirect Network Incoherence Direct Indirect Network Incoherence meta-analysis meta-analysis 10 estimate estimate estimate (P-value) estimate estimate estimate (P-value) 11 (fixed) (fixed) 12 0.61 0.18 0.52 0.62 0.59 0.26 0.52 0.63 PPIs versus placebo 0.02† 0.07 13 (0.42, 0.89) (0.07, 0.46) (0.36, 0.74) (0.42, 0.90) (0.41, 0.85) (0.11, 0.58) (0.37, 0.72) (0.44, 0.91) 14 0.33 1.06 0.53 0.41 0.55 1.08 0.69 0.48 15 H2RAs versus placebo 0.03† 0.14 16 (0.18, 0.61) (0.48, 2.40) (0.34, 0.84) (0.22, 0.75) (0.35, 0.84) (0.53, 2.32) (0.48, 0.98) (0.29, 0.79) 17 5.01 1.20 1.33 3.36 4.89 1.40 1.43 3.26 18 Sucralfate versus placebo Confidential:0.32 For Review Only 0.82 19 (0.13, 196.00)* (0.50, 3.02) (0.60, 3.00) (0.34, 33.13) (0.09, 252.30)* (0.64, 3.26) (0.71, 2.96) (0.35, 30.61) 20 0.56 1.72 0.98 0.58 0.48 1.09 0.75 0.59 PPIs versus H2RAs 0.02† 0.055 21 (0.26, 1.13) (0.86, 3.49) (0.60, 1.59) (0.29, 1.16) (0.24, 0.93) (0.62, 1.93) (0.49, 1.13) (0.30, 1.16) 22 0.21 0.42 0.39 0.31 0.21 0.36 0.36 0.32 23 PPIs versus sucralfate 0.68 0.63 24 (0.005, 8.18)* (0.16, 1.01) (0.17, 0.87) (0.03, 3.05) (0.004, 10.76)* (0.15, 0.81) (0.17, 0.75) (0.03, 3.00) 25 0.44 0.40 0.46 0.52 0.48 0.47 26 H2RAs versus sucralfate NA NA NA NA 27 (0.21, 0.88) (0.19, 0.79) (0.23, 0.91) (0.26, 1.00) (0.25, 0.90) (0.24, 0.92) 28 NA=not applicable. 29 *The direct estimate from the network is wildly wider than the direct estimate from the conventional direct comparison. 30 31 †incoherence between direct and indirect estimate. 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 89 of 109 BMJ

1 2 3 4 5 Table 3 Overt gastrointestinal bleeding with low risk of bias (random-effects model) 6 7 OR RR 8 Conventional Conventional 9 Direct Indirect Network Incoherence Direct Indirect Network Incoherence meta-analysis meta-analysis 10 estimate estimate estimate (P-value) estimate estimate estimate (P-value) 11 (random) (random) 12 0.66 0.14 0.35 0.59 0.59 0.28 0.42 0.61 PPIs versus placebo 0.04 0.13 13 (0.23, 2.04)* (0.04, 0.46) (0.14, 0.86) (0.45, 0.77) (0.28, 1.24) (0.12, 0.60) (0.23, 0.75) (0.48, 0.78) 14 0.29 1.29 0.37 0.34 0.51 1.02 0.60 0.41 15 H2RAs versus placebo 0.07 0.21 16 (0.15, 0.52) (0.29, 7.11) (0.19, 0.69) (0.19, 0.61) (0.35, 0.80) (0.36, 3.12) (0.41, 0.91) (0.26, 0.65) 17 5.03 0.82 1.07 3.36 4.83 1.22 1.26 3.26 18 Sucralfate versus placebo Confidential:0.38 For Review Only 0.78 19 (0.13, 197.10)* (0.07, 8.62) (0.21, 5.57) (0.34, 33.13) (0.10, 242.90)* (0.29, 5.54) (0.41, 3.96) (0.35, 30.61) 20 0.50 2.24 0.93 0.54 0.54 1.18 0.71 0.62 PPIs versus H2RAs 0.06 0.11 21 (0.16, 1.36) (0.64, 8.97) (0.38, 2.36) (0.30, 0.99) (0.24, 1.00) (0.49, 2.55) (0.38, 1.21)† (0.38, 1.00) 22 0.21 0.39 0.32 0.31 0.21 0.33 0.34 0.32 23 PPIs versus sucralfate 0.79 0.68 24 (0.005, 8.15)* (0.03, 5.35) (0.06, 1.86) (0.03, 3.05) (0.004, 11.20)* (0.06, 1.59) (0.10, 1.07)‡ (0.03, 3.00) 25 0.45 0.35 0.46 0.55 0.48 0.47 26 H2RAs versus sucralfate NA NA NA NA 27 (0.10, 2.18)* (0.07, 1.68) (0.23, 0.91) (0.20, 1.69)* (0.16, 1.40) (0.24, 0.92) 28 NA=not applicable. 29 *The direct estimate from the network is wildly wider than the direct estimate from the conventional direct comparison. 30 31 †The indirect widens the credible interval from the network in comparison to the direct more than it should. 32 ‡The point estimate is not between the direct and indirect and adding the indirect to the direct narrows the credible interval more than seems intuitively reasonable given the very wide credible interval 33 around the indirect. 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 90 of 109

1 2 3 4 5 Table 4 Overt gastrointestinal bleeding with low risk of bias (fixed-effect model) 6 7 OR RR 8 Conventional Conventional 9 Direct Indirect Network Incoherence Direct Indirect Network Incoherence meta-analysis meta-analysis 10 estimate estimate estimate (P-value) estimate estimate estimate (P-value) 11 (fixed) (fixed) 12 0.59 0.15 0.45 0.59 0.59 0.28 0.48 0.61 PPIs versus placebo 0† 0.002† 13 (0.45, 0.76) (0.09, 0.26) (0.35, 0.57) (0.45, 0.77) (0.46, 0.76) (0.19, 0.41) (0.39, 0.59) (0.48, 0.78) 14 0.29 1.07 0.48 0.33 0.48 0.98 0.60 0.40 15 H2RAs versus placebo 0† 0.003† 16 (0.20, 0.41) (0.68, 1.70) (0.37, 0.63) (0.22, 0.48) (0.36, 0.62) (0.67, 1.45) (0.49, 0.74) (0.30, 0.55) 17 5.04 1.11 1.21 3.36 4.83 1.25 1.28 3.26 18 Sucralfate versus placebo Confidential:0.28 For Review Only 0.74 19 (0.13, 198.50)* (0.51, 2.59) (0.59, 2.56) (0.34, 33.13) (0.10, 240.40)* (0.60, 2.74) (0.66, 2.53) (0.35, 30.61) 20 0.54 1.98 0.93 0.56 0.59 1.25 0.80 0.63 PPIs versus H2RAs 6.67e-06† 0.001† 21 (0.37, 0.79) (1.27, 3.11) (0.70, 1.23) (0.38, 0.82) (0.43, 0.79) (0.87, 1.79) (0.64, 0.99) (0.46, 0.85) 22 0.21 0.39 0.37 0.31 0.22 0.38 0.38 0.32 23 PPIs versus sucralfate 0.71 0.66 24 (0.005, 8.08)* (0.17, 0.86) (0.17, 0.76) (0.03, 3.05) (0.004, 10.68)* (0.17, 0.79) (0.19, 0.73) (0.03, 3.00) 25 0.44 0.40 0.46 0.52 0.47 0.47 26 H2RAs versus sucralfate NA NA NA NA 27 (0.21, 0.87) (0.19, 0.78) (0.23, 0.91) (0.26, 0.99) (0.24, 0.89) (0.24, 0.92) 28 NA=not applicable. 29 *The direct estimate from the network is wildly wider than the direct estimate from the conventional direct comparison. 30 31 †incoherence between direct and indirect estimate. 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 91 of 109 BMJ

1 2 3 4 Appendix 7: Cumulative ranking of interventions for different outcomes 5 6 Outcome Rank Intervention SUCRA 7 8 Mortality 1 Sucralfate 83.7% 9 2 H2RAs 56.6% 10 3 Placebo 44.6% 11 12 Confidential:4 For ReviewPPIs Only15.2% 13 Pneumonia 1 Sucralfate 90.8% 14 2 Placebo 71.2% 15 16 3 H2RAs 27.7% 17 4 PPIs 10.3% 18 Clostroides difficile infection 1 PPIs 60.5% 19 2 H2RAs 48.5% 20 21 3 Placebo 41.0% 22 Length of ICU stay 1 Sucralfate 72.6% 23 2 H2RAs 60.8% 24 25 3 PPIs 47.9% 26 4 Placebo 18.7% 27 Length of hospital stay 1 Sucralfate 69.2% 28 29 2 H2RAs 64.8% 30 3 Placebo 49.3% 31 4 PPIs 16.7% 32 33 Duration of mechanical ventilation 1 H2RAs 86.0% 34 2 Sucralfate 55.6% 35 3 PPIs 42.7% 36 37 4 Placebo 15.7% 38 SUCRA=surface under the cumulative ranking curve; H2RAs=histamine-2 receptor antagonists; 39 PPIs=proton pump inhibitors; ICU=intensive care unit. 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 92 of 109

1 2 3 4 5 Appendix 8: GRADE summary of findings for other outcomes 6 7Table 1 GRADE summary of findings for mortality for different comparisons 8 Confidential: For Review Only 9Comparison Direct estimate (95% Indirect estimate (95% Network estimate (95% Absolute effect estimates† (95% CrI) (per 1000) 10 CrI); Certainty of CrI); Certainty of CrI); Certainty of 11 evidence evidence evidence* 12 13PPIs versus placebo 1.05 (0.86 to 1.30); 1.12 (0.76 to 1.69); 1.06 (0.90 to 1.27); Placebo: 304 PPIs: 317 (281 to 359) 14 High High Moderate‡ Difference: 13 (-23 to 55) 15H2RAs versus placebo 0.95 (0.73 to 1.24); 0.99 (0.71 to 1.39); 0.96 (0.79 to 1.17); Placebo: 304 H2RAs: 296 (258 to 340) 16 17 High High Moderate‡ Difference: -8 (-46 to 36) 18Sucralfate versus placebo 0.99 (0.63 to 1.56); 0.84 (0.61 to 1.14); 0.90 (0.71 to 1.14); Placebo: 304 Sucralfate: 281 (235 to 332) 19 High High Very low§ Difference: -23 (-69 to 28) 20 21PPIs versus H2RAs 1.16 (0.85 to 1.63); 1.07 (0.78 to 1.49); 1.10 (0.90 to 1.36); H2RAs: 296 (258 to 340) PPIs: 317 (281 to 359) 22 High High Very low§ Difference: 22 (-27 to 66) 23PPIs versus sucralfate 1.06 (0.53 to 2.16); 1.24 (0.94 to 1.68); 1.19 (0.92 to 1.53); Sucralfate: 281 (235 to 332) PPIs: 317 (281 to 359) 24 25 High High Very low§ Difference: 36 (-18 to 90) 26H2RAs versus sucralfate 1.10 (0.88 to 1.39); 0.91 (0.52 to 1.67); 1.07 (0.89 to 1.30); Sucralfate: 281 (235 to 332) H2RAs: 296 (258 to 340) 27 High High Moderate‡ 28 Difference: 15 (-25 to 54) 29CrI=credible interval; PPIs=proton pump inhibitors; H2RAs=histamine-2 receptor antagonists. 30*Higher of direct or indirect confidence (without consider imprecision), and then considered imprecision and incoherence. 31†We used as baseline risk in the placebo group of the SUP-ICU trial. 32 33‡Rated down for imprecision. 34§Rated down 3 levels for imprecision. 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 93 of 109 BMJ

1 2 3 4 5Table 2 GRADE summary of findings for Clostroides difficile infection for different comparisons 6Comparison Direct estimate (95% Indirect estimate (95% Network estimate (95% Absolute effect estimates† (95% CrI) (per 1000) 7 CrI); Certainty of CrI); Certainty of CrI); Certainty of 8 Confidential: For Review Only 9 evidence evidence evidence* 10PPIs versus placebo 0.82 (0.31 to 2.47); NA 0.83 (0.31 to 2.48); Placebo: 15 PPIs: 12 (4 to 48) 11 High Moderate§ Difference: -3 (-11 to 33) 12 13H2RAs versus placebo NA 0.95 (0.06 to 15.48); 0.95 (0.06 to 15.48); Placebo: 15 H2RAs: 14 (1 to 257) 14 Moderate‡ Very low‡¶ Difference: -1 (-14 to 242) 15 16PPIs versus H2RAs 0.88 (0.09 to 8.19); NA 0.87 (0.07 to 10.93); H2RAs: 14 (1 to 257) PPIs: 12 (4 to 48) 17 Moderate‡ Low‡§ Difference: -2 (-229 to 22) 18CrI=credible interval; PPIs=proton pump inhibitors; NA=not applicable; H2RAs=histamine-2 receptor antagonists. 19 20*Higher of direct or indirect confidence (without consider imprecision), and then considered imprecision and incoherence. 21†We used as baseline risk in the placebo group of the SUP-ICU trial. 22‡Rated down for risk of bias. 23§Rated down for imprecision. 24 25¶Rated down 2 levels for imprecision. 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 94 of 109

1 2 3 4 5Table 3 GRADE summary of findings for overt gastrointestinal bleeding for different comparisons 6 7Comparison Study results (95% CI) Absolute effect estimates (per Absolute Certainty Plain text summary 8 and measurementsConfidential:1000) Fordifference Review (95% in effect Only 9 CI) (per 1000) estimates 10PPIs versus placebo OR 0.59 (0.45 to 0.77). Low risk Placebo: PPIs: 16 -10 (-14 to -6) High PPIs reduce overt bleeding 11 12 3874 patients in 5 26 13 studies with low risk of Moderate risk 75 46 -29 (-40 to -16) High PPIs reduce overt bleeding 14 bias. High risk 125 78 -47 (-65 to -26) High PPIs reduce overt bleeding 15 16 Highest risk 190 122 -68 (-95 to -37) High PPIs reduce overt bleeding 17H2RAs versus OR 0.38 (0.24 to 0.59). Low risk Placebo: H2RAs: -16 (-20 to -10) Moderate* H2RAs probably reduce overt bleeding 18placebo 2442 patients in 29 26 10 19 20 studies. Moderate risk 75 30 -45 (-56 to -29) Moderate* H2RAs probably reduce overt bleeding 21 High risk 125 51 -74 (-92 to -47) Moderate* H2RAs probably reduce overt bleeding 22 Highest risk 190 82 -108 (-137 to -68) Moderate* H2RAs probably reduce overt bleeding 23 24Sucralfate versus OR 0.58 (0.30 to 1.11). Low risk Placebo: Suc: 15 -11 (-18 to 3) Moderate† Sucralfate probably reduces overt bleeding 25placebo 900 patients in 7 studies. 26 26 Moderate risk 75 45 -30 (-51 to 8) Moderate† Sucralfate probably reduces overt bleeding 27 28 High risk 125 77 -48 (-84 to 12) Moderate† Sucralfate probably reduces overt bleeding 29 Highest risk 190 120 -70 (-124 to 17) Moderate† Sucralfate probably reduces overt bleeding 30PPIs versus H2RAs OR 0.44 (0.28 to 0.69). Low risk H2RAs: PPIs: 16‡ -20 (-39 to -7) High PPIs reduce overt bleeding more than H2RAs 31 32 2092 patients in 17 36 33 studies. Moderate risk 99 46‡ -53 (-101 to -19) High PPIs reduce overt bleeding more than H2RAs 34 High risk 161 78‡ -83 (-154 to -31) High PPIs reduce overt bleeding more than H2RAs 35 36 Highest risk 240 122‡ -118 (-210 to -46) High PPIs reduce overt bleeding more than H2RAs 37PPIs versus OR 0.21 (0.03 to 1.30). Low risk Suc: 72 PPIs: 16‡ -56 (-335 to 4) Moderate† PPIs probably reduce overt bleeding compared 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 95 of 109 BMJ

1 2 3 4 5sucralfate 211 patients in 2 studies. with sucralfate 6 Moderate risk 187 46‡ -141 (-570 to 10) Moderate† PPIs probably reduce overt bleeding compared 7 with sucralfate 8 Confidential: For Review Only 9 High risk 287 78‡ -209 (-660 to 17) Moderate† PPIs probably reduce overt bleeding compared 10 with sucralfate 11 12 Highest risk 398 122‡ -276 (-700 to 25) Moderate† PPIs probably reduce overt bleeding compared 13 with sucralfate 14H2RAs versus OR 0.46 (0.23 to 0.91). Low risk Suc: 21 H2RAs: -11 (-32 to -1) High H2RAs reduce overt bleeding compared with 15sucralfate 1340 patients in 2 10§ sucralfate 16 17 studies with low risk of Moderate risk 63 30§ -33 (-89 to -3) High H2RAs reduce overt bleeding compared with 18 bias. sucralfate 19 High risk 104 51§ -53 (-138 to -5) High H2RAs reduce overt bleeding compared with 20 21 sucralfate 22 Highest risk 162 82§ -80 (-198 to -7) High H2RAs reduce overt bleeding compared with 23 sucralfate 24 25CI=confidence interval; GI=gastrointestinal; PPIs=proton pump inhibitors; OR=odds ratio; H2RAs=histamine-2 receptor antagonists. 26*Rated down for inconsistency (I2 = 55%). 27†Rated down for imprecision. 28 29‡We used the point estimate of absolute effect for PPIs, obtained from PPIs versus placebo, to calculate absolute effect for PPIs versus H2RAs and PPIs versus sucralfate. 30§We used the point estimate of absolute effect for H2RAs, obtained from H2RAs versus placebo, to calculate absolute effect for H2RAs versus sucralfate. 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 96 of 109

1 2 3 4 5 Table 4 GRADE summary of findings for length of ICU stay for different comparisons 6 Comparison Direct estimate* (95% CrI); Indirect estimate* (95% CrI); Network estimate* (95% CrI); 7 Certainty of evidence Certainty of evidence Certainty of evidence† 8 Confidential: For Review Only 9 PPIs versus placebo -0.58 (-1.69 to 0.64); High 0.25 (-1.40 to 2.26); Moderate‡ -0.33 (-1.22 to 0.62); Moderate§ 10 H2RAs versus placebo 0.20 (-1.39 to 1.86); Moderate‡ -0.90 (-2.21 to 0.40); High -0.45 (-1.38 to 0.60); Moderate§ 11 12 Sucralfate versus placebo -0.90 (-3.24 to 1.45); Moderate‡ -0.46 (-2.04 to 1.33); Moderate‡ -0.61 (-1.82 to 0.63); Low‡§ 13 PPIs versus H2RAs 0.22 (-0.73 to 1.27); High -0.36 (-2.16 to 1.61); Moderate‡ 0.12 (-0.72 to 0.89); High 14 PPIs versus sucralfate -0.12 (-3.66 to 3.55); High 0.33 (-0.92 to 1.73); High 0.27 (-0.86 to 1.45); Moderate§ 15 16 H2RAs versus sucralfate 0.13 (-0.91 to 1.29); High 0.38 (-2.39 to 3.18); Moderate‡ 0.16 (-0.75 to 1.17); Moderate§ 17 ICU=intensive care unit; CrI=credible interval; PPIs=proton pump inhibitors; H2RAs=histamine-2 receptor antagonists. 18 *Mean difference (day). 19 20 †Higher of direct or indirect confidence (without consider imprecision), and then considered imprecision and incoherence. 21 ‡Rated down for risk of bias. 22 §Rated down for imprecision. 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 97 of 109 BMJ

1 2 3 4 5 Table 5 GRADE summary of findings for length of hospital stay for different comparisons 6 Comparison Direct estimate* (95% CrI); Indirect estimate* (95% CrI); Network estimate* (95% CrI); 7 Certainty of evidence Certainty of evidence Certainty of evidence† 8 Confidential: For Review Only 9 PPIs versus placebo 0.96 (-1.71, 3.73); High NA 0.97 (-1.74, 3.69); Moderate§ 10 H2RAs versus placebo NA -0.66 (-4.58, 3.34); Moderate‡ -0.66 (-4.58, 3.34); Low‡§ 11 12 Sucralfate versus placebo NA -0.87 (-5.08, 3.48); Moderate‡ -0.87 (-5.08, 3.48); Low‡§ 13 PPIs versus H2RAs 1.92 (-0.76, 4.60); Moderate‡ 0.71 (-5.18, 6.57); Moderate‡ 1.63 (-1.29, 4.56); Low‡§ 14 PPIs versus sucralfate 1.23 (-3.06, 5.52); Moderate‡ 2.44 (-2.44, 7.26); Moderate‡ 1.83 (-1.52, 5.13); Low‡§ 15 16 H2RAs versus sucralfate 0.48 (-2.93, 3.87); Moderate‡ -0.69 (-6.55, 5.14); Moderate‡ 0.21 (-2.69, 3.06); Low‡§ 17 CrI=credible interval; PPIs=proton pump inhibitors; NA=not applicable; H2RAs=histamine-2 receptor antagonists. 18 *Mean difference (day). 19 20 †Higher of direct or indirect confidence (without consider imprecision), and then considered imprecision and incoherence. 21 ‡Rated down for risk of bias. 22 §Rated down for imprecision. 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 98 of 109

1 2 3 4 5 Table 6 GRADE summary of findings for duration of mechanical ventilation different comparisons 6 Comparison Direct estimate* (95% CrI); Indirect estimate* (95% CrI); Network estimate* (95% CrI); 7 Certainty of evidence Certainty of evidence Certainty of evidence† 8 Confidential: For Review Only 9 PPIs versus placebo -0.18 (-1.47 to 1.27); High -1.26 (-3.23 to 0.74); Moderate‡ -0.40 (-1.40 to 0.71); Moderate§ 10 H2RAs versus placebo -1.55 (-3.50 to 0.38); Moderate‡ -0.58 (-2.13 to 1.16); Moderate‡ -0.78 (-1.85 to 0.30); Low‡§ 11 12 Sucralfate versus placebo 0.004 (-2.58 to 2.49); High -0.86 (-2.53 to 0.60); Moderate‡ -0.53 (-1.74 to 0.62); Low¶‡§ 13 PPIs versus H2RAs 0.29 (-0.45 to 1.18); Moderate‡ 1.09 (-0.71 to 2.96); Moderate‡ 0.37 (-0.26 to 1.19); Low‡§ 14 PPIs versus sucralfate 0.49 (-2.17 to 3.21); High 0.12 (-0.79 to 1.38); Moderate‡ 0.12 (-0.69 to 1.22); Low¶‡§ 15 16 H2RAs versus sucralfate -0.22 (-0.90 to 0.58); Moderate‡ 0.66 (-3.13 to 4.27); Moderate‡ -0.25 (-0.87 to 0.47); Low‡§ 17 CrI=credible interval; PPIs=proton pump inhibitors; H2RAs=histamine-2 receptor antagonists. 18 *Mean difference (day). 19 20 †Higher of direct or indirect confidence (without consider imprecision), and then considered imprecision and incoherence. 21 ‡Rated down for risk of bias. 22 §Rated down for imprecision. 23 ¶The indirect evidence contributed much more than direct evidence, so certainty of network started rating from the quality of indirect estimate. 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 99 of 109 BMJ

1 2 3 4 5 Appendix 9: Meta-regression and subgroup analysis 6 7 Table 1 Network meta-regression for risk of bias 8 Confidential: For Review Only 9 Outcome 50% (95% CrI) 10 Mortality Adjusted for sequence generation beta[1] -0.07 (-27.09, 27.57) 11 beta[3] 0.19 (-2.86, 3.38) 12 13 beta[4] -0.22 (-3.32, 2.92) 14 Adjusted for allocation sequence concealment beta[1] 0.001 (-0.41, 0.41) 15 beta[3] 0.08 (-0.50, 0.66) 16 17 beta[4] -0.06 (-0.55, 0.42) 18 Adjusted for blinding beta[1] 0.17 (-0.24, 0.57) 19 beta[3] 0.04 (-0.42, 0.51) 20 21 beta[4] 0.02 (-0.49, 0.52) 22 Adjusted for missing outcome data beta[1] -0.34 (-0.78, 0.11) 23 beta[3] -0.21 (-0.79, 0.37) 24 25 beta[4] -0.44 (-1.04, 0.17) 26 Adjusted for other bias All are low risk of bias 27 Pneumonia Adjusted for sequence generation beta[1] 0.59 (-2.18, 3.42) 28 29 beta[3] 0.24 (-2.54, 3.04) 30 beta[4] -0.87 (-3.73, 1.89) 31 Adjusted for allocation sequence concealment beta[1] 0.31 (-0.46, 1.12) 32 33 beta[3] 0.22 (-0.66, 1.06) 34 beta[4] 0.07 (-0.90, 0.99) 35 Adjusted for blinding beta[1] 0.51 (-0.18, 1.21) 36 beta[3] 0.56 (-0.25, 1.29) 37 38 beta[4] 0.23 (-0.73, 1.13) 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 100 of 109

1 2 3 4 5 Adjusted for missing outcome data beta[1] -0.08 (-0.90, 0.71) 6 beta[3] 0.08 (-0.85, 0.97) 7 beta[4] -0.07 (-1.06, 0.88) 8 Confidential: For Review Only 9 Adjusted for other bias All are low risk of bias 10 Clostroides difficile infection Adjusted for sequence generation All are low risk of bias 11 12 Adjusted for allocation sequence concealment beta[1] 0.01 (-7.40, 7.45) 13 beta[3] -0.01 (-7.93, 6.95) 14 Adjusted for blinding beta[1] 0.01 (-7.40, 7.45) 15 16 beta[3] -0.01 (-7.93, 6.95) 17 Adjusted for missing outcome data All are low risk of bias 18 Adjusted for other bias All are low risk of bias 19 20 Length of ICU stay Adjusted for sequence generation All are low risk of bias 21 Adjusted for allocation sequence concealment beta[1] 0.43 (-1.81, 2.64) 22 beta[3] 0.54 (-1.75, 2.61) 23 beta[4] -0.03 (-2.58, 2.48) 24 25 Adjusted for blinding beta[1] 0.19 (-2.19, 2.45) 26 beta[3] 0.62 (-1.48, 2.70) 27 beta[4] -0.07 (-2.62, 2.50) 28 29 Adjusted for missing outcome data beta[1] 0.69 (-1.29, 2.59) 30 beta[3] -0.10 (-64.80, 26.07) 31 beta[4] 0.26 (-2.54, 3.08) 32 33 Adjusted for other bias All are low risk of bias 34 Length of hospital stay Adjusted for sequence generation All are low risk of bias 35 Adjusted for allocation sequence concealment beta[1] 0.26 (-4.88, 5.98) 36 37 beta[3] 0.01 (-18.43, 17.71) 38 beta[4] -0.24 (-6.01, 4.77) 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 101 of 109 BMJ

1 2 3 4 5 Adjusted for blinding beta[1] 0.10 (-12.34, 24.99) 6 beta[3] 0.22 (-4.77, 6.28) 7 beta[4] -0.24 (-6.01, 5.01) 8 Confidential: For Review Only 9 Adjusted for missing outcome data All are low risk of bias 10 Adjusted for other bias All are low risk of bias 11 12 Duration of mechanical ventilation Adjusted for sequence generation beta[1] -0.19 (-48.54, 40.50) 13 beta[3] 0.52 (-8.84, 10.40) 14 beta[4] -0.38 (-9.84, 9.42) 15 Adjusted for allocation sequence concealment beta[1] -0.79 (-3.15, 1.45) 16 17 beta[3] -0.81 (-3.11, 1.45) 18 beta[4] -1.61 (-4.11, 0.84) 19 Adjusted for blinding beta[1] -0.36 (-2.59, 1.76) 20 21 beta[3] -0.30 (-2.33, 1.68) 22 beta[4] -1.36 (-3.75, 0.93) 23 Adjusted for missing outcome data beta[1] -1.65 (-4.29, 0.86) 24 25 beta[3] -1.53 (-4.67, 1.44) 26 beta[4] -1.65 (-4.74, 1.35) 27 Adjusted for other bias All are low risk of bias 28 29 CrI=credible interval; [1]=H2RAs versus placebo; [3]=PPIs versus placebo; [4]=sucralfate versus placebo; ICU=intensive care unit. 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 102 of 109

1 2 3 4 5 Table 2 Meta-regression for mechanical ventilation 6 7 Number of studies P-value 8 Clinically importantConfidential: gastrointestinal bleeding PPIs versus placebo For ReviewNA Only- 9 10 H2RAs versus placebo 7 0.54 11 Sucralfate versus placebo NA - 12 13 PPIs versus H2RAs 10 0.45 14 PPIs versus sucralfate NA - 15 16 H2RAs versus sucralfate NA - 17 Overt gastrointestinal bleeding PPIs versus placebo 6 0.13 18 19 H2RAs versus placebo 17 0.61 20 Sucralfate versus placebo 5 0.92 21 22 PPIs versus H2RAs 12 0.85 23 PPIs versus sucralfate NA - 24 25 H2RAs versus sucralfate 24 0.38 26 NA=not applicable. 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 103 of 109 BMJ

1 2 3 4 5 Table 3 Meta-regression for enteral nutrition 6 7 Number of studies P-value 8 Clinically important gastrointestinalConfidential: bleeding PPIs versus placebo For7 Review0.70 Only 9 10 H2RAs versus placebo 5 0.88 11 Sucralfate versus placebo 4 0.67 12 13 PPIs versus H2RAs 5 0.47 14 PPIs versus sucralfate 1 - 15 16 H2RAs versus sucralfate 10 0.13 17Overt gastrointestinal bleeding PPIs versus placebo 8 0.14 18 19 H2RAs versus placebo 12 0.03 (the subgroup which suggested that H2RAs had smaller 20 relative effect on overt bleeding when patients received enteral 21 22 nutrition is of low credibility)* 23 Sucralfate versus placebo 4 0.48 24 25 PPIs versus H2RAs 8 0.17 26 PPIs versus sucralfate 1 - 27 28 H2RAs versus sucralfate 11 0.18 29*P <0.05. As the proportion of enteral nutrition increases, the relative effect of H2RAs versus placebo on overt bleeding will closer to 1, which represents the possibility 30 1 31that H2RAs had smaller relative effect on overt bleeding when patients received enteral nutrition. We applied the criteria to assess the credibility of the subgroup 32analysis: 33 341. The subgroup variable is a characteristic measured at baseline. 352. The effect is suggested by comparisons between studies. 36 373. The hypothesis that enteral nutrition may provide protection against gastrointestinal bleeding and the impact of prophylaxis in patients receiving enteral nutrition may 38therefore be negligible or absent was specified a priori. 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 104 of 109

1 2 3 4 5 4. The direction of the subgroup effect (the impact of prophylaxis in patients receiving enteral nutrition may therefore be negligible or absent) was specified a priori. 6 7 5. The subgroup effect was one of a small number of hypothesised effects tested. 8 6. The interaction test partially Confidential:suggest a low likelihood that chance explains Forthe apparent Reviewsubgroup effect. Only 9 7. Whether the significant subgroup effect is independent is not relevant. 10 118. The size of the subgroup effect is large. 129. The interaction is partly consistent across studies. 13 1410. Whether the interaction is consistent across closely related outcomes within studies or not is of uncertainty. 1511. Indirect evidence supports the hypothesized interaction (biological rationale). 16 17Therefore, the subgroup of low credibility. 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 105 of 109 BMJ

1 2 3 4 5Table 4 Subgroup analysis for critically ill patients versus critically ill patients with risk factor 6 7 Critically ill patients Critically ill patients Subgroup difference (p value) 8 Confidential: Forwith risk factor Review Only 9 10Clinically important PPIs versus placebo NA 0.61 (0.42, 0.89) - 11gastrointestinal H2RAs versus placebo 0.43 (0.20, 0.94) 0.51 (0.22, 1.20) 0.77 12 13bleeding Sucralfate versus placebo 0.60 (0.26, 1.38) 2.28 (0.39, 13.49) 0.18 14 PPIs versus H2RA 2.73 (0.11, 68.36) 0.38 (0.21, 0.68) 0.24 15 16 PPIs versus sucralfate NA 0.30 (0.05, 1.92) - 17 H2RAs versus sucralfate 0.80 (0.32, 2.02) 0.83 (0.47, 1.48) 0.94 18 19Overt gastrointestinal PPIs versus placebo 0.22 (0.09, 0.54) 0.58 (0.45, 0.75) 0.04 (the subgroup which suggested that PPIs reduce overt bleeding 20bleeding more in studies which only specified inclusion of critically ill 21 22 patients than in studies included critically ill patients with risk 23 factors for gastrointestinal bleeding is of low credibility)* 24 25 H2RAs versus placebo 0.41 (0.23, 0.73) 0.35 (0.17, 0.72) 0.77 26 Sucralfate versus placebo 0.59 (0.29, 1.19) 0.59 (0.11, 3.06) 1.00 27 28 PPIs versus H2RA 0.71 (0.16, 3.11) 0.39 (0.23, 0.67) 0.47 29 PPIs versus sucralfate NA 0.21 (0.03, 1.30) - 30 31 H2RAs versus sucralfate 1.05 (0.54, 2.03) 1.36 (0.82, 2.27) 0.53 32*P <0.05. For PPIs versus placebo, studies in which investigators specified that patients had risk factors had apparently smaller effects on overt bleeding that those in which 33 1 34there was no mention or risk factors. We applied the criteria to assess the credibility of the subgroup analysis: 351. The subgroup variable is a characteristic measured at baseline. 36 372. The effect is suggested by comparisons between studies. 383. The hypothesis was not specified a priori. 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 106 of 109

1 2 3 4 54. The direction of the subgroup effect was not specified a priori. 6 75. The subgroup effect was one of a small number of hypothesised effects tested. 86. The interaction test partially suggestConfidential: a low likelihood that chance explains the For apparent subgroup Review effect. Only 97. Whether the significant subgroup effect is independent is not relevant. 10 118. The size of the subgroup effect is large. 129. The interaction is partly consistent across studies. 13 1410. Whether the interaction is consistent across closely related outcomes within studies or not is of uncertainty. 1511. Indirect evidence does not support the hypothesized interaction (biological rationale). 16 17Therefore, the subgroup of low credibility. 18 19 20 21 22 23 24 25 26 27 References 28 29 1. Sun X, Briel M, Walter SD, et al. Is a subgroup effect believable? Updating criteria to evaluate the credibility of subgroup analyses. BMJ 2010;340:c117. 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 107 of 109 BMJ

1 2 3 4 5 Table 3 Low versus high risk of bias results for sequence generation for overt gastrointestinal bleeding 6 Comparison Low risk of bias result High risk of bias result Subgroup difference (p value) 7 8 PPIs versus placeboConfidential:All studies were at low risk Forof bias Review Only 9 Table 1 Low versus high risk of bias results for sequence generation for clinically important gastrointestinal bleeding 10 Comparison Low risk of bias result High risk of bias result Subgroup difference (p value) 11 12 PPIs versus placebo All are low risk of bias 13 H2RAs versus placebo All are low risk of bias 14 Sucralfate versus placebo All are low risk of bias 15 16 PPIs versus H2RAs All are low risk of bias 17 PPIs versus sucralfate All are low risk of bias 18 H2RAs versus sucralfate All are low risk of bias 19 20 Appendix 10: Subgroup analysis for risk of bias for clinically important gastrointestinal bleeding and overt gastrointestinal bleeding 21 22 23 24 25 Table 2 Low versus high risk of bias results for missing outcome data for clinically important gastrointestinal bleeding 26 Comparison Low risk of bias result High risk of bias result Subgroup difference (p value) 27 28 PPIs versus placebo All are low risk of bias 29 H2RAs versus placebo 0.38 (0.19, 0.73) 0.84 (0.32, 2.17) 0.17 30 Sucralfate versus placebo 0.73 (0.33, 1.60) 1.26 (0.07, 21.27) 0.72 31 32 PPIs versus H2RAs 0.37 (0.19, 0.71) 0.56 (0.16, 1.97) 0.56 33 PPIs versus sucralfate All are low risk of bias 34 H2RAs versus sucralfate 0.71 (0.43, 1.17) 1.19 (0.43, 3.28) 0.37 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 108 of 109

1 2 3 4 5 H2RAs versus placebo 0.39 (0.25, 0.62) 0.11 (0.01, 0.86) 0.23 6 Sucralfate versus placebo All studies were at low risk of bias 7 8 PPIs versus H2RAsConfidential:0.43 (0.27, 0.71) For0.26 Review(0.01, 6.77) Only0.76 9 PPIs versus sucralfate All studies were at low risk of bias 10 H2RAs versus sucralfate All studies were at low risk of bias 11 12 13 14 Table 4 Low versus high risk of bias results for missing outcome data for overt gastrointestinal bleeding 15 16 Comparison Low risk of bias result High risk of bias result Subgroup difference (p value) 17 PPIs versus placebo 0.58 (0.45, 0.75) 0.22 (0.09, 0.54) 0.04* 18 H2RAs versus placebo 0.29 (0.17, 0.50) 0.73 (0.33, 1.60) 0.06 19 20 Sucralfate versus placebo 0.55 (0.27, 1.13) 1.26 (0.07, 21.27) 0.58 21 PPIs versus H2RAs 0.36 (0.22, 0.59) 0.73 (0.36, 1.51) 0.11 22 PPIs versus sucralfate All studies were at low risk of bias 23 24 H2RAs versus sucralfate 1.40 (0.79, 2.48) 1.03 (0.62, 1.70) 0.43 25 *Important difference between low and high risk of bias result, so we used low risk of bias result as best estimate. 26 27 28 29 Subgroup analysis for concealment/blinding for clinically important gastrointestinal bleeding in table 1. 30 Subgroup analysis for concealment/blinding for overt gastrointestinal bleeding in table 4. 31 All studies are at low risk of bias for other bias. 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 109 of 109 BMJ

1 2 3 4 5 Appendix 11: Baseline risk of clinically important gastrointestinal bleeding and overt gastrointestinal bleeding for each risk group 6 Risk group Risk factors Clinically important gastrointestinal bleeding Overt gastrointestinal bleeding 7 Baseline risk of Representative risk Baseline risk of Representative 8 Confidential: For Review Only 9 clinically important chosen for evidence overt risk chosen for 10 gastrointestinal profile gastrointestinal evidence profile 11 bleeding (per 1000) bleeding (per 12 13 1000) 14Low risk (including Critically ill patients without any risk factor; 10-20 12 20-60 26 15proposed risk factors 16 17without evidence that they Acute hepatic failure; 18substantially increase the 19risk of bleeding) Use of steroids/immunosuppression; 20 21 Use of anticoagulants; 22 23 24 Cancer; 25 26 Male gender 27 28Moderate risk Mechanical ventilation with enteral nutrition; 21-40 30 61-90 75 29 30 Shock; 31 32 33 Sepsis; 34 35 Acute kidney injury 36 37High risk Coagulopathy 41-80 60 91-160 125 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 110 of 109

1 2 3 4 5 Highest risk Mechanical ventilation without enteral 81-100 90 161-220 190 6 nutrition; 7 8 Confidential: For Review Only 9 Chronic liver disease 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60