Skip to main content

Foundational Trials

SUP-ICU

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Pantoprazole in Patients at Risk for Gastrointestinal Bleeding in the ICU
  • Acronym: SUP-ICU
  • Year: 2018
  • Journal published in: New England Journal of Medicine
  • Citation: Krag M, Marker S, Perner A, Wetterslev J, Wise MP, Schefold JC, et al. Pantoprazole in Patients at Risk for Gastrointestinal Bleeding in the ICU. N Engl J Med. 2018;379:2199-2208.

Context & Rationale

  • Background
    • Stress-related mucosal disease and upper gastrointestinal bleeding were historically frequent in critical illness, but modern ICU practice (earlier enteral feeding, lung-protective ventilation, improved haemodynamic support) reduced bleeding incidence, shifting the risk–benefit balance for routine prophylaxis.
    • Acid-suppressive stress ulcer prophylaxis (H2RAs/PPIs) remained common despite uncertainty about patient-important benefits (especially mortality) and plausible harms (e.g., infection through gastric decontamination loss).
    • Prior trials were often small, heterogeneous, focused on bleeding definitions with limited adjudication, and conducted in older ICU eras, limiting confidence about net benefit in contemporary practice.
  • Research Question/Hypothesis
    In acutely admitted adult ICU patients at increased risk of gastrointestinal bleeding, does prophylactic intravenous pantoprazole (vs placebo) reduce 90-day all-cause mortality and clinically important gastrointestinal bleeding without increasing clinically important harms?
  • Why This Matters
    • Stress ulcer prophylaxis is a high-volume ICU intervention with potential for both benefit (bleeding prevention) and harm (infectious and ischaemic complications), so even small absolute effects matter at scale.
    • Establishing contemporary efficacy and safety is essential for rational, risk-stratified prescribing and for deprescribing once ICU bleeding risk abates.
    • The trial tests a core ICU “default” behaviour using rigorous, blinded methodology and patient-centred outcomes, directly informing guidelines and embedded ICU care bundles.

Design & Methods

  • Research Question: Among adult ICU patients at increased risk of gastrointestinal bleeding, does pantoprazole 40 mg IV daily (vs placebo) change 90-day all-cause mortality?
  • Study Type: Investigator-initiated, multicentre, parallel-group, randomised, placebo-controlled, double-blind trial conducted in 33 adult ICUs across Denmark, Finland, the Netherlands, Norway, Switzerland, and the United Kingdom; enrolment January 2016 to October 2017.
  • Population:
    • Inclusion: Adults (≥18 years) acutely admitted to the ICU and considered at increased risk of gastrointestinal bleeding, defined by ≥1 pre-specified risk factor (e.g., shock/vasopressor requirement; coagulopathy thresholds; renal-replacement therapy; anticoagulant treatment; invasive mechanical ventilation expected >24 hours; acute or chronic hepatic failure). 1
    • Key exclusions: Existing/ongoing gastrointestinal bleeding at ICU admission; a clinical indication for acid suppression; current use of a PPI/H2RA at ICU admission (or inability to discontinue); prior randomisation; pregnancy; inability to receive trial intervention per protocol constraints. 1
  • Intervention:
    • Intravenous pantoprazole 40 mg once daily, initiated after randomisation, continued until ICU discharge, death, or day 90 (whichever occurred first).
  • Comparison:
    • Matching intravenous placebo once daily, same initiation and discontinuation rules as the intervention arm; open-label acid suppression permitted only if clinically indicated (protocol violations tracked).
  • Blinding: Double-blind (patients, clinicians, investigators, and statisticians); identical trial drug and placebo; an independent adjudication committee assessed key clinical events while blinded to allocation.
  • Statistics: A total of 3350 patients were required to detect a 20% relative risk reduction in 90-day mortality (from 25% to 20%) with 90% power at the 5% significance level; primary analysis was (modified) intention-to-treat with relative risk and 95% CI; interim monitoring planned with O’Brien–Fleming boundaries.
  • Follow-Up Period: 90 days for the primary outcome; in-ICU outcomes collected during the ICU stay up to day 90.

Key Results

This trial was not stopped early. Planned interim monitoring occurred, but the second interim analysis was not performed because recruitment completed before 90-day follow-up was available for the relevant threshold.

Outcome Pantoprazole Placebo Effect p value / 95% CI Notes
All-cause mortality at day 90 (primary) 510/1642 (31.1%) 499/1640 (30.4%) RR 1.02 95% CI 0.91 to 1.13; P=0.76 Analysed population excludes mis-randomised patients before first dose; 7 withdrew consent for data use and 9 lost to follow-up for this outcome.
Clinically important gastrointestinal bleeding (day 90) 41/1644 (2.5%) 69/1647 (4.2%) RR 0.58 95% CI 0.40 to 0.86 Defined as overt bleeding plus haemodynamic/transfusion/endoscopic consequences; secondary outcomes presented without multiplicity adjustment.
Composite of clinically important events
GI bleeding, pneumonia, C. difficile infection, acute myocardial ischaemia		 360/1644 (21.9%) 372/1647 (22.6%) RR 0.97 95% CI 0.83 to 1.12 Designed to capture net clinical effect across key benefit and harm domains.
One or more infectious adverse events 276/1644 (16.8%) 279/1647 (16.9%) RR 0.99 95% CI 0.84 to 1.16 Infections were prospectively collected; not powered for rare specific pathogens or syndromes.
Days alive without life support (day 90)
Median % of days alive without ventilation, vasopressors, or RRT		 92 (IQR 60–97) 92 (IQR 65–97) Not reported Not reported No evidence of clinically meaningful separation; metric may be insensitive to small bleeding-prevention benefits.
  • Mortality at 90 days was virtually identical (RR 1.02; 95% CI 0.91 to 1.13; P=0.76), excluding a large survival benefit but consistent with both small benefit or small harm.
  • Clinically important gastrointestinal bleeding was reduced (2.5% vs 4.2%; RR 0.58; 95% CI 0.40 to 0.86), demonstrating biological efficacy of acid suppression in a contemporary ICU cohort.
  • No signal of net benefit in the composite “clinically important events” outcome (21.9% vs 22.6%; RR 0.97; 95% CI 0.83 to 1.12), implying that bleeding reduction did not translate into an overall improvement in the prespecified benefit–harm balance.

Internal Validity

  • Randomisation and allocation: Central computer-generated allocation sequence; stratified by site and presence/absence of haematological malignancy; variable block sizes; concealment maintained via indistinguishable pantoprazole and placebo.
  • Drop out / exclusions (post randomisation): 3350 were randomised, but 52 were excluded because randomisation occurred in error before first dose (pantoprazole 23 vs placebo 29); an additional 7 withdrew consent for use of data and 9 were lost to follow-up for 90-day mortality (analysed n=1642 vs 1640 for primary outcome), converting the primary analysis to a modified ITT framework.
  • Performance/detection bias: Double-blinding and placebo control reduce performance bias; primary outcome (mortality) is objective; clinically important bleeding required overt bleeding plus consequence and was adjudicated blinded.
  • Protocol adherence: Major protocol violations were uncommon (46/1645 [2.8%] vs 43/1653 [2.6%]); trial regimen was discontinued early in 274/1645 (16.7%) vs 319/1653 (19.4%).
  • Baseline characteristics: Groups were broadly comparable and severely ill (median SAPS II 49 [IQR 38–62] vs 48 [IQR 37–60]; median SOFA 8 [IQR 5–10] vs 8 [IQR 5–10]); small imbalances (e.g., chronic lung disease 20.7% vs 17.0%) were present but unlikely to drive the primary outcome.
  • Heterogeneity: Prespecified subgroup analyses showed largely consistent mortality effects; a borderline subgroup interaction for higher SAPS II was reported in the main publication (interpretation requires caution due to multiple comparisons and limited biological specificity).
  • Timing: Intervention began after randomisation and was continued only during ICU stay; median exposure duration was short, reflecting contemporary ICU lengths of stay and potentially limiting impact on later complications.
  • Dose: Pantoprazole 40 mg IV once daily is a standard prophylactic dose; the regimen tests typical real-world prescribing rather than high-dose acid suppression.
  • Separation of the variable of interest: Median duration of trial drug 4 days (IQR 2–7) in both groups; concurrent open-label acid suppression while still receiving trial drug was rare (17/1645 [1.0%] vs 17/1653 [1.0%]); enteral nutrition on ICU day 1 was similar (958/1645 [58.2%] vs 926/1653 [56.4%]).
  • Outcome assessment: Mortality is robust; clinically important bleeding definition required objective consequences (haemodynamic instability, transfusion requirement, endoscopic/surgical intervention); the composite outcome explicitly integrated potential harms.
  • Statistical rigour: Primary effect estimate reported with CI and P value; secondary outcomes presented with CIs and without multiplicity adjustment, appropriately limiting inferential weight of secondary comparisons.

Conclusion on Internal Validity: Strong overall: rigorous concealment and blinding, objective primary outcome, and high follow-up completeness support internal validity; the main threats are the modified ITT exclusions (mis-randomisation before first dose) and non-trivial early discontinuation, which modestly reduce confidence in very small effect estimates.

External Validity

  • Population representativeness: Broad, pragmatic inclusion of acutely admitted adult ICU patients with common bleeding-risk features and high illness severity supports applicability to general mixed ICUs in high-income settings.
  • Important exclusions: Patients already receiving (or requiring) acid suppression at admission and those with active gastrointestinal bleeding were excluded; therefore results apply primarily to prophylaxis initiation in acid-suppression–naïve patients rather than continuation strategies.
  • Healthcare system context: Conducted across six European countries; differences in baseline bleeding risk, transfusion practice, endoscopy availability, and infection surveillance may influence absolute effects in other systems.
  • Applicability across subpopulations: Findings are most applicable to unplanned ICU admissions; external validity is less certain in elective postoperative ICUs, highly specialised units, and resource-limited settings with different bleeding/infection epidemiology.

Conclusion on External Validity: Generalisability is good for mixed adult ICUs initiating prophylaxis in high-risk admissions, but limited for patients already on chronic acid suppression or for settings with markedly different baseline bleeding or infection burdens.

Strengths & Limitations

  • Strengths: Large sample size; multicentre, international ICU setting; placebo-controlled double-blind design; robust and clinically meaningful bleeding definition; patient-centred primary outcome (90-day mortality); prespecified interim monitoring; adjudication of key clinical events.
  • Limitations: Modified ITT analysis due to post-randomisation exclusions before first dose; moderate early discontinuation of assigned regimen; secondary outcomes not adjusted for multiplicity; median exposure duration was short (median 4 days), potentially limiting impact on later ICU/hospital events; applicability limited for patients already receiving acid suppression at ICU admission.

Interpretation & Why It Matters

  • Clinical signal
    Pantoprazole reduced clinically important gastrointestinal bleeding (2.5% vs 4.2%) without improving 90-day mortality (31.1% vs 30.4%), implying that bleeding prevention is real but the downstream survival benefit (if any) is small in contemporary ICU care.
  • Decision framing
    The trial supports a risk-stratified approach: prophylaxis may be justified where even small absolute reductions in bleeding are valued, but routine “default” prophylaxis for all ICU patients is not supported by a mortality benefit.
  • Methodological contribution
    SUP-ICU demonstrates that long-standing ICU practices can and should be tested with large, blinded trials using clinically important definitions and composite benefit–harm endpoints rather than relying on historical, surrogate-rich evidence.

Controversies & Subsequent Evidence

  • Modified ITT and post-randomisation exclusions: Excluding mis-randomised patients before first dose strengthens “received treatment” interpretability but weakens strict randomisation-based inference; because exclusions occurred after allocation, this remains a key methodological point when interpreting small effect sizes.
  • Low absolute bleeding risk and clinical value judgement: The absolute reduction in clinically important bleeding (2.5% to 4.2%) is modest; whether this is “worth” prophylaxis depends on patient values, competing risks, and local harms (e.g., infection rates) rather than on mortality alone.
  • Multiplicity and subgroup signals: Secondary outcomes were reported without multiplicity adjustment; subgroup heterogeneity (including a borderline interaction by illness severity in the primary report) should be treated as hypothesis-generating, not practice-changing, given the number of comparisons.
  • Follow-up randomised evidence: A large cluster-crossover trial comparing PPI-based versus H2RA-based prophylaxis (PEPTIC) did not demonstrate a clear mortality advantage of one strategy over the other, supporting the view that any survival effect is likely small. 2
  • Placebo-controlled contemporary replication: A subsequent large placebo-controlled trial in invasively mechanically ventilated patients reported no reduction in 90-day mortality with pantoprazole versus placebo (hazard ratio 0.91; 95% CI 0.77 to 1.08) but fewer clinically important upper gastrointestinal bleeds (1.9% vs 3.5%), reinforcing the “bleeding benefit, mortality-neutral” pattern. 3
  • Evidence synthesis: A 2024 individual-trial–anchored synthesis in NEJM Evidence concluded that PPIs reduce clinically important upper gastrointestinal bleeding in critically ill adults, with little or no effect on mortality; remaining uncertainty centres on rare harms and how best to target prophylaxis to those with greatest net benefit. 4
  • Guideline incorporation: The 2024 SCCM/ASHP guideline emphasises restricting prophylaxis to critically ill adults at increased bleeding risk and discontinuing prophylaxis when risk factors resolve, explicitly incorporating contemporary placebo-controlled evidence. 5

Summary

  • In a large, multicentre, double-blind ICU RCT, pantoprazole prophylaxis did not reduce 90-day mortality versus placebo (31.1% vs 30.4%; RR 1.02; 95% CI 0.91 to 1.13).
  • Pantoprazole reduced clinically important gastrointestinal bleeding (2.5% vs 4.2%; RR 0.58; 95% CI 0.40 to 0.86), confirming efficacy for bleeding prevention in contemporary ICU practice.
  • No difference was seen in the composite “clinically important events” outcome (21.9% vs 22.6%; RR 0.97; 95% CI 0.83 to 1.12), suggesting no net benefit across prespecified benefit–harm domains.
  • Protocol adherence was high (major protocol violations ~2.6–2.8%), but early discontinuation was common (~17–19%) and the primary analysis used a modified ITT due to post-randomisation exclusions.
  • SUP-ICU reframed stress ulcer prophylaxis as a targeted intervention for selected high-risk patients rather than a universal ICU default.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • For several older and observational references, DOI-level details were not available from the sources accessible in this workspace; PubMed search links are provided to support rapid retrieval and verification.

Overall Takeaway

SUP-ICU is a landmark because it challenged a ubiquitous ICU default using a large, rigorous, double-blind, placebo-controlled design and showed a clear reduction in clinically important gastrointestinal bleeding without any signal of improved 90-day survival. Its net message is not that prophylaxis “does nothing”, but that benefits are modest and must be targeted to those at highest bleeding risk, a conclusion reinforced by subsequent trials, meta-analyses, and modern guidelines.

Overall Summary

  • Pantoprazole prophylaxis reduced clinically important GI bleeding but did not reduce 90-day mortality.
  • Net clinical benefit was not demonstrated on a prespecified composite capturing key benefits and harms.
  • Modern practice should emphasise risk stratification and stopping prophylaxis when ICU bleeding risk resolves.

Bibliography