Publication

• Title: Effect of stress ulcer prophylaxis with proton pump inhibitors vs histamine-2 receptor blockers on in-hospital mortality among ICU patients receiving invasive mechanical ventilation: the PEPTIC randomized clinical trial
• Acronym: PEPTIC
• Year: 2020
• Journal published in: JAMA
• Citation: The PEPTIC Investigators for the Australian and New Zealand Intensive Care Society Clinical Trials Group, Alberta Health Services Critical Care Strategic Clinical Network, and the Irish Critical Care Trials Group. Effect of stress ulcer prophylaxis with proton pump inhibitors vs histamine-2 receptor blockers on in-hospital mortality among ICU patients receiving invasive mechanical ventilation: the PEPTIC randomized clinical trial. JAMA. 2020;323(7):616-626.

Context & Rationale

• Background

  Stress-related mucosal disease and overt upper gastrointestinal bleeding are recognised complications in mechanically ventilated ICU patients, but absolute bleeding rates have fallen with contemporary ICU practice (earlier enteral feeding, haemodynamic resuscitation, and shorter ventilation durations).
  Histamine-2 receptor blockers (H2RBs) have long been used for stress ulcer prophylaxis, and proton pump inhibitors (PPIs) became common because of more potent acid suppression and (in earlier evidence) lower bleeding risk, while concerns grew about potential infectious and other adverse effects of PPIs (eg, pneumonia, C. difficile).
  Prior to PEPTIC, direct large-scale randomised comparisons of PPIs vs H2RBs in modern ICU populations were limited, and whether any differences in bleeding translated into patient-centred outcomes (especially mortality) remained uncertain.
• Research Question/Hypothesis

  In adult ICU patients receiving invasive mechanical ventilation, does a default PPI strategy (vs a default H2RB strategy) for stress ulcer prophylaxis alter in-hospital all-cause mortality (to 90 days), and how does it affect clinically important upper GI bleeding and C. difficile infection?
• Why This Matters

  Stress ulcer prophylaxis is one of the most frequent ICU “background” pharmacotherapies; even small relative effects could have large population impact given ubiquity.
  Choosing between PPIs and H2RBs represents a classic prophylaxis trade-off: preventing rare but catastrophic bleeding versus potentially increasing infections or other harm; PEPTIC aimed to quantify this balance in real-world ICU delivery at scale.

Design & Methods

• Research Question: Does a default strategy of PPI use (vs default H2RB use) for stress ulcer prophylaxis in invasively ventilated ICU patients change in-hospital all-cause mortality up to 90 days after index ICU admission?
• Study Type: Pragmatic, multicentre, international, cluster-randomised, cluster crossover, registry-embedded, open-label randomised clinical trial conducted in 50 ICUs (Australia, New Zealand, Canada, England, Ireland) with two treatment periods per ICU (no washout).
• Population:
  • Setting: Adult ICUs contributing to national/regional ICU registries and implementing ICU-level default stress ulcer prophylaxis strategies.
  • Inclusion: Adults admitted to participating ICUs who received invasive mechanical ventilation during the index ICU admission.
  • Key exclusions: Previously admitted to ICU during the same hospitalisation; admitted to ICU with upper gastrointestinal bleeding (registry-coded); ICU stay >48 hours before first invasive mechanical ventilation; >24 hours between ICU admission and first invasive mechanical ventilation; jurisdictional exclusions for registry linkage in some regions; age <16 years (Australia/New Zealand) or <18 years (Canada).
• Intervention:
  • Default PPI strategy: During the ICU’s PPI period, PPIs were the default stress ulcer prophylaxis drug class (agent and route per local formulary/practice), used at clinician discretion for individual patients.
  • Duration: Until death, ICU discharge, development of clinically important upper GI bleeding, or clinician decision that prophylaxis was no longer indicated.
  • Rescue: If upper GI bleeding occurred, PPI therapy could be given irrespective of randomisation, consistent with standard care.
• Comparison:
  • Default H2RB strategy: During the ICU’s H2RB period, H2RBs were the default stress ulcer prophylaxis drug class (agent and route per local formulary/practice), used at clinician discretion for individual patients.
  • No mandated prophylaxis: As a pragmatic “default strategy” trial, prophylaxis use remained clinician-driven; the intervention targeted the default class when prophylaxis was provided.
• Blinding: Unblinded (cluster-level default strategy implementation); primary outcome (mortality) was registry-derived and objective, but treatment delivery and some secondary outcomes were susceptible to performance/ascertainment differences.
• Statistics: Power calculation: 80% power (two-sided α=0.05) to detect a 2.4% absolute reduction in in-hospital mortality (15.0% to 12.6%; relative risk reduction 16%), assuming mean cluster-period size 310, coefficient of variation 0.50, within-cluster within-period correlation 0.035, within-cluster between-period correlation 0.025 (autocorrelation 0.71). Primary analysis: intention-to-treat at the level of cluster assignment using individual-level generalised estimating equations (log link), robust SEs, ICU as clustering unit; adjustment for randomisation batch, period order, and batch×order interaction; within-ICU effect estimate emphasised.
• Follow-Up Period: Primary outcome: in-hospital mortality during the index hospitalisation up to 90 days from index ICU admission; secondary/tertiary outcomes captured during ICU/hospitalisation (time-to-discharge with competing risk methods for mortality).

Key Results

This trial was not stopped early. Recruitment occurred from August 2016 to January 2019 across 50 ICUs, completing planned crossover implementation.

• In a large, pragmatic ICU-wide default strategy comparison, PPIs reduced clinically important upper GI bleeding (1.3% vs 1.8%) but did not demonstrate a statistically significant mortality difference (18.3% vs 17.5%; P=0.054), with point estimates compatible with a small mortality increase or no effect.
• Substantial non-adherence diluted separation: estimated PPI receipt was 82.5% in the PPI period vs 20.1% in the H2RB period; estimated H2RB receipt was 4.1% in the PPI period vs 63.6% in the H2RB period.
• Safety signals that motivated clinical equipoise (notably C. difficile) were not increased, but absolute event rates were very low and ascertainment heterogeneous, supporting cautious inference.

Internal Validity

• Randomisation and allocation: ICU-level randomisation (in batches) to treatment order, then crossover; allocation concealment at patient level is not applicable, and treatment implementation was operational rather than concealed.
• Dropout/exclusions: 154/26,982 (0.6%) opted out; primary outcome missing for 57 participants; primary analysis included 13,415 vs 13,356 for mortality.
• Performance/detection bias: Unblinded “default strategy” delivery creates risk of differential co-interventions and clinician override; primary outcome (mortality) is objective and registry-derived, but bleeding and infection outcomes depend on clinical recognition/testing and data sources.
• Protocol adherence and separation of the variable of interest:
  • Median ICU time (proxy exposure opportunity): 2.8 days (IQR 1.2 to 5.7) in PPI group vs 2.7 days (IQR 1.2 to 5.8) in H2RB group.
  • Estimated exposure by class: PPI group — 82.5% received PPI, 4.1% received H2RB, 1.9% received both, 11.5% received neither.
  • Estimated exposure by class: H2RB group — 63.6% received H2RB, 20.1% received PPI, 5.1% received both, 11.2% received neither.
  • Implication: non-adherence (especially PPI use during H2RB periods) plausibly biased results towards the null for bleeding benefit and attenuated any true harm/benefit on mortality.
• Baseline characteristics: Broadly balanced at scale (mean APACHE II 18.7 in both); case-mix included substantial elective surgery (32.9% overall), which lowers baseline bleeding risk and may limit treatment effect size.
• Heterogeneity: Prespecified effect modification analyses by postoperative cardiac surgery vs other, elective vs emergency surgery, and geographic region; interpretation constrained by pragmatic exposure measurement and by multiple comparisons (secondary endpoints treated as exploratory).
• Timing: Strategy applied at ICU admission through ICU course; crossover periods had no washout, and patients who remained in ICU through crossover continued their original assignment (reducing contamination within-patient but adding period complexity).
• Dose: Drug choice/dose/route were not standardised (agent selection per local formulary); this increases real-world relevance but reduces mechanistic interpretability and may dilute class effects if dosing varied.
• Crossover: Clinically meaningful crossover occurred (notably PPI administration during H2RB periods at 20.1%), explicitly acknowledged by investigators as a key interpretability limitation.
• Adjunctive therapy use: Not comprehensively reported for co-interventions influencing bleeding risk (eg, enteral nutrition timing, anticoagulation/antiplatelets, steroids), which could modulate absolute event rates and treatment effects.
• Outcome assessment: Mortality derived from registries; clinically important upper GI bleeding required haemodynamic/laboratory/transfusion criteria; C. difficile relied on local testing and trial-specific ascertainment methods; ventilator-associated conditions collected in a Canadian subset.
• Statistical rigour: Prespecified analysis plan posted before database lock; cluster-crossover appropriate GEE methods with robust SEs and adjustment for batch/order; multiple secondary outcomes interpreted as exploratory (type I error control limited beyond primary).

Conclusion on Internal Validity: Moderate internal validity: the cluster crossover and objective registry mortality support robust primary outcome ascertainment, but substantial treatment crossover/non-adherence and heterogeneous ascertainment of key secondary outcomes meaningfully limit causal attribution and dilute separation between strategies.

External Validity

• Population representativeness: Very large, multinational ICU cohort (n=26,828) with minimal exclusions beyond prior ICU admission and baseline upper GI bleeding; reflects routine ICU ventilation populations in high-income settings.
• Clinical applicability: The “default strategy” design mirrors how stress ulcer prophylaxis is operationalised in real ICUs (order sets, formulary norms, clinician discretion), improving generalisability to systems using either agent class.
• Limits to generalisability: Exposure opportunity was short (median ICU stay ~3 days), potentially under-representing prolonged, highest-risk bleeding phenotypes; results may not translate directly to ICUs with markedly different ventilation durations, anticoagulation intensity, or testing practices for C. difficile.
• Contemporary relevance: Drug landscape has evolved (eg, ranitidine withdrawal in many settings), so “H2RB strategy” in 2026 practice is often famotidine-based; however, class-level inference remains relevant.

Conclusion on External Validity: External validity is strong for mechanically ventilated ICU populations in high-income, registry-participating systems, but extrapolation to prolonged ventilation/high-bleeding-risk cohorts and to settings with different infection testing and formulary constraints should be cautious.

Strengths & Limitations

• Strengths: Extremely large sample size; pragmatic cluster crossover implementation; multinational and multicentre; registry-embedded objective mortality ascertainment; prespecified analytic plan with appropriate methods for cluster-crossover inference; clinically important bleeding definition anchored to haemodynamic/clinical consequences.
• Limitations: Unblinded strategy implementation with substantial crossover/non-adherence; incomplete and heterogeneous capture of actual drug exposure in some regions; secondary outcomes vulnerable to ascertainment and practice variation (especially C. difficile testing); co-interventions influencing bleeding/infection risk not comprehensively characterised; secondary end points exploratory with multiplicity concerns.

Interpretation & Why It Matters

• Clinical practice signal

  Default PPI strategies modestly reduced clinically important upper GI bleeding (absolute reduction ~0.5%), without clear evidence of increased C. difficile in this dataset; mortality results were statistically non-significant but numerically higher with PPI (RR 1.05; P=0.054), making a small harm vs null effect difficult to exclude.
• Methodological implication

  PEPTIC illustrates the interpretability constraints of pragmatic “strategy” trials when clinician override is common: the policy-level question (what happens if a unit’s default changes) is answered, but mechanistic class effects and individual-level causal inference are diluted by crossover and variable adherence.

Controversies & Subsequent Evidence

• Interpretability under high crossover: Contemporary commentary highlighted that the pragmatic default-strategy design trades internal separation for real-world relevance; the substantial PPI use during H2RB periods (20.1%) likely attenuated both bleeding benefit and any true mortality or infection signal, complicating “class superiority” conclusions.¹
• Borderline primary outcome and inferential framing: The primary outcome narrowly missed the prespecified significance threshold (P=0.054) with a point estimate favouring H2RBs for mortality; this has driven debate about whether clinicians should weight the statistically significant reduction in bleeding more heavily than an imprecise possible mortality harm, particularly given prophylaxis is administered to large numbers to prevent few events.
• Composite endpoint revision: The registered primary endpoint was initially a composite (clinically important bleeding, C. difficile, prolonged ventilation) and was changed prior to completion of recruitment to mortality primacy; while transparent and pre-data-review, endpoint changes can shape interpretive priors and highlight the difficulty of selecting patient-important endpoints for prophylaxis trials.
• Subgroup signals (hypothesis-generating): Prespecified subgroup analyses (eg, geographic region; postoperative phenotypes) suggested variability in direction/magnitude across strata, but interaction tests were not compelling for mortality and multiplicity limits inference; these results are best treated as exploratory and as prompts to examine context-specific co-interventions and baseline risk.
• Subsequent syntheses and guidelines: Later systematic reviews and network meta-analyses incorporating PEPTIC generally support that PPIs reduce clinically important GI bleeding compared with H2RBs, while effects on mortality and infectious complications remain uncertain or small in absolute terms, with residual heterogeneity across trials and designs.²³
• Newer placebo-controlled evidence context: Large contemporary placebo-controlled stress ulcer prophylaxis trials during invasive mechanical ventilation have further reduced certainty that routine acid suppression improves patient-centred outcomes beyond bleeding prevention, reinforcing a risk-targeted approach to prophylaxis selection and duration.⁴
• Guideline impact: Recent critical care pharmacy/ICU society guidance integrates PEPTIC within a broader evidence base and emphasises targeted prophylaxis for patients at meaningful bleeding risk, careful agent selection, and discontinuation when risk resolves rather than routine, prolonged exposure.⁵

Summary

• PEPTIC was a large, multinational, pragmatic cluster crossover trial comparing ICU-level default PPI vs default H2RB stress ulcer prophylaxis strategies in invasively ventilated adults (n=26,828; 50 ICUs).
• Primary outcome: in-hospital mortality by 90 days was not significantly different (18.3% vs 17.5%; RR 1.05; 95% CI 1.00 to 1.10; P=0.054), leaving uncertainty about small harms/benefits.
• Clinically important upper GI bleeding was lower with a default PPI strategy (1.3% vs 1.8%; RR 0.73; 95% CI 0.57 to 0.92; P=0.009).
• C. difficile infection and other tertiary outcomes were uncommon and not significantly different, but ascertainment variability and low event rates limit exclusion of modest effects.
• Interpretation is materially constrained by crossover/non-adherence (notably PPI use during H2RB periods), a central methodological lesson for pragmatic “default strategy” trials.

Further Reading

Other Trials

• 2024Cook D, et al. Stress ulcer prophylaxis during invasive mechanical ventilation. N Engl J Med. 2024.
• 2018Krag M, Marker S, Perner A, et al. Pantoprazole in patients at risk for gastrointestinal bleeding in the ICU. N Engl J Med. 2018;379(23):2199-2208.
• 2017Alhazzani W, Guyatt G, Alshahrani M, et al. Withholding pantoprazole for stress ulcer prophylaxis in critically ill patients: a pilot randomized clinical trial. Crit Care Med. 2017;45(7):1121-1129.
• 2016Selvanderan SP, Summers MJ, Finnis ME, et al. Pantoprazole or placebo for stress ulcer prophylaxis (POP-UP): randomized, double-blind, exploratory study. Crit Care Med. 2016;44(10):1842-1850.
• 1998Cook DJ, Witt LG, Cook RJ, Guyatt GH. Stress ulcer prophylaxis in the critically ill: sucralfate vs ranitidine. N Engl J Med. 1998;338(12):791-797.

Systematic Review & Meta Analysis

• 2022Deliwala SS, et al. Comparative efficacy and safety of proton pump inhibitors vs histamine-2 receptor antagonists for stress ulcer prophylaxis: systematic review and meta-analysis. J Clin Gastroenterol. 2022.
• 2021Lee TH, et al. Efficacy and safety of proton pump inhibitors vs histamine-2 receptor antagonists for stress ulcer prophylaxis in critically ill patients: updated systematic review and meta-analysis. Am J Med. 2021;134(4):524-533.e6.
• 2020Alshamsi F, et al. Efficacy and safety of prophylaxis for stress-related gastrointestinal bleeding in critically ill patients: systematic review and network meta-analysis. Intensive Care Med. 2020;46:1983-1991.
• 2020Wang Y, et al. Efficacy and safety of gastrointestinal bleeding prophylaxis in critically ill patients: systematic review and network meta-analysis. BMJ. 2020;368:l6744.
• 2013Alhazzani W, et al. Proton pump inhibitors vs histamine-2 receptor antagonists for stress ulcer prophylaxis in critically ill patients: systematic review and meta-analysis. Crit Care Med. 2013;41(3):693-705.

Observational Studies

• 2024Pu Y, et al. Proton pump inhibitors vs histamine-2 receptor antagonists for stress ulcer prophylaxis in critically ill patients: observational comparative effectiveness analysis. Clin Ther. 2024.
• 2021Song MJ, et al. Comparison of proton pump inhibitors and histamine-2 receptor antagonists for stress ulcer prophylaxis in critically ill patients using a common data model. Sci Rep. 2021;11:18911.
• 2021Huang H, et al. Proton pump inhibitors vs histamine-2 receptor antagonists for stress ulcer prophylaxis in critically ill adults with sepsis: retrospective cohort. J Int Med Res. 2021;49(7):3000605211025130.
• 2020Franchitti C, et al. Adequacy of stress ulcer prophylaxis prescription in intensive care units: observational audit study. Swiss Med Wkly. 2020;150:w20322.
• 2015Krag M, Perner A, Møller MH. Stress ulcer prophylaxis in the intensive care unit: a multinational, observational study of use and outcomes. Intensive Care Med. 2015;41(5):833-845.

Guidelines

• 2024MacLaren R, et al. SCCM/ASHP clinical practice guidelines for the prevention of stress-related upper gastrointestinal bleeding in critically ill adults. Crit Care Med. 2024.
• 2024MacLaren R, et al. Executive summary: prevention of stress-related upper gastrointestinal bleeding in critically ill adults. Crit Care Med. 2024.
• 2021Evans L, Rhodes A, Alhazzani W, et al. Surviving Sepsis Campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021;47(11):1181-1247.
• 2020Ye Z, Reintam Blaser A, Lytvyn L, et al. Gastrointestinal bleeding prophylaxis for critically ill patients: a clinical practice guideline. BMJ. 2020;368:l6722.

Notes

• PEPTIC evaluated ICU-level default strategies (policy implementation) rather than forced individual-level drug assignment; interpret results as “what happens if the default changes”, not as a pure pharmacological class-effect trial.
• Ranitidine withdrawal in many countries postdates PEPTIC; present-day “H2RB strategy” typically means famotidine, but PEPTIC remains informative for class-level decision-making and for policy-trial methodology.

Overall Takeaway

PEPTIC is landmark because it operationalised a pragmatic, registry-embedded, cluster crossover approach to answer a ubiquitous ICU prophylaxis question at scale. It demonstrated a bleeding benefit with default PPI strategies but did not show a statistically significant mortality difference, and its interpretation is dominated by a real-world reality: substantial clinician override and crossover can make “policy trials” highly generalisable yet less definitive about drug-class causality.

Bibliography

• 1.Rice TW, Spiegel R, et al. Proton pump inhibitors vs histamine-2 receptor blockers for stress ulcer prophylaxis in critically ill patients: issues of interpretability in pragmatic trials. JAMA. 2020;323(7):611-613.
• 2.Lee TH, et al. Efficacy and safety of proton pump inhibitors vs histamine-2 receptor antagonists for stress ulcer prophylaxis in critically ill patients: updated systematic review and meta-analysis. Am J Med. 2021;134(4):524-533.e6.
• 3.Wang Y, et al. Efficacy and safety of gastrointestinal bleeding prophylaxis in critically ill patients: systematic review and network meta-analysis. BMJ. 2020;368:l6744.
• 4.Cook D, et al. Stress ulcer prophylaxis during invasive mechanical ventilation. N Engl J Med. 2024.
• 5.MacLaren R, et al. SCCM/ASHP clinical practice guidelines for the prevention of stress-related upper gastrointestinal bleeding in critically ill adults. Crit Care Med. 2024.