Skip to main content

Foundational Trials

WATERFALL

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Aggressive or Moderate Fluid Resuscitation in Acute Pancreatitis
  • Acronym: WATERFALL
  • Year: 2022
  • Journal published in: New England Journal of Medicine
  • Citation: de-Madaria E, Buxbaum JL, Maisonneuve P, et al. Aggressive or moderate fluid resuscitation in acute pancreatitis. N Engl J Med. 2022;387(11):989-1000.

Context & Rationale

  • Background
    • Early i.v. crystalloid is a near-universal component of acute pancreatitis supportive care, aiming to mitigate hypovolaemia from vomiting, third spacing, and reduced intake.
    • Many guidelines historically promoted “early aggressive hydration”, but the evidentiary base consisted largely of physiology, observational associations (haemoconcentration/BUN), and small heterogeneous trials.
    • Clinicians increasingly recognised a competing hazard: iatrogenic fluid overload (pulmonary oedema/respiratory compromise), especially when severity prediction is imperfect and monitoring intensity varies.
    • A pragmatic, early-presenting ED population trial was needed to adjudicate benefit-versus-harm of weight-based aggressive hydration compared with a more conservative, reassessment-driven strategy.
  • Research Question/Hypothesis
    • Does early aggressive, weight-based lactated Ringer’s resuscitation reduce the development of moderately severe or severe acute pancreatitis compared with a moderate, reassessment-guided strategy, without increasing clinically important harms?
  • Why This Matters
    • Fluid is among the earliest and most ubiquitous “treatments” in acute pancreatitis, often initiated before definitive severity stratification.
    • Even modest absolute increases in fluid-related respiratory morbidity could translate into substantial downstream ICU utilisation and prolonged hospitalisation at population scale.
    • WATERFALL directly tests a guideline-relevant dosing paradigm (bolus + high maintenance rate) against a practical alternative (lower rate, bolus only if hypovolaemic).

Design & Methods

  • Research Question: In adults with early acute pancreatitis presenting to the ED, does aggressive lactated Ringer’s resuscitation (bolus 20 mL/kg then 3 mL/kg/h) reduce moderately severe/severe disease versus a moderate strategy (1.5 mL/kg/h with bolus only if hypovolaemic), and what is the trade-off in fluid overload?
  • Study Type: Randomised, multicentre, open-label, parallel-group, investigator-initiated superiority trial; ED-based enrolment across 18 centres in four countries (India, Italy, Mexico, Spain); enrolment May 2020 to September 2021.
  • Population:
    • Setting: Emergency departments (early disease window).
    • Key inclusion: age ≥18 years; acute pancreatitis diagnosed using standard criteria; symptom onset within 24 hours; diagnosis within 8 hours prior to randomisation; written informed consent.
    • Key exclusions (selected): pancreatitis severity already declaring itself (baseline organ failure); decompensated heart failure; chronic renal failure; other conditions at high risk of fluid complications or requiring alternative resuscitation strategies.
  • Intervention:
    • Aggressive lactated Ringer’s: 20 mL/kg bolus over 2 hours followed by 3 mL/kg/hour infusion.
    • Reassessment checkpoints (3, 12, 24, 48, 72 hours): infusion adjusted per protocolised clinical targets (including signs of hypovolaemia/overload) and stopped/reduced as oral intake resumed.
    • Fluid overload management: mandated reduction/cessation of infusion, with diuretics and supportive therapies at clinician discretion when overload criteria met.
  • Comparison:
    • Moderate lactated Ringer’s: 10 mL/kg bolus over 2 hours only if baseline hypovolaemia present; otherwise no initial bolus.
    • Maintenance 1.5 mL/kg/hour with the same scheduled reassessment and protocolised adjustment/cessation rules as the intervention arm.
    • Rescue/co-interventions: additional therapies (imaging, nutrition, antibiotics, procedures) per treating teams; protocol focused on fluid strategy rather than downstream management algorithms.
  • Blinding: Open-label (fluid rates/boluses not feasibly blinded); raises risk of performance bias for discretionary outcomes (e.g., timing of imaging/discharge), but primary and many key components were clinically anchored and protocol-defined.
  • Statistics: A total sample size of 744 patients was planned to detect a 10% absolute difference in the primary outcome (35% vs 25%) with 80% power at a two-sided 5% significance level, incorporating O’Brien–Fleming interim monitoring and an allowance for attrition, with a modified intention-to-treat primary analysis.1
  • Follow-Up Period: Protocolised clinical reassessments through 72 hours; primary and major secondary outcomes assessed during the index hospitalisation.

Key Results

This trial was stopped early. It was terminated after the first planned interim analysis (249 randomised) because aggressive resuscitation increased fluid overload without evidence of benefit in the primary outcome.

Outcome Aggressive LR (N=122) Moderate LR (N=127) Effect p value / 95% CI Notes
Moderately severe or severe acute pancreatitis (primary) 27/122 (22.1%) 22/127 (17.3%) Adjusted RR 1.30 95% CI 0.78 to 2.18; P=0.32 Composite severity endpoint during index admission
Fluid overload (safety) 25/122 (20.5%) 8/127 (6.3%) Adjusted RR 2.85 95% CI 1.36 to 5.94; P=0.004 Protocol-defined syndrome (symptoms/signs/imaging/haemodynamic evidence)
ICU admission 8/122 (6.6%) 2/127 (1.6%) Adjusted RR 2.71 95% CI 0.64 to 11.51; P=0.17 Low event counts; wide CI
Any organ failure 9/122 (7.4%) 5/127 (3.9%) Adjusted RR 1.23 95% CI 0.47 to 3.23; P=0.67 Organ failure by modified Marshall criteria
Persistent organ failure (>48 h) 8/122 (6.6%) 2/127 (1.6%) Adjusted RR 2.69 95% CI 0.56 to 12.88; P=0.22 Severity-defining component; wide CI
Death 4/122 (3.3%) 1/127 (0.8%) Adjusted RR 3.05 95% CI 0.32 to 28.76; P=0.33 Underpowered; no multiplicity adjustment
Death, persistent organ failure, or infected necrotising pancreatitis 9/122 (7.4%) 4/127 (3.1%) Adjusted RR 1.60 95% CI 0.50 to 5.10; P=0.43 Clinically “hard” composite
Median duration of hospital stay (IQR) 6 (4–8) days 5 (3–7) days Ratio of medians 1.31 95% CI 0.98 to 1.75; Not reported Longer stay consistent with higher overload/respiratory complications
  • No signal of benefit for the primary severity endpoint (22.1% vs 17.3%; Adjusted RR 1.30; 95% CI 0.78 to 2.18; P=0.32).
  • A clear safety signal emerged: fluid overload tripled in relative terms (20.5% vs 6.3%; Adjusted RR 2.85; 95% CI 1.36 to 5.94; P=0.004).
  • Early stopping limits precision for efficacy and low-frequency endpoints (ICU admission, death), but harm was sufficiently pronounced to drive termination.

Internal Validity

  • Randomisation and allocation: Central randomisation with stratification (centre; baseline systemic inflammatory response syndrome; baseline hypovolaemia) and concealed allocation via electronic system; reduces selection bias.
  • Dropout/exclusions: All 249 randomised were included in the reported randomised comparison; missing outcome data were minimal for primary and most clinical endpoints (0 missing for many Table 2 endpoints).
  • Performance/detection bias: Open-label fluid strategy may influence discretionary care (e.g., discharge decisions), but core outcomes (organ failure, death, ICU admission) are relatively objective and protocol-defined.
  • Protocol adherence: Protocol mandated scheduled reassessment at 3, 12, 24, 48, and 72 hours with prespecified rate adjustments; fluid overload triggered reduction/cessation and diuretic use, which likely attenuated ongoing exposure in the aggressive arm after harm emerged.
  • Baseline characteristics: Generally balanced in age (54.7 vs 56.4 years) and BMI (26.8 vs 26.3 kg/m2), but a modest sex imbalance existed (men 44.3% vs 53.5%); aetiology broadly similar (biliary 54.1% vs 54.3%; alcoholic 23.0% vs 21.3%).
  • Heterogeneity: Multinational, 18-centre design improves robustness; stratified randomisation reduces centre effects, but practice variation (imaging thresholds, discharge criteria) could still contribute to heterogeneity in softer endpoints.
  • Timing: Enrolment within 24 hours of symptom onset and early ED initiation directly tests the “early hydration” window that guidelines emphasise.
  • Dose: The aggressive arm is a high-intensity, weight-based regimen (20 mL/kg bolus + 3 mL/kg/h) that maps closely to “aggressive hydration” concepts; the moderate arm still provides substantial fluid via 1.5 mL/kg/h and selective bolus.
  • Separation of the variable of interest: Lactated Ringer’s volume (median, IQR) demonstrated clear early separation, especially 0–12 hours: 3.7 (2.6–4.8) L vs 2.0 (1.5–2.9) L; by 0–72 hours, separation persisted: 8.3 (6.5–11.0) L vs 6.6 (4.8–9.4) L.
  • Key delivery aspects: Protocolised reassessment reduces the risk that “aggressive” simply means “uncalibrated”; however, the safety-driven de-escalation after overload likely reduced cumulative exposure, potentially biasing efficacy estimates towards the null while preserving harm detection.
  • Adjunctive therapy use: Imaging frequency was similar (contrast-enhanced CT 41.8% vs 39.4%), suggesting limited detection bias for necrosis-related endpoints; diuretic use and oxygen supplementation were more common in the aggressive arm, consistent with the overload mechanism.
  • Outcome assessment: Primary endpoint is a composite that includes clinically meaningful severity states (organ failure/local/systemic complications), but composites can dilute interpretability if components vary in importance and measurement intensity.
  • Statistical rigour: Prespecified interim monitoring and planned sample size were appropriate, but early stopping markedly reduced power for efficacy and renders several secondary outcomes hypothesis-generating (especially with multiplicity).

Conclusion on Internal Validity: Moderate. Allocation methods and protocolisation were strong, and the harm signal is internally credible; however, open-label care and early termination materially limit the precision of efficacy and many secondary outcome estimates.

External Validity

  • Population representativeness: Patients were enrolled early in the ED and largely had mild disease at baseline (no baseline organ failure), reflecting a common clinical entry point, but excluding many high-risk patients (cardiac/renal comorbidity; established organ failure) who often complicate real-world fluid decisions.
  • Aetiology mix: Predominantly biliary and alcohol-related pancreatitis; applicability to hypertriglyceridaemia, post-ERCP, or complex iatrogenic cases is less certain.
  • Healthcare system context: Conducted across four countries with heterogeneous ward/ICU thresholds; nevertheless, the biological harm mechanism (volume-related pulmonary congestion) should generalise.
  • Applicability: Findings most directly inform early ED/ward management of uncomplicated acute pancreatitis; extrapolation to patients in shock, with evolving organ failure, or requiring ICU-level resuscitation should be cautious and individualized.

Conclusion on External Validity: Moderate. Results are highly relevant to early ED-presenting acute pancreatitis without established organ failure, but are less generalisable to older, multimorbid patients and those with early severe physiology where aggressive resuscitation may be unavoidable.

Strengths & Limitations

  • Strengths: Pragmatic ED enrolment in the early window; clear weight-based dosing regimens; protocolised reassessment; clinically meaningful safety endpoint; multicentre multinational design; prespecified interim monitoring.
  • Limitations: Stopped early (underpowered for efficacy); open-label (risk of performance bias for discretionary outcomes); composite primary outcome may complicate mechanistic interpretation; exclusions limit generalisability to high-risk cardiopulmonary/renal populations.

Interpretation & Why It Matters

  • Clinical signal
    Across 249 randomised patients, an aggressive LR strategy did not reduce moderately severe/severe pancreatitis but substantially increased fluid overload, shifting the risk–benefit balance against routine “aggressive hydration”.
  • Practice implication
    A default approach of moderate maintenance (≈1.5 mL/kg/h) with boluses reserved for demonstrable hypovolaemia is better aligned with safety, especially on general wards where continuous haemodynamic monitoring is limited.
  • Mechanistic framing
    The harm signal is coherent with early large-volume exposure (0–12 h median 3.7 L vs 2.0 L) and downstream respiratory support/diuretic use, emphasising frequent reassessment and early de-escalation when oral intake returns or overload appears.

Controversies & Subsequent Evidence

  • Early stopping and inference: The DSMB-driven halt strengthens confidence in the safety signal but makes efficacy estimates imprecise; the absence of benefit cannot be interpreted as definitive equivalence.
  • Composite primary outcome interpretability: The primary endpoint aggregates different severity constructs (local complications, systemic complications, organ failure, comorbidity exacerbation), which can dilute mechanistic specificity even when clinically meaningful.
  • Open-label downstream decisions: Discretionary endpoints (hospital length of stay) could be influenced by clinician awareness of assignment; however, imaging use was similar and objective outcomes (fluid overload criteria, organ failure, death) anchor the main message.
  • Editorial perspective: Contemporary commentary emphasised that WATERFALL provides high-quality evidence against routine aggressive hydration and highlighted the practical message: bolus for hypovolaemia only, and reassess frequently to avoid pulmonary congestion.2
  • Meta-analytic synthesis: Post-WATERFALL meta-analyses focusing on RCT evidence generally report no improvement in clinically important outcomes with aggressive strategies but increased fluid overload/respiratory complications, reinforcing the trial’s harm signal.34
  • Guideline evolution: Newer guidance increasingly favours more conservative, goal-directed fluid with lactated Ringer’s and explicit avoidance of indiscriminate “aggressive” resuscitation, reflecting incorporation of WATERFALL’s safety findings into recommendations.56

Summary

  • ED-enrolled, multinational RCT comparing aggressive vs moderate lactated Ringer’s resuscitation in early acute pancreatitis.
  • Stopped early after first interim analysis (249 randomised) due to excess fluid overload without benefit signal.
  • Primary outcome not improved: moderately severe/severe pancreatitis 22.1% vs 17.3% (Adjusted RR 1.30; 95% CI 0.78 to 2.18; P=0.32).
  • Harm increased: fluid overload 20.5% vs 6.3% (Adjusted RR 2.85; 95% CI 1.36 to 5.94; P=0.004).
  • Best-supported practice message: moderate maintenance with bolus only for hypovolaemia, coupled with frequent reassessment and early de-escalation.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • The WATERFALL regimen is “aggressive” in a very specific way (weight-based bolus + high infusion rate) and was tested in early ED-presenting patients without baseline organ failure; extrapolation beyond this phenotype should be cautious.

Overall Takeaway

WATERFALL is a landmark because it replaced decades of physiologic intuition and weak evidence about “early aggressive hydration” in acute pancreatitis with a pragmatic, protocolised randomised comparison that detected a clinically important harm signal. In early ED-presenting pancreatitis without baseline organ failure, aggressive LR resuscitation did not reduce severity but substantially increased fluid overload, supporting a reassessment-driven, moderate strategy as the safer default.

Overall Summary

  • Aggressive weight-based fluid resuscitation increased fluid overload (20.5% vs 6.3%) without improving clinically important severity outcomes.
  • Early, repeated clinical reassessment and de-escalation are central: bolus only for hypovolaemia, avoid automatic high-rate infusions.
  • Generalises best to early ED/ward pancreatitis without baseline organ failure; apply cautiously to ICU-level resuscitation phenotypes.

Bibliography

  • 1Bolado F, de-Madaria E, López-Font I, et al. Aggressive or moderate fluid resuscitation in acute pancreatitis (WATERFALL): design and rationale of a randomised controlled trial. Front Med (Lausanne). 2020;7:440. DOI
  • 2Gardner TB. Fluid resuscitation in acute pancreatitis — going over the WATERFALL. N Engl J Med. 2022;387(11):1038-1039. DOI
  • 3Li XW, Liu X, Li S, et al. Comparison of clinical outcomes between aggressive and non-aggressive intravenous hydration for acute pancreatitis: a systematic review and meta-analysis. Crit Care. 2023;27:122. DOI
  • 4He K, He Y, Qian X, et al. Aggressive versus controlled fluid resuscitation in acute pancreatitis: a systematic review and meta-analysis. Chin Med J (Engl). 2023;136(10):1166-1173. DOI
  • 5Tenner S, Vege SS, Abou Saleh M, et al. American College of Gastroenterology guidelines: management of acute pancreatitis. Am J Gastroenterol. 2024;119(7):1048-1076. DOI
  • 6Schepers NJ, Besselink MG, Bruno MJ, et al. International Association of Pancreatology guideline: evidence-based recommendations for acute pancreatitis. Pancreatology. 2025;25(6):e1-e32. DOI