Evolution of early fluid therapy in acute pancreatitis: Balancing perfusion needs with fluid toxicity

Abstract

Hemoconcentration, vomiting, and third-space losses make early fluid administration in acute pancreatitis (AP) seem almost reflexive. The first few hours often justify this approach. What changes faster than most ward rounds capture, however, is the physiological target itself. By the time a severe case is re-examined, capillary leak and endothelial barrier injury may already be redirecting that same crystalloid into the interstitium. The first 24 hours therefore carry two opposing imperatives: Correcting hypoperfusion and recognizing the moment at which infusion becomes injury. The WATERFALL trial reframed this trade-off by counting fluid overload as a treatment effect rather than a late mishap of usual care. Aggressive hydration produced more harm without measurable clinical gain. The trial nonetheless left a familiar bedside scenario unanswered: The patient whose hematocrit refuses to fall while a new oxygen requirement appears overnight, or whose abdomen tenses while urine output drops. Such patients were too few in the trial to define a rescue rule. Endothelial injury and glycocalyx loss explain why crystalloid escapes the vessel; for the clinician, what matters is the point at which continued infusion has stopped helping. We frame this review around fluid tolerance rather than total volume. Lactated Ringer’s remains a defensible first crystalloid, though its edge over saline is clearer for inflammatory and intermediate endpoints than for harder outcomes. The bedside practice we advocate is brief, repeated reassessment—typically near the 2-hour and 6-hour marks—drawing on blood urea nitrogen trajectory, urine output, oxygenation, abdominal status, and cardiorenal reserve as a combined picture. Infusion should be slowed, paused, or redirected when tolerance markers worsen. The margin is narrowest in phenotypes that constrain the patient before AP does: Obesity, limited respiratory reserve, and prior cardiac or renal disease.

Key Words: Acute pancreatitis; Fluid therapy; Dextran; Ringer’s lactate; Randomized controlled trial; Systemic inflammatory response syndrome; C-reactive protein

Core Tip: In early acute pancreatitis, fluid is usually started for the right reason—and continued for the wrong one. The same admission order may persist while the patient underneath it has changed. A patient with persistent hemoconcentration or oliguria may still appear under-resuscitated, but a new oxygen requirement, basal crackles, or a tense abdomen should interrupt the next bolus, not justify it. The bedside question shifts from “Is perfusion adequate?” to “Can this patient still tolerate volume?”—and the answer rarely comes from a single number. Our argument is simple: Reassess early, adjust by tolerance, and resist the inertia of the admission prescription.

INTRODUCTION

Acute pancreatitis (AP) has been diagnosed more often over the past decade, paralleling the rise of metabolic disease and the persistence of gallstone- and alcohol-related disease as principal triggers[1,2]. Most cases settle on conservative care. A meaningful minority does not—and the deterioration may begin within the first 12 hours, before any severity score can be trusted. In that window, physiology runs ahead of the rhythm of routine ward review.

Guidelines, unsurprisingly, place hemodynamic stability and fluid therapy at the center of early care[3-5]. What makes those early decisions unusual is their downstream weight: Escalation to higher-level care, mechanical ventilation, abdominal complications, and length of stay all bend on what is decided in the first few hours[6,7]. An admission infusion that no one revisits is not neutral. When harm accumulates here, it accumulates quickly.

The harder question is rarely whether to start fluid, since early supportive care and fluid resuscitation remain central components of AP management[8-10]. It is when to stop assuming the answer is still “yes”. Infusion rate, crystalloid type, and monitoring endpoints have each been studied in isolation, but the ward patient does not present these as separate problems. Volume depletion at hour 2 may have given way to capillary leak by hour 8—and at that point, each additional liter is increasingly diverted to the interstitium. The evidence base thins precisely here, where the clinician must judge whether an existing prescription is still serving the patient.

This review therefore avoids recommending a fixed volume target. We ask instead a recursive question: After each reassessment, does another increment of fluid still buy effective perfusion without producing clinically meaningful intolerance? The remainder of the review reads the literature through that lens. The overall evolution of early fluid-resuscitation practice in AP, from routine aggressive hydration toward monitored moderate and individualized strategies, is summarized in Figure 1. Recent guidelines and pivotal studies shaping the current approach to early fluid resuscitation in AP are summarized in Table 1.

Figure 1 Evolution of fluid-resuscitation strategies in acute pancreatitis. LR: Lactated Ringer’s solution.

Table 1 Recent guidelines and pivotal turning points shaping early fluid resuscitation in acute pancreatitis.

Ref.	Year	Study type/scope	Main message	Comment
Párniczky et al[3]	2025	International guideline	Moderate monitored hydration	Guideline
Tenner et al[4]	2024	ACG guideline	Early fluid therapy remains central	Guideline
de-Madaria et al[11]	2022	Multicenter RCT	Overload without clear benefit	Pivotal

ACG: American College of Gastroenterology; RCT: Randomized controlled trial.

WHAT HAS THE FIELD LEARNED FROM THE FAILURE OF AGGRESSIVE HYDRATION?

The case for aggressive hydration was never opaque. Hemoconcentration, pancreatic ischemia, intravascular depletion were the textbook drivers of early injury, and generous infusion looked like the most direct lever against them. That logic held in some patients. It held less reliably in those whose pancreatitis was severe enough to matter most: Vascular leak and fluid intolerance can develop in hours, and the same crystalloid that stabilized the circulation early can later sit in the wrong compartment, feeding edema rather than perfusion[11-13].

The WATERFALL trial made this harm signal harder to dismiss. In that trial, aggressive hydration led to more fluid-overload events and did not provide a clear clinical advantage[11]. Its message is therefore broader than a comparison of two infusion rates. Fluid toxicity needs to be treated as a consequence of the prescription itself and recorded early, before overload evolves into obvious respiratory or abdominal failure[14]. After this trial, routine aggressive hydration has little place as a default order.

WATERFALL also needs reading inside its enrollment limits. The trial does not speak to the rescue strategy for severe AP with established organ failure, marked hypovolemia, or rapidly evolving shock—populations in which under-resuscitation remains genuinely dangerous. The lesson is therefore not to make every patient deliberately “dry”, but it is also not to treat the trial's harm signal as confined to the tail of the severity spectrum. Each bolus, each rate change, has to earn its place against the same recurring question: Is the next increment still doing what was intended?

The concern about harm did not begin with WATERFALL. Earlier aggressive protocols had already been associated with increased need for respiratory support, without improvement in clinical outcomes[15-17]. Later observational and randomized studies linked excess fluid to pulmonary edema, abdominal complications, and prolonged hospitalization, particularly after capillary permeability increases[18]. Meta-analyses have pointed consistently in the same direction[19-21].

The literature is uneven—patient selection, severity profile, baseline volume status, and endpoint definitions vary enough across trials[22] to make it hard to pinpoint exactly which patients drive the harm signal. The conclusion that survives this unevenness is narrower than enthusiasts want and firmer than sceptics admit: Aggressive hydration cannot be defended as a default, but it remains a legitimate rescue option when hypovolemia is documented and perfusion is failing. The older intuition—that bigger early volumes are inherently safer—has run out of evidence to stand on.

None of this turns AP into a low-fluid disease. The opening move should still be moderate Lactated Ringer's solution (LR)-based resuscitation; boluses still earn their place when hypoperfusion is concrete, not assumed. What changes is the threshold for continuing. A patient whose pressure, urine output, blood urea nitrogen (BUN), and hematocrit move in the right direction—without their lungs, abdomen, or venous system pushing back—can be carried forward cautiously. The moment any of those second-line signals turns, the order needs revision, not renewal. Fluid plans that survive the patient's physiology are the ones that get rewritten with it.

WHICH FLUID, WHICH RATE, AND WHICH ENDPOINT SHOULD GUIDE RESUSCITATION TODAY?

With aggressive hydration no longer the default, the active questions move one step closer to the bedside: Which crystalloid, at what rate, and judged by what response. The randomized trial by Lee et al[23], reinforced by earlier trial-level evidence and subsequent meta-analyses, has generally favored LR over normal saline for inflammatory and selected clinical endpoints[24-26]. Representative contemporary studies informing fluid selection, reassessment, and precision-oriented practice are summarized in Table 2. The biological argument is consistent: Balanced crystalloid avoids hyperchloremic acidosis and is less likely to amplify the inflammatory response.

Table 2 Representative contemporary studies informing fluid selection, reassessment, and precision-oriented practice in acute pancreatitis.

Ref.	Year	Study design/population	Main findings	Comment
Li et al[15]	2020	Severe hemoconcentrated AP	More noninvasive ventilation	Pulmonary risk
Lee et al[23]	2021	Randomized trial; LR vs NS	LR favored	Key trial
Chen et al[27]	2022	Updated meta-analysis; LR vs NS	LR superior to NS	Supports LR
Froghi et al[50]	2022	Feasibility RCT of ward-based GDFT	Ward-based reassessment feasible	Practical
Ehsan et al[17]	2024	Updated meta-analysis	No reproducible net benefit	Restrictive
Hong et al[29]	2024	Updated meta-analysis; LR vs NS	Recent evidence favored LR	Consistent
Evans et al[51]	2024	Meta-analysis and trial sequential analysis	Routine aggressive therapy unsupported	Sequential
Kong et al[41]	2025	Machine learning-guided fluid study	ML-guided strategy improved outcomes	Frontier

AP: Acute pancreatitis; GDFT: Goal-directed fluid therapy; LR: Lactated Ringer’s solution; ML: Machine learning; NS: Normal saline; RCT: Randomized controlled trial.

Later syntheses restricted to randomized evidence have reported broadly compatible findings[27-31], though not all updates have demonstrated consistent benefit across every endpoint. Differences in trial design, sample size, timing, disease severity, and outcome selection account for much of this variation. The resulting literature supports a practical preference without justifying overstatement. Representative randomized trials and meta-analyses comparing LR solution with normal saline in AP are summarized in Table 3.

Table 3 Representative studies comparing Lactated Ringer’s solution with normal saline in acute pancreatitis.

Ref.	Year	Study type/population	Main findings	Comment
Lee et al[23]	2021	Randomized trial; mild AP	LR favored	Key trial
Zhou et al[56]	2021	Meta-analysis of RCTs	Pooled signal favored LR	Synthesis
Kow et al[25]	2022	Systematic review and meta-analysis	LR remained favorable	Consistent
Vedantam et al[28]	2022	Updated meta-analysis	Benefit varied by endpoint	Mixed
Hong et al[29]	2024	Updated meta-analysis	Recent pooled data favored LR	Supports LR
Guilabert et al[33]	2024	Multicenter RCT protocol	Prospective testing ongoing	Ongoing

AP: Acute pancreatitis; LR: Lactated Ringer’s solution; RCT: Randomized controlled trial.

In clinical terms, LR should remain the usual first crystalloid for most patients with AP[32,33]. LR provides a safer starting platform than normal saline across much of the available literature, but crystalloid composition cannot compensate for an inappropriate infusion rate, an unattended order set, or failure to recognize early fluid intolerance.

The ongoing WATERLAND program is important for a different reason: The remaining questions are mainly bedside questions. LR is already a reasonable default, but it is less clear whether its advantages are maintained across severity strata, organ vulnerability, and different time windows. It is also unclear how crystalloid choice should interact with bolus use, rate reduction, withholding of fluid, and stopping criteria. For this reason, the most useful data will not be the assigned fluid alone, but also the timing and clinical rationale for dose changes after reassessment.

Take two patients with identical numbers-falling urine output, rising BUN. The decision diverges almost entirely on what is happening above the diaphragm and below it. Clear lungs and stable oxygenation make a bolus defensible; new B-lines, increasing oxygen requirement, or a tense abdomen turn the same bolus into a wager the patient may not survive.

Figure 2 summarizes this central biologic tension: Too little effective volume sustains microcirculatory ischemia, whereas too much volume amplifies capillary-leak-mediated edema and organ dysfunction.

Figure 2 Under-resuscitation sustains microcirculatory ischemia, whereas over-resuscitation worsens edema and organ dysfunction; the therapeutic goal is to preserve effective perfusion while limiting capillary leak.

An admission infusion is a hypothesis, not a standing order—and like any hypothesis, it deserves frequent retesting. Pressure, urine output, mental state, breathing, BUN and hematocrit, and any sign of accumulating volume should be read together rather than in isolation, at intervals shorter than the ward round naturally allows. Bedside ultrasound—assessing the lungs, inferior vena cava, and signs of venous congestion—adds resolution in trained hands, but in most non-intensive care unit (ICU) settings, its standardization lags behind equipment and operator availability. Clinicians are encouraged to treat these tools as adjuncts to clinical reassessment, not substitutes.

WHICH PATIENTS REQUIRE A DIFFERENT STRATEGY, AND HOW SHOULD THE FIELD MOVE TOWARD PRECISION?

Tolerance to early crystalloid is not uniform across AP patients. In mild disease with clear volume depletion, crystalloid may restore effective circulating volume without causing immediate harm. The situation is different in evolving severe AP. As systemic inflammation progresses, endothelial barrier injury and glycocalyx loss may increase capillary permeability early in the disease course[34-36]. At that point, the volume written on the chart and the volume that actually remains useful for perfusion may no longer be the same.

The bedside consequence is uncomfortable. After several liters, the same patient may still have hemoconcentration and oliguria, yet also develop crackles, a tense abdomen, or rising venous pressures. If assessed narrowly—through BUN, hematocrit, and urine output alone—another bolus may appear defensible. However, when the same patient is evaluated with attention to the chest, abdomen, and venous system, that same bolus may be dangerous. Few clinical scenarios in AP are as exposing of single-axis thinking as this.

Experimental and translational studies provide a biologic explanation for this pattern. Endothelial injury and glycocalyx shedding make it easier for crystalloid to leave the vascular space[34-36]. Clinically, the important message is not that glycocalyx injury can be measured and used as a bedside target; it usually cannot. The more useful lesson is simpler: Administered volume should not be assumed to equal effective intravascular volume. Once capillary leak is established, additional crystalloid may worsen lung water, bowel edema, abdominal pressure, and organ dysfunction without producing a durable improvement in perfusion. The proposed mechanistic link between endothelial injury, glycocalyx disruption, capillary leak, reduced effective circulating volume, and fluid-related harm is shown in Figure 3.

Figure 3 Endothelial and glycocalyx injury as a mechanism linking inflammation, capillary leak, reduced effective circulating volume, and fluid-related harm in acute pancreatitis. LR: Lactated Ringer’s solution.

Severe AP is rarely one hemodynamic state for long. The same patient may need active resuscitation at hour 4 and active restraint by hour 12, which is why the practical question is never how much fluid has been given, but whether the next litre still has somewhere useful to go.

Hypertriglyceridemia-associated AP is one example of why phenotype matters. It is increasingly recognized as a distinct clinical entity, with specific metabolic drivers, recurrence patterns, and possibly different inflammatory or microcirculatory behavior[37-39]. Current evidence is not strong enough to justify a separate fluid protocol for these patients. However, it is strong enough to argue against blind use of a standard pathway without early reassessment.

Obesity, limited thoracic reserve, and prior cardiac or renal disease do not argue for withholding fluid; they argue for noticing trouble earlier, because in these patients the warning signs arrive late and reverse slowly. A modest rise in lung water matters more in someone with poor respiratory reserve. A modestly positive balance accumulates faster in someone with reduced cardiac or renal headroom. Admission scores rarely capture either constraint well.

Guidelines call for individualised fluid therapy, but on a real ward, “individualized” has to mean something portable: A structured second look, not an algorithm. For most patients, that means revisiting the same axes that should have prompted the original prescription—namely, whether perfusion is still moving in the right direction and whether the patient continues tolerating the volume that got him/her there? Unstable patients escalate to invasive monitoring or critical care. The rest need a more ordinary discipline: The admission order must be questioned out loud whenever the physiology shifts.

The term “precision fluid therapy” should therefore be used carefully. At present, precision does not mean that a biomarker or prediction model can dictate the infusion rate in real time. Recent studies have explored early fluid requirements, machine-learning-guided fluid strategies, and model-based severity prediction in AP[40-43]. Other studies have evaluated early severity prediction, respiratory failure risk, radiomics, and interpretable prediction models as tools for identifying patients at risk of deterioration[44-47]. These approaches are promising, but most remain retrospective, single-center, or insufficiently validated for immediate bedside decision-making. None has yet shown, in a prospective clinical setting, that model-guided adjustment improves patient-centered outcomes.

Recent studies have extended precision-oriented fluid management in AP beyond conventional clinical reassessment. Automated machine-learning models have been developed for early severity prediction in AP[48,49], and ward-based goal-directed fluid therapy has shown that structured reassessment outside the ICU is feasible[50]. Trial sequential analysis of randomized evidence has further questioned the clinical benefit of aggressive fluid therapy[51]. Imaging-based radiomics, lung injury indices, invasive hemodynamic monitoring, and computed tomography-based prediction of acute kidney injury have been investigated as approaches for identifying patients at risk of deterioration or organ failure[52-55]. Additional randomized-trial meta-analysis comparing LR solution with normal saline supports the use of crystalloid selection as part of a broader reassessment-based strategy[56].

Alongside crystalloid selection, serial BUN and hematocrit trends remain clinically accessible markers that may help identify prognostic risk or treatment response, but they should be interpreted as reassessment signals rather than isolated triggers for continued infusion[57,58].

Interpretability matters as much as accuracy. A score that fires without explaining whether the alarm came from the perfusion side or the tolerance side is hard to act on—especially overnight, when the patient may be oliguric and hypoxemic at the same time. In that situation, what the clinician needs is not a number but a label: Which signal moved, and what it asks the team to do next.

A workable sequence keeps three moves in order: (1) Start with moderate LR when early hypoperfusion is plausible, not assumed; (2) Reserve boluses for documented deficits and let the response—within 1-2 hours—judge whether to continue; and (3) Carry the infusion forward only while the perfusion axis improves and the tolerance axis holds; the moment oxygen requirement, abdominal pressure, positive balance, or venous congestion starts moving the wrong way, slow, pause, or reverse rather than persist. In some patients, the next step is no longer fluid at all but organ support or active deresuscitation.

Figure 4 presents the organ-level consequences of under-resuscitation and over-resuscitation, and summarizes a moderate, goal-directed approach to early fluid therapy in AP.

Figure 4 Organ-level consequences of under-resuscitation and over-resuscitation as well as a moderate, goal-directed approach to early fluid therapy in acute pancreatitis. MAP: Mean arterial pressure; UO: Urine output; BUN: Blood urea nitrogen; RESR: Respiratory tolerance.

The largest remaining gap is the lack of reliable stopping rules. Clinicians increasingly accept that indiscriminate aggressive hydration is unsafe, but they still lack a reproducible threshold for deciding when rescue therapy is becoming fluid toxicity. This uncertainty is most important in severe or evolving AP, where capillary leak, obesity, limited respiratory reserve, cardiorenal disease, and rising abdominal pressure can change the balance within hours.

The clinical conflict is familiar enough that most pancreatitis-active wards can name it. Hematocrit stays stubbornly high; oxygen requirement or abdominal girth keeps creeping up. One number says “still under-filled”. The other says “the next litre is going somewhere it should not”. Trial evidence speaks to almost everything around this moment, and almost nothing inside it. That gap, more than the volume question, is where useful trials should now go.

Future trials should test moderate LR-based strategies together with explicit rules for slowing, stopping, or reversing fluid therapy. The outcomes should be practical and clinically meaningful: Respiratory deterioration, need for organ support, abdominal complications, length of stay, and treatment-related harm. Biomarkers, phenotype stratification, and decision-support tools may be useful, but only if they clarify the stopping-point problem. They should not become separate agendas detached from bedside decisions.

Figure 5 brings these elements together by presenting fluid therapy as a changing balance between perfusion need and fluid tolerance. This is a narrow but useful version of precision medicine. Before relying on more complex biomarkers or algorithms, clinicians need better ways to recognize that additional crystalloid has stopped helping.

Figure 5 Dynamic balance between perfusion need and fluid tolerance in acute pancreatitis, linking vascular leak to both benefit and harm during resuscitation.

Table 4 highlights the same principle from the perspective of patient phenotype. In AP, precision fluid therapy should not mean automatic escalation through an algorithm. It should mean recognizing when fluid deficit is no longer the dominant danger and fluid intolerance has become more immediate. Moving this concept into practice will require prospective trials, standardized bedside endpoints, and decision tools that clinicians can understand and reproduce. Figure 6 provides a phenotype-aware bedside roadmap integrating moderate LR-based resuscitation, repeated reassessment, and recognition of fluid responsiveness or intolerance.

Figure 6 Phenotype-aware bedside roadmap for fluid therapy in acute pancreatitis, integrating moderate Lactated Ringer’s-based resuscitation, repeated reassessment, and recognition of fluid responsiveness or intolerance. LR: Lactated ringer’s solution; MAP: Mean arterial pressure; AP: Acute pancreatitis; AI: Artificial intelligence.

Table 4 High-risk clinical phenotypes and practical implications for fluid strategy in acute pancreatitis.

Phenotype	Why risk is higher	Practical implication	Ref.
Severe AP with evolving organ failure	Inflammation may uncouple fluid load from effective perfusion	Reassess frequently	[34,36,53]
Hypertriglyceridemia-associated AP	Distinct metabolic drivers may alter fluid tolerance	Avoid a generic algorithm	[37-39]
Obesity and reduced thoracic reserve	Positive balance may trigger respiratory compromise	Use smaller test boluses
Cardiorenal or unstable high-risk presentations	Fluid tolerance may be low and organ-support needs may evolve rapidly	Escalate support earlier	[54,55,57,58]

AP: Acute pancreatitis.

CONCLUSION

Treating fluid in AP as a procedure with a stop-criterion, rather than a target, is the harder cognitive move. Once endothelial injury, capillary leak, and third-spacing dominate, the next liter is no longer volume support; it is lung water, bowel edema, and rising intra-abdominal pressure. The point at which this conversion occurs is not visible on any single bedside variable, and clinicians know this long before the literature catches up.

Moderate LR-based resuscitation with early reassessment is the most defensible default available today. Defending a default, however, is not the same as solving the problem. What the field still owes the bedside is a usable rule for stopping—one that survives the patient who looks under-filled and fluid-intolerant in the same hour. Phenotype shapes that rule more than admission scores do: Obesity, limited respiratory reserve, and prior cardiac or renal disease narrow the window long before any prognostic tool reflects it.

What useful trials should now address is not larger vs smaller early volumes. It is the discipline of stopping: Explicit thresholds for slowing, pausing, and reversing fluid therapy, tested against outcomes that translate to the ward-respiratory deterioration, abdominal complications, the need for organ support, and length of stay. Biomarkers and decision-support models will earn their place only if they speak to that threshold question rather than running parallel to it. Until they do, the most consequential clinical skill remains a quiet one: Noticing, before the chart does, that the next liter has stopped helping.