Which fluid resuscitation strategy is associated with shorter length of stay in mild acute pancreatitis? Interaction between crystalloid type and administration rate Authors: Erhan Bozkurt, Osman Saglam, Alper Sarı, Murat Ay Published: 2026 ABSTRACT BackgroundWhile fluid resuscitation is the cornerstone of acute pancreatitis (AP) management, uncertainty remains regarding the optimal crystalloid selection and administration rate. This study assessed the independent and interactive associations of crystalloid type and resuscitation rate with clinical outcomes in mild AP.MethodsThis single-center, retrospective cohort study (2018–2024) included 204 patients hospitalized with mild AP. Patients were categorized by crystalloid type (Lactated Ringer’s (LR), normal saline (NS), Isolyte S, or dextrose-containing solutions) and resuscitation rate (aggressive, moderate, minimal, or low). The primary endpoint was length of stay (LOS). Multivariable linear regression and analyses stratified by Bedside Index for Severity in Acute Pancreatitis (BISAP) score were performed to assess main associations and interactions.ResultsLR and NS were associated with comparable hospital stays, with no significant difference in LOS (P = 0.24). Both were associated with significantly shorter LOS than Isolyte S and dextrose (p < 0.001). A significant interaction occurred between fluid type and resuscitation rate: LR and NS at moderate-to-aggressive rates yielded the shortest stays, whereas Isolyte S and dextrose at low rates yielded the longest. In multivariable analysis, fluid type, resuscitation rate, BISAP score, age, and hypertension were independent predictors of LOS (R² = 0.739). These associations remained consistent across severity subgroups.ConclusionsIn mild AP, LR and NS showed comparable outcomes, both of which were associated with shorter hospital stays than Isolyte S and dextrose. The significant interaction between fluid type and rate suggests that crystalloid composition and infusion rate may need to be considered jointly when optimizing resuscitation. NS appears to be a reasonable alternative when LR availability is limited.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12876-026-05016-9. Introduction Acute pancreatitis (AP) is a common gastrointestinal emergency ranging from mild, self-limiting inflammation to life-threatening disease with mortality rates approaching 30% [1, 2]. Early fluid resuscitation remains the cornerstone of management, maintaining hemodynamic stability and pancreatic perfusion while reducing the risk of organ failure and complications [3–5]. The optimal crystalloid choice remains debated. While normal saline (NS) has historically been standard, Lactated Ringer’s (LR) has emerged as a preferred alternative with mounting evidence of reduced progression to severe AP, fewer ICU admissions, and attenuated systemic inflammation [6–17]. Current IAP/APA and ACG guidelines recommend LR as the first-line resuscitation fluid, supported by several meta-analyses [18–26]. Despite this, substantial heterogeneity persists: acetate/gluconate-buffered solutions and dextrose-containing fluids remain widely used due to regional availability and institutional protocols, despite limited comparative data in AP [27, 28]. These fluids have distinct electrolyte compositions and metabolic effects that may differentially affect inflammatory responses and clinical outcomes. The optimal resuscitation rate is equally uncertain. The WATERFALL trial demonstrated that aggressive fluid administration did not reduce progression to severe AP compared to moderate resuscitation but significantly increased fluid overload [29]. Subsequent systematic reviews confirmed no benefit from aggressive strategies [20, 26–28]. Importantly, these studies evaluated resuscitation rate independently of fluid composition. Despite this evidence base, critical knowledge gaps persist [30]. First, Isolyte S and dextrose-containing solutions have not been systematically evaluated in AP [31, 32]. Second, no prior study has examined whether fluid type and resuscitation rate interact in their associations with clinical outcomes — an interaction that is biologically plausible, as optimal infusion rates may vary by crystalloid composition [22, 27, 30, 33]. Third, it remains unknown whether fluid-related associations vary according to AP severity category as defined by the revised Atlanta classification. We conducted this single-center retrospective cohort study to address these gaps. We compared four crystalloid solutions (LR, NS, Isolyte S, and dextrose) administered at four resuscitation rates (aggressive, moderate, minimal, and low) in patients with mild AP. We hypothesized that fluid composition and resuscitation rate would interact in their associations with length of stay (LOS), and that these associations would remain consistent across different BISAP-defined severity categories. Materials and methods Study design and patient selection This single-center retrospective cohort study was conducted at the Departments of Internal Medicine and Gastroenterology, Faculty of Medicine, Afyonkarahisar Health Sciences University. Patients were identified through a structured search of the institutional electronic medical records (EMR) using ICD-10 codes for acute pancreatitis. All retrieved records were independently reviewed by two investigators to confirm the diagnosis based on at least two of the following criteria: characteristic abdominal pain; serum amylase or lipase levels exceeding three times the upper limit of normal; or characteristic imaging findings on ultrasonography or contrast-enhanced computed tomography [2]. A total of 419 patients hospitalized for mild AP between March 2018 and February 2024 were initially screened. We included patients aged ≥ 18 years with mild AP [2]—defined by the absence of organ failure or local/systemic complications—who were admitted during the study period and had complete documentation regarding crystalloid type and resuscitation rate. The primary endpoint was LOS (in days). Exclusion criteria were moderate or severe AP at presentation, chronic pancreatitis, chronic kidney disease, congestive heart failure, active malignancy or chemotherapy, transfer from another facility within 24 h of admission (due to potential protocol differences), pregnancy, age < 18 years, hospital stay of less than 24 h, and sepsis. Patients with hospital stays shorter than 24 h were excluded because complete intravenous fluid administration records were unavailable for systematic analysis. This group predominantly comprised self-limiting presentations in which a full resuscitation protocol was not initiated, as well as patients transferred to other facilities due to capacity constraints; the precise clinical disposition of these excluded patients could not be fully ascertained from the available retrospective records (Fig. 1). Fig. 1Patient flow diagram. AP, acute pancreatitis; EMR, electronic medical record; IV, intravenous; LR, Lactated Ringer’s; NS, normal saline Patient flow diagram. AP, acute pancreatitis; EMR, electronic medical record; IV, intravenous; LR, Lactated Ringer’s; NS, normal saline Clinical variables and assessment of disease severity Demographic variables included age, sex, and comorbidities at presentation (diabetes mellitus, hypertension, hyperlipidemia, and alcohol use disorder). Disease severity was assessed using the Bedside Index for Severity in AP (BISAP), calculated within the first 24 h of admission. The BISAP score consists of five criteria (1 point each): blood urea nitrogen > 25 mg/dL, altered mental status (Glasgow Coma Scale < 15), systemic inflammatory response syndrome (≥ 2 of: heart rate > 90/min, respiratory rate > 20/min or PaCO₂ <32 mmHg, temperature < 36 °C or > 38 °C, white blood cell count < 4,000 or > 12,000/µL), age > 60 years, and pleural effusion on imaging [34]. Fluid type and resuscitation rate During the study period, no institution-wide mandatory protocol governed crystalloid selection or resuscitation rate in mild AP. Fluid choice and infusion rate were at the discretion of the attending physician, guided by clinical assessment, patient comorbidities, and evolving local practice. In all 204 patients included in the final analysis, a single crystalloid type was administered throughout the initial 48-hour resuscitation period; fluid type was determined at admission and was not changed during the resuscitation phase, reflecting a protocol-consistent practice in our institution. Classification was therefore unambiguous for every included patient, with no risk of misclassification bias related to mixed crystalloid exposure. This approach minimizes misclassification and reflects real-world resuscitation practice. Patients were divided into four groups based on the crystalloid administered: LR, NS, Isolyte S, and dextrose-containing solutions. Dextrose-containing solutions included 5% dextrose in water (D5W) and dextrose-saline combinations (5% dextrose in 0.9% NaCl or 0.45% NaCl); these were analyzed as a single category given the small aggregate sample size (n = 7) and their shared metabolic characteristic of providing exogenous glucose without an alkalinizing buffer. The resuscitation rate was calculated as the total crystalloid volume administered during the first 24 h divided by body weight (mL/kg/h). Two observers independently classified resuscitation strategies using objective numerical thresholds; rare disagreements (< 5%) were resolved by consensus with a third reviewer. Resuscitation categories were based on the WATERFALL trial protocols [29]: aggressive (20 mL/kg bolus + ≥ 3.0 mL/kg/h or ≥ 200 mL/h) and moderate (10 mL/kg bolus + 1.5–2.9 mL/kg/h or 100–199 mL/h). Because WATERFALL defined only these two levels, we established a priori categories for minimal (1.0–1.4 mL/kg/h or 75–99 mL/h) and low (< 1.0 mL/kg/h or < 75 mL/h) resuscitation to stratify patients below the moderate threshold into clinically relevant subgroups. Statistical analysis Continuous variables are expressed as mean ± standard deviation, and categorical variables as frequencies and percentages. Normality was assessed using the Kolmogorov–Smirnov test. For non-normally distributed variables, comparisons between groups were performed using the Kruskal–Wallis test, followed by Dunn’s post hoc test for pairwise comparisons. We performed a multivariable linear regression analysis to identify independent predictors of LOS. Fluid type was modeled as an ordinal variable (1 = LR, 2 = NS, 3 = Isolyte S, 4 = dextrose). Model A included fluid type, resuscitation rate, and BISAP score. Model B added age, sex, diabetes mellitus, hypertension, hyperlipidemia, and alcohol use disorder. To assess the pre-specified hypothesis that the optimal resuscitation rate may vary by crystalloid composition, an interaction term (fluid type × resuscitation rate) was included in the model. A multiplicative interaction term in multivariable linear regression is the standard approach for continuous outcomes. As a sensitivity analysis, fluid type was additionally modeled as a categorical variable using dummy coding with LR as the reference category and Aggressive rate as the reference, to confirm that findings were not dependent on the equal-spacing assumption of the ordinal model. Stratified analyses were conducted based on BISAP risk categories (low-risk: BISAP 0–1; high-risk: BISAP 2–3). A p-value of < 0.05 was considered statistically significant. All analyses were performed using IBM SPSS Statistics version 25.0 (IBM Corp., Armonk, NY, USA). Results Study population and descriptive statistics Of the 419 patients screened for mild AP, 215 were excluded. Reasons for exclusion included moderate-to-severe AP at presentation (n = 16), chronic pancreatitis (n = 4), chronic kidney disease (n = 14), congestive heart failure (n = 19), active malignancy or chemotherapy (n = 7), transfer from another facility within 24 h (n = 36), sepsis (n = 12), incomplete data or hospital stay < 24 h (n = 102), and pregnancy (n = 5). The final analysis included 204 patients. Table 1 shows the demographic and clinical characteristics. The cohort was sex-balanced, and the majority of patients had at least one comorbidity (most commonly hypertension, diabetes mellitus, hyperlipidemia, and alcohol use disorder). Disease severity was mild, with most patients classified as low-risk according to the BISAP score (Table 1). Table 1Baseline characteristics of the study populationCharacteristicOverall (N = 204)Age (years), mean ± SD56.4 ± 20.6Male sex, n (%)109 (53.4)BISAP score, mean ± SD0.98 ± 0.93BISAP 0, n (%)72 (35.3)BISAP 1, n (%)83 (40.7)BISAP 2, n (%)31 (15.2)BISAP 3, n (%)18 (8.8)Diabetes mellitus, n (%)68 (33.3)Hyperlipidemia, n (%)50 (24.5)Hypertension, n (%)96 (47.1)Alcohol use disorder, n (%)22 (10.8)Hospital LOS (days), mean ± SD6.62 ± 2.63Prolonged stay (> 7 days), n (%)65 (31.9) Table 1Baseline characteristics of the study populationCharacteristicOverall (N = 204)Age (years), mean ± SD56.4 ± 20.6Male sex, n (%)109 (53.4)BISAP score, mean ± SD0.98 ± 0.93BISAP 0, n (%)72 (35.3)BISAP 1, n (%)83 (40.7)BISAP 2, n (%)31 (15.2)BISAP 3, n (%)18 (8.8)Diabetes mellitus, n (%)68 (33.3)Hyperlipidemia, n (%)50 (24.5)Hypertension, n (%)96 (47.1)Alcohol use disorder, n (%)22 (10.8)Hospital LOS (days), mean ± SD6.62 ± 2.63Prolonged stay (> 7 days), n (%)65 (31.9) Baseline characteristics of the study population Comparison of clinical outcomes based on fluid type and resuscitation rate Age distribution did not differ by fluid type (p = 0.51). However, patients receiving Isolyte S or dextrose-containing solutions had significantly longer hospital stays (9.05 ± 2.56 and 9.43 ± 2.23 days, respectively) and higher BISAP scores compared to those receiving LR (6.04 ± 2.15 days) or NS (6.43 ± 2.71 days) (p < 0.001 for both). Outcomes were comparable between the LR and NS groups, with no significant differences in LOS or BISAP scores (p > 0.05) (Table 2). Table 2Comparison of clinical outcomes by fluid typeFluid Type n Agemean ± SDLOS (days) mean ± SDBISAPScoreLactated Ringer’s9454.39 ± 21.036.04 ± 2.150.86 ± 0.91Normal Saline8256.50 ± 20.696.43 ± 2.710.88 ± 0.89Isolyte S2167.24 ± 15.079.05 ± 2.561.71 ± 0.90Dextrose749.86 ± 20.009.43 ± 2.231.43 ± 0.53p value—0.51*< 0.001*< 0.001*< 0.001a< 0.001a0.002b0.047b< 0.001c< 0.001c0.004d*: Kruskal–Wallis test, pairwise comparisons: a: Lactated Ringer’s solution vs. Isolyte S, p < 0.05; b: Lactated Ringer’s vs. dextrose, p < 0.05; c: normal saline vs. Isolyte S, p < 0.05; d: normal saline vs. dextrose, p < 0.05 Table 2Comparison of clinical outcomes by fluid typeFluid Type n Agemean ± SDLOS (days) mean ± SDBISAPScoreLactated Ringer’s9454.39 ± 21.036.04 ± 2.150.86 ± 0.91Normal Saline8256.50 ± 20.696.43 ± 2.710.88 ± 0.89Isolyte S2167.24 ± 15.079.05 ± 2.561.71 ± 0.90Dextrose749.86 ± 20.009.43 ± 2.231.43 ± 0.53p value—0.51*< 0.001*< 0.001*< 0.001a< 0.001a0.002b0.047b< 0.001c< 0.001c0.004d*: Kruskal–Wallis test, pairwise comparisons: a: Lactated Ringer’s solution vs. Isolyte S, p < 0.05; b: Lactated Ringer’s vs. dextrose, p < 0.05; c: normal saline vs. Isolyte S, p < 0.05; d: normal saline vs. dextrose, p < 0.05 Comparison of clinical outcomes by fluid type *: Kruskal–Wallis test, pairwise comparisons: a: Lactated Ringer’s solution vs. Isolyte S, p < 0.05; b: Lactated Ringer’s vs. dextrose, p < 0.05; c: normal saline vs. Isolyte S, p < 0.05; d: normal saline vs. dextrose, p < 0.05 In the LR cohort, higher resuscitation rates were associated with shorter lengths of stay (p < 0.001). Patients receiving lower hydration rates were older, had longer hospital stays, and had higher BISAP scores (p < 0.001) (Table 3). Similar patterns were observed in the NS group: low-rate administration correlated with prolonged hospitalization and worse severity markers (p < 0.001), while aggressive or moderate rates were associated with earlier discharge (Table 4). Table 3Clinical outcomes by resuscitation rate in the Lactated Ringer’s groupFluid Type n Agemean ± SDLOS (days) mean ± SDBISAP ScoreAggressive3648.61 ± 23.405.06 ± 2.070.67 ± 0.86Moderate2949.17 ± 17.646.00 ± 1.930.79 ± 0.98Minimal1560.13 ± 15.056.93 ± 1.581.07 ± 0.80Low1473.93 ± 13.217.71 ± 2.091.29 ± 0.91p value—< 0.001*0.06*0.001*0.001a< 0.001b< 0.001b< 0.001d0.010 d*: Kruskal–Wallis test, pairwise comparisons: a: aggressive vs. minimal, p < 0.05; b: aggressive vs. low, p < 0.05; c: moderate vs. minimal, p < 0.05; d: moderate vs. low, p < 0.05 Table 3Clinical outcomes by resuscitation rate in the Lactated Ringer’s groupFluid Type n Agemean ± SDLOS (days) mean ± SDBISAP ScoreAggressive3648.61 ± 23.405.06 ± 2.070.67 ± 0.86Moderate2949.17 ± 17.646.00 ± 1.930.79 ± 0.98Minimal1560.13 ± 15.056.93 ± 1.581.07 ± 0.80Low1473.93 ± 13.217.71 ± 2.091.29 ± 0.91p value—< 0.001*0.06*0.001*0.001a< 0.001b< 0.001b< 0.001d0.010 d*: Kruskal–Wallis test, pairwise comparisons: a: aggressive vs. minimal, p < 0.05; b: aggressive vs. low, p < 0.05; c: moderate vs. minimal, p < 0.05; d: moderate vs. low, p < 0.05 Clinical outcomes by resuscitation rate in the Lactated Ringer’s group *: Kruskal–Wallis test, pairwise comparisons: a: aggressive vs. minimal, p < 0.05; b: aggressive vs. low, p < 0.05; c: moderate vs. minimal, p < 0.05; d: moderate vs. low, p < 0.05 Table 4Comparison of clinical outcomes based on normal saline infusion rateFluid Type n Agemean ± SDLOS (days)mean ± SDBISAP ScoreAggressive838.50 ± 24.604.38 ± 2.070.38 ± 0.52Moderate3347.73 ± 14.854.94 ± 1.750.61 ± 0.70Minimal966.22 ± 13.277.00 ± 1.941.22 ± 1.09Low3267.31 ± 19.888.31 ± 2.621.19 ± 0.97p value—0.001*< 0.001*0.020*0.007a0.012a0.023b0.001b< 0.001b0.012d0.012c0.015c< 0.001d< 0.001d*: Kruskal–Wallis test, pairwise comparisons: a: aggressive vs. minimal, p < 0.05; b: aggressive vs. low, p < 0.05; c: moderate vs. minimal, p < 0.05; d: moderate vs. low, p < 0.05 Table 4Comparison of clinical outcomes based on normal saline infusion rateFluid Type n Agemean ± SDLOS (days)mean ± SDBISAP ScoreAggressive838.50 ± 24.604.38 ± 2.070.38 ± 0.52Moderate3347.73 ± 14.854.94 ± 1.750.61 ± 0.70Minimal966.22 ± 13.277.00 ± 1.941.22 ± 1.09Low3267.31 ± 19.888.31 ± 2.621.19 ± 0.97p value—0.001*< 0.001*0.020*0.007a0.012a0.023b0.001b< 0.001b0.012d0.012c0.015c< 0.001d< 0.001d*: Kruskal–Wallis test, pairwise comparisons: a: aggressive vs. minimal, p < 0.05; b: aggressive vs. low, p < 0.05; c: moderate vs. minimal, p < 0.05; d: moderate vs. low, p < 0.05 Comparison of clinical outcomes based on normal saline infusion rate *: Kruskal–Wallis test, pairwise comparisons: a: aggressive vs. minimal, p < 0.05; b: aggressive vs. low, p < 0.05; c: moderate vs. minimal, p < 0.05; d: moderate vs. low, p < 0.05 Multivariable regression analysis In the multivariable linear regression analysis (Model A), fluid type, resuscitation rate, and BISAP score independently predicted the LOS (R² = 0.615; p < 0.001). Each increase in fluid category (LR→NS→Isolyte S→dextrose) was associated with 0.537 additional days (β = +0.537; 95% confidence interval (CI): 0.28–0.79; p < 0.001). Higher resuscitation rates predicted a shorter LOS (β = −0.634 per 1 mL/kg/h increase; 95% CI: −0.85 to − 0.42; p < 0.001). Every 1-point increase in the BISAP score was associated with 1.685 additional days (β = +1.685; 95% CI: 1.41–1.96; p < 0.001) (Table 5). Table 5Multivariable linear regression models for length of stayVariableβ Coefficient95% CIP valueModel AFluid type (per category increase)+ 0.5370.28 to 0.79< 0.001Resuscitation rate (per category increase)−0.634−0.85 to − 0.42< 0.001BISAP score (per point)+ 1.6851.41 to 1.96< 0.001R² = 0.615, p < 0.001Model BFluid type (per category increase)+ 0.5760.35 to 0.80< 0.001Resuscitation rate (per category increase)−0.292−0.49 to − 0.090.004BISAP score (per point)+ 0.8300.53 to 1.13< 0.001Age (per year)+ 0.0470.03 to 0.06< 0.001Hypertension+ 1.1860.66 to 1.71< 0.001Male sex+ 0.292−0.26 to 0.840.302Diabetes mellitus−0.339−0.96 to 0.280.283Hyperlipidemia+ 0.419−0.17 to 1.010.163Alcohol use disorder+ 0.749−0.03 to 1.530.060R² = 0.739, p < 0.001Fluid types: 1 = Lactated Ringer’s, 2 = normal saline, 3 = Isolyte S, 4 = dextrose. Resuscitation rates: 1 = aggressive, 2 = moderate, 3 = minimal, 4 = low. CI, confidence interval Table 5Multivariable linear regression models for length of stayVariableβ Coefficient95% CIP valueModel AFluid type (per category increase)+ 0.5370.28 to 0.79< 0.001Resuscitation rate (per category increase)−0.634−0.85 to − 0.42< 0.001BISAP score (per point)+ 1.6851.41 to 1.96< 0.001R² = 0.615, p < 0.001Model BFluid type (per category increase)+ 0.5760.35 to 0.80< 0.001Resuscitation rate (per category increase)−0.292−0.49 to − 0.090.004BISAP score (per point)+ 0.8300.53 to 1.13< 0.001Age (per year)+ 0.0470.03 to 0.06< 0.001Hypertension+ 1.1860.66 to 1.71< 0.001Male sex+ 0.292−0.26 to 0.840.302Diabetes mellitus−0.339−0.96 to 0.280.283Hyperlipidemia+ 0.419−0.17 to 1.010.163Alcohol use disorder+ 0.749−0.03 to 1.530.060R² = 0.739, p < 0.001Fluid types: 1 = Lactated Ringer’s, 2 = normal saline, 3 = Isolyte S, 4 = dextrose. Resuscitation rates: 1 = aggressive, 2 = moderate, 3 = minimal, 4 = low. CI, confidence interval Multivariable linear regression models for length of stay Fluid types: 1 = Lactated Ringer’s, 2 = normal saline, 3 = Isolyte S, 4 = dextrose. Resuscitation rates: 1 = aggressive, 2 = moderate, 3 = minimal, 4 = low. CI, confidence interval Model B (R² = 0.739; p < 0.001) included age, sex, and comorbidities. Fluid type (β = +0.576; 95% CI: 0.35–0.80; p < 0.001), resuscitation rate (β = −0.292; 95% CI: −0.49 to − 0.09; p = 0.004), BISAP score (β = +0.830; 95% CI: 0.53–1.13; p < 0.001), age (β = +0.047; 95% CI: 0.03–0.06; p < 0.001), and hypertension (β = +1.186; 95% CI: 0.66–1.71; p < 0.001) remained independent predictors. Sex, diabetes mellitus, hyperlipidemia, and alcohol use disorder were not significant (p > 0.05) (Table 5). As a sensitivity analysis, fluid type was modeled as a categorical variable using dummy coding with LR as the reference category and Aggressive resuscitation rate as the reference. The dummy-coded model yielded a marginally higher R² (0.765 vs. 0.739 in Model B), confirming that the ordinal assumption introduced minimal distortion. Isolyte S remained independently associated with longer LOS compared with LR (β = +1.18; 95% CI: 0.50–1.86; p = 0.001), and low-rate resuscitation was associated with longer stays compared with aggressive rate (β = +1.32; 95% CI: 0.70–1.93; p < 0.001). BISAP score (β = +0.73; 95% CI: 0.44–1.02; p < 0.001), age (β = +0.045; 95% CI: 0.029–0.061; p < 0.001), and hypertension (β = +1.26; 95% CI: 0.74–1.78; p < 0.001) remained fully consistent with the primary model. Interaction between fluid type and resuscitation rate We found a statistically significant interaction between fluid type and resuscitation rate (β = −0.287; 95% CI: −0.47 to − 0.10; p = 0.003; Table 6). LOS was shortest with LR or NS administered at moderate-to-aggressive rates (NS + Aggressive: 4.38 ± 2.07 days; NS + Moderate: 4.94 ± 1.75 days; LR + Aggressive: 5.06 ± 2.07 days). In contrast, Isolyte S at low rates resulted in the longest stays (11.50 ± 2.51 days), with 100% of patients experiencing prolonged hospitalization (> 7 days). This interaction suggests that the optimal resuscitation rate may vary by fluid composition. Table 6Analysis of the interaction between fluid type and resuscitation rateFluid × Rate Combination n LOS (days), mean ± SDProlonged Stay (> 7 days), n (%)Best Combinations (Shortest Hospital Stay)NS + Aggressive84.38 ± 2.071 (12.5)NS + Moderate334.94 ± 1.754 (12.1)LR + Aggressive365.06 ± 2.075 (13.9)LR + Moderate296.00 ± 1.933 (10.3)Worst Combinations (Longest Hospital Stay)Isolyte S + Low611.50 ± 2.516 (100.0)Isolyte S + Minimal58.60 ± 2.974 (80.0)NS + Low328.31 ± 2.6217 (53.1)LR + Low147.71 ± 2.096 (42.9)Interaction term (fluid type × resuscitation rate): β = −0.287; 95% CI: −0.47 to − 0.10; p = 0.003. LR Lactated Ringer’s, NS normal saline Table 6Analysis of the interaction between fluid type and resuscitation rateFluid × Rate Combination n LOS (days), mean ± SDProlonged Stay (> 7 days), n (%)Best Combinations (Shortest Hospital Stay)NS + Aggressive84.38 ± 2.071 (12.5)NS + Moderate334.94 ± 1.754 (12.1)LR + Aggressive365.06 ± 2.075 (13.9)LR + Moderate296.00 ± 1.933 (10.3)Worst Combinations (Longest Hospital Stay)Isolyte S + Low611.50 ± 2.516 (100.0)Isolyte S + Minimal58.60 ± 2.974 (80.0)NS + Low328.31 ± 2.6217 (53.1)LR + Low147.71 ± 2.096 (42.9)Interaction term (fluid type × resuscitation rate): β = −0.287; 95% CI: −0.47 to − 0.10; p = 0.003. LR Lactated Ringer’s, NS normal saline Analysis of the interaction between fluid type and resuscitation rate Interaction term (fluid type × resuscitation rate): β = −0.287; 95% CI: −0.47 to − 0.10; p = 0.003. LR Lactated Ringer’s, NS normal saline Analyses stratified by BISAP score Analyses stratified by BISAP score assessed whether fluid type associations varied by BISAP-defined severity category (Table 7). Given the small sample sizes of the Isolyte S (n = 21) and dextrose (n = 7) groups and the significant baseline imbalances relative to LR and NS, findings related to these fluids should be considered exploratory and hypothesis-generating, requiring confirmation in adequately powered prospective studies. Among low-risk patients (BISAP 0–1, n = 155), LR (5.41 ± 1.69 days, n = 75) and NS (5.59 ± 2.14 days, n = 64) demonstrated significantly shorter lengths of stay than Isolyte S (7.50 ± 1.51 days, n = 12) and dextrose (9.75 ± 1.26 days, n = 4) (p < 0.001 for all comparisons). Only 9.3% of low-risk patients treated with LR had prolonged hospitalizations (> 7 days), compared with 18.8% for NS, 50% for Isolyte S, and 100% for dextrose. Table 7Analysis stratified by BISAP risk categoryFluid Type n LOS (days), mean ± SDProlonged Stay (> 7 days), n (%)P valueLow-Risk Patients (BISAP 0–1), n = 155Lactated Ringer’s755.41 ± 1.697 (9.3)< 0.001*< 0.001a< 0.001cNormal Saline645.59 ± 2.1412 (18.8)Isolyte S127.50 ± 1.516 (50.0)Dextrose49.75 ± 1.264 (100.0)High-Risk Patients (BISAP 2–3), n = 49Lactated Ringer’s198.53 ± 2.0112 (63.2)< 0.001*< 0.001aNormal Saline189.39 ± 2.4514 (77.8)Isolyte S911.11 ± 2.209 (100.0)Dextrose39.00 ± 3.461 (33.3)*: Kruskal–Wallis test, pairwise comparisons: a: Lactated Ringer’s vs. Isolyte S, p < 0.05; b: Lactated Ringer’s vs. dextrose, p < 0.05; c: normal saline vs. Isolyte S, p < 0.05; d: normal saline vs. dextrose, p < 0.05 Table 7Analysis stratified by BISAP risk categoryFluid Type n LOS (days), mean ± SDProlonged Stay (> 7 days), n (%)P valueLow-Risk Patients (BISAP 0–1), n = 155Lactated Ringer’s755.41 ± 1.697 (9.3)< 0.001*< 0.001a< 0.001cNormal Saline645.59 ± 2.1412 (18.8)Isolyte S127.50 ± 1.516 (50.0)Dextrose49.75 ± 1.264 (100.0)High-Risk Patients (BISAP 2–3), n = 49Lactated Ringer’s198.53 ± 2.0112 (63.2)< 0.001*< 0.001aNormal Saline189.39 ± 2.4514 (77.8)Isolyte S911.11 ± 2.209 (100.0)Dextrose39.00 ± 3.461 (33.3)*: Kruskal–Wallis test, pairwise comparisons: a: Lactated Ringer’s vs. Isolyte S, p < 0.05; b: Lactated Ringer’s vs. dextrose, p < 0.05; c: normal saline vs. Isolyte S, p < 0.05; d: normal saline vs. dextrose, p < 0.05 Analysis stratified by BISAP risk category < 0.001* < 0.001a < 0.001c < 0.001* < 0.001a *: Kruskal–Wallis test, pairwise comparisons: a: Lactated Ringer’s vs. Isolyte S, p < 0.05; b: Lactated Ringer’s vs. dextrose, p < 0.05; c: normal saline vs. Isolyte S, p < 0.05; d: normal saline vs. dextrose, p < 0.05 In high-risk patients (BISAP 2–3, n = 49), the absolute LOS was longer across all fluid types. LR yielded the shortest stays (8.53 ± 2.01 days, n = 19), significantly shorter than Isolyte S (11.11 ± 2.20 days, n = 9; p < 0.001). NS (9.39 ± 2.45 days, n = 18) and dextrose (9.00 ± 3.46 days, n = 3) showed intermediate durations. Notably, all high-risk patients treated with Isolyte S (100%, 9/9) had prolonged hospital stays, compared with 63.2% in the LR group and 77.8% in the NS group. Discussion This retrospective cohort study of 204 patients with mild AP identified three key findings. LR and NS showed comparable clinical outcomes with no significant difference in LOS or BISAP scores. Both were associated with significantly shorter hospital stays than Isolyte S and dextrose-containing solutions (p < 0.001 for all comparisons). A statistically significant interaction was observed between fluid type and resuscitation rate (β = −0.287, p = 0.003), suggesting that the optimal infusion rate may vary by crystalloid composition. These fluid-related associations persisted across all severity subgroups stratified by BISAP. These findings represent associations and should not be interpreted as causal. The comparable outcomes of LR and NS are consistent with several meta-analyses reporting no significant differences in SIRS, mortality, or organ failure [20, 21, 24, 26]. Although some analyses favor LR for reducing progression to moderate-to-severe AP and ICU admission [7, 35], these benefits may be most pronounced where systemic inflammation is more marked — a condition less likely in our exclusively mild AP cohort (mean BISAP 0.86–0.88). Nevertheless, guidelines continue to recommend LR [18, 19], and NS may be a reasonable alternative when LR availability is limited. This study presents the first systematic comparison of Isolyte S and dextrose-containing solutions in AP. Both were associated with significantly longer hospital stays compared to LR and NS (p < 0.001; Table 2), and these associations persisted after multivariable adjustment (β = +0.576, p < 0.001; Table 5). Regarding dextrose-containing fluids, these findings align with Wu et al., who reported that glucose-containing fluids significantly increased organ failure and ICU admission rates in 1,146 AP patients, potentially through hyperglycemia-induced oxidative stress and endothelial dysfunction [36]. The Isolyte S results warrant careful interpretation. To our knowledge, no prior studies have compared acetate/gluconate-buffered crystalloids with LR or NS in AP [7]. The significantly higher baseline BISAP scores (1.71 ± 0.90 vs. 0.86–0.88), older mean age (67.2 vs. 54.4 years), and higher hypertension prevalence (71.4% vs. 39.4%) in the Isolyte S group strongly suggest confounding by indication — clinicians likely selected this fluid for patients perceived to be more severely ill. Despite multivariable adjustment, residual confounding cannot be excluded, and the small sample size (n = 21) limits statistical precision. Findings for Isolyte S should therefore be regarded as hypothesis-generating only, and prospective randomized trials are needed to clarify the role of acetate-buffered solutions in AP. The primary novel finding is the statistically significant interaction between fluid type and resuscitation rate (β = −0.287, 95% CI: −0.47 to − 0.10, p = 0.003; Table 6), suggesting that optimal outcomes may be associated with specific fluid-rate combinations rather than either variable alone. To our knowledge, no prior study has formally tested this interaction in AP; previous systematic reviews compared fluid types and resuscitation rates separately but did not examine their combined associations [22, 33]. Notably, the WATERFALL trial employed an individualized protocol in which each patient was assessed for the need for bolus fluid administration, underscoring that no single resuscitation rate is optimal for all patients [29]. Furthermore, the trial evaluated resuscitation rates using LR exclusively, leaving open the question of whether rate-outcome associations differ by crystalloid type; our findings suggest that the absence of benefit from aggressive resuscitation observed in that trial may be specific to LR and may not generalize to other crystalloids, consistent with 2024 ACG guidelines emphasizing individualized therapy [19]. Stratification by BISAP risk category confirmed consistent fluid type associations across severity levels (Table 7). In both low-risk (BISAP 0–1, n = 155) and high-risk (BISAP 2–3, n = 49) subgroups, LR and NS were associated with shorter hospital stays than Isolyte S and dextrose (p < 0.001). LR and NS showed comparable outcomes across both severity levels, consistent with Lipinski et al. [37]. The ongoing WATERLAND study (NCT05781243) [38] may provide definitive answers regarding fluid selection across AP severity subgroups. In the multivariable analysis (Table 5, Model B), age, BISAP score, and hypertension were independent predictors of LOS, consistent with the existing literature [39–41]. Each 1-point increase in BISAP score predicted nearly one additional day of hospitalization (β = +0.830), validating its prognostic utility in mild AP [42, 43]. Hypertension may delay recovery through microvascular dysfunction or may reflect a broader comorbidity burden, rather than acting as an independent determinant of mortality in mild AP [44]. Sex, diabetes mellitus, hyperlipidemia, and alcohol use disorder were not significant predictors. Limitations This study has several important limitations. First, the retrospective design introduces selection bias; fluid type and infusion rates were driven by clinical judgment rather than randomization, and unmeasured confounders (e.g., initial hemodynamic trajectory, hemodynamic instability at presentation) may have influenced both treatment decisions and outcomes. The differential baseline profiles of the Isolyte S and dextrose groups are consistent with confounding by indication, and residual confounding cannot be fully excluded despite multivariable adjustment. Second, small sample sizes in the Isolyte S (n = 21) and dextrose (n = 7) groups limit statistical precision; findings for these fluids are exploratory and hypothesis-generating only. Third, the inclusion of only mild AP cases limits applicability to moderate or severe disease. Fourth, the single-center design may limit external validity. Fifth, LOS is influenced by non-physiological factors (e.g., discharge criteria, institutional culture), introducing potential heterogeneity. Sixth, time from symptom onset to admission — a critical determinant of resuscitation outcomes, as early fluid administration is associated with superior outcomes — was not consistently documented in the medical records and therefore could not be included in the analysis or adjusted for in the multivariable models. This represents a potentially important source of residual confounding that may limit the interpretability of the fluid-rate associations reported. Seventh, composite comorbidity indices (CCI, ASA) were not systematically available; residual comorbidity-related confounding cannot be excluded. Finally, while exclusion criteria were applied at the time of admission (moderate or severe AP at presentation, sepsis), systematic prospective tracking of in-hospital disease progression was not a pre-specified component of this retrospective study, and progression data were not systematically extracted from medical records as part of the study protocol; a major imbalance in progression across groups is unlikely but cannot be formally excluded. This represents an important limitation, and future prospective studies should systematically document in-hospital disease progression as a co-primary or secondary outcome. Strengths and clinical implications Despite these limitations, this study makes significant contributions. To the best of our knowledge, this is the first study to: (1) systematically compare four types of crystalloids (LR, NS, Isolyte S, and dextrose) in AP; (2) demonstrate a statistically significant interaction between fluid type and resuscitation rate (β = −0.287, p = 0.003); and (3) show consistent associations across all severity subgroups stratified by BISAP. There are three key clinical implications: LR and NS were associated with comparable outcomes—and with shorter LOS than Isolyte S and dextrose—in mild AP; the optimal resuscitation rate may be fluid-dependent; and NS may be a reasonable alternative when LR availability is limited. These findings underscore critical gaps that warrant prospective randomized trials using factorial designs to validate fluid-rate interactions and guide individualized protocols. Conclusions This single-center retrospective study of 204 patients with mild AP found that LR and NS were associated with comparable clinical outcomes, while both were associated with significantly shorter hospital stays than Isolyte S and dextrose-containing solutions. A key novel finding was the statistically significant interaction between fluid type and resuscitation rate (β = −0.287, p = 0.003), suggesting that outcomes may vary by specific fluid-rate combinations rather than by either variable in isolation. Given the observational design, potential residual confounding, and small sample sizes for Isolyte S (n = 21) and dextrose (n = 7), these findings should be interpreted as hypothesis-generating. Prospective randomized factorial trials are needed before these results can inform clinical practice guidelines. Supplementary Information Supplementary Material 1. Supplementary Material 1.