This hospital-based retrospective cohort study included patients admitted with acute pancreatitis between January 2018 and December 2023. Data were extracted from the hospital's electronic medical record system. Time zero was defined as the first qualifying hospital admission for acute pancreatitis during the study period. The prediction horizon was defined as the occurrence of moderately severe or severe acute pancreatitis (MSAP/SAP) during the same index hospitalization, in accordance with the Revised Atlanta Classification.

The following demographic and clinical variables were collected: sex, age, body mass index (BMI), waist circumference, body temperature, heart rate, respiratory rate, systolic blood pressure, diastolic blood pressure, history of hypertension, diabetes, fatty liver, hyperlipidemia, alcohol use, smoking, etiology (biliary, hyperlipidemia, alcohol-related, unknown), complete blood count parameters, liver and renal function tests, lipid profile, pancreatic enzymes, CRP, procalcitonin (PCT), heparin-binding protein (HBP), lactate, and coagulation indices. Laboratory parameters used in the primary analyses (including CRP, serum calcium, complete blood count, albumin, PCT, HBP, lactate, and coagulation indices) were obtained from the first blood sample collected within 24 h of hospital admission, prior to the development of persistent organ failure.

Inclusion: all patients hospitalized with AP from January 2018 to December 2023.

Exclusion: (1) chronic pancreatitis; (2) multiple malignancies (pancreatic, esophageal, colorectal, breast, etc.); (3) pregnancy; (4) incomplete medical records; (5) age <18 or >80 years; (6) admission >7 days after symptom onset.

A total of 1,981 patients were included. According to the revised Atlanta classification [20], patients were categorized into mild AP (MAP, n = 1,058) and moderately severe or severe AP (MSAP/SAP, n = 923). The study flowchart and patient selection process are shown in Figure 1. MAP is defined as the absence of organ failure and complications; MSAP is characterized by transient (<48 h) organ failure or local/systemic complications without persistent organ failure; SAP is defined as persistent (≥48 h) single or multiple organ failure.

This study was designed as an etiologic association analysis rather than a predictive modeling study. Therefore, using the “10 events per variable” principle (21), the 923 MSAP/SAP cases exceeded the required minimum of 280, confirming adequate sample size.

The study was conducted in compliance with the Declaration of Helsinki (22) and approved by the Ethics Committee of the First Affiliated Hospital of Bengbu Medical College (Approval No.: 2020KY073). As anonymized retrospective data were used, informed consent was waived (23). Patient confidentiality was maintained through encryption and strict privacy protocols.

To assess the relationship between inflammatory status and AP severity, the following inflammatory indices were calculated:

NLR = Neutrophils/Lymphocytes (15);

PLR = Platelets/Lymphocytes (15);

MLR = Monocytes/Lymphocytes (24);

LMR = Lymphocytes/Monocytes (25);

CAR = CRP/Albumin (26);

CALLY = CRP/(Albumin × Lymphocytes) (27);

CCR = CRP/Calcium (13);

CLR = CRP/Lymphocytes (28),

RDW/Alb = RDW/Albumin (25),

SII = Platelets × Neutrophils/Lymphocytes (20),

NAR = Neutrophils/Albumin (29),

SIRI = Neutrophils × Monocytes/Lymphocytes (29),

Baseline patient characteristics were described according to AP severity (MAP vs. MSAP/SAP). Continuous variables were expressed as mean ± standard deviation for normally distributed data or as median (interquartile range) for non-normally distributed data, while categorical variables were expressed as frequencies and percentages. Group differences were assessed using ANOVA for normally distributed continuous variables, the Kruskal–Wallis test for non-normally distributed continuous variables, and the chi-square test for categorical variables. For comparability of effect estimates across predictors with different measurement scales, continuous variables included in the logistic regression models were standardized using z-score transformation (mean = 0, standard deviation = 1). Accordingly, odds ratios (ORs) derived from logistic regression analyses represent the change in odds per one–standard deviation increase in the corresponding standardized predictor.

To evaluate the association between the 12 inflammatory indices and the incidence of MSAP/SAP, univariate and multivariate logistic regression models were used to calculate odds ratios (ORs) and 95% confidence intervals (CIs). Model 1 was unadjusted, Model 2 was adjusted for gender and age, and Model 3 was further adjusted for the following covariates: gender, age, BMI, waist circumference, body temperature, heart rate, respiratory rate, systolic blood pressure, diastolic blood pressure, history of hypertension, diabetes, fatty liver, hyperlipidemia, alcohol consumption, smoking, etiology, hematocrit, platelet count, creatinine, BUN, sodium, potassium, chloride, PCT, HBP, lactate, PT, APTT, TT, and INR, to assess the robustness of the associations after extensive adjustment, however, the possibility of overadjustment cannot be completely excluded. VIF analysis was performed to test for multicollinearity. Restricted cubic spline (RCS) analysis was applied to explore nonlinear associations between inflammatory indices and MSAP/SAP incidence. RCS models were fitted using four knots placed at the 5th, 35th, 65th, and 95th percentiles of each marker distribution, following Harrell's recommended default settings. Standardization was not applied to CCR in analyses involving nonlinear relationships or absolute cutoff determination. ROC curve analysis was conducted to evaluate the discrimination performance of each inflammatory index. For the index with the greater discrimination performance, a segmented logistic regression model was applied to identify potential threshold effects, in which the change point was estimated using an iterative algorithm to determine the optimal cutoff value. To assess the robustness of the discrimination performance and account for potential optimism, internal validation was performed using bootstrap resampling with 1,000 iterations. Optimism-corrected area under the receiver operating characteristic curve (AUC) was calculated for each index. Differences in AUCs between inflammatory markers were formally compared using the DeLong test for correlated ROC curves, with corresponding P values reported.

All statistical analyses were performed using R software (version 4.4.1, R Foundation, http://www.R-project.org), and statistical significance was defined as a two-sided P < 0.05.

This study included patients with MAP (n = 1,058) and MSAP/SAP (n = 923), and baseline characteristics are summarized in Table 1. Compared to the MAP group, the MSAP/SAP group exhibited significantly higher values for the following indicators: age, BMI, waist circumference, heart rate, respiratory rate, prevalence of diabetes, fatty liver, history of hyperlipidemia, history of alcohol consumption, history of smoking, white blood cell count, red blood cell count, neutrophil count, monocyte count, hemoglobin, hematocrit, RDW, triglycerides, CRP, PCT, HBP, PT, INR, fibrinogen, NLR, PLR, MLR, CAR, CCR, CLR, RDW/Alb, SII, NAR, and SIRI.

Conversely, the MSAP/SAP group demonstrated significantly lower values for the following indicators: prevalence of hypertension, ALP, albumin, sodium, calcium, HDL-C, LMR, and CALLY.

VIF analysis indicated that none of the covariates exhibited multicollinearity, as all VIF values were less than 4 (Supplementary Table 1). In the logistic regression analysis, three models were constructed sequentially: Model 1 (unadjusted), Model 2 (adjusted for gender and age), and Model 3 (further adjusted for 28 covariates including BMI, waist circumference, vital signs, comorbidities, laboratory indicators, and coagulation function). Model 3 passed collinearity detection with all VIF values less than 5. In Model 3, MLR (OR = 1.29, 95%CI: 1.15–1.45, P < 0.001), LMR (OR = 0.75, 95%CI: 0.66–0.85, P < 0.001), CAR (OR = 3.82, 95%CI: 3.18–4.64, P < 0.001), CALLY (OR = 0.56, 95%CI: 0.49–0.64, P < 0.001), CCR (OR = 4.84, 95%CI: 3.98–5.96, P < 0.001), CLR (OR = 2.12, 95%CI: 1.84–2.46, P < 0.001), RDW/Alb (OR = 1.74, 95%CI: 1.54–1.99, P < 0.001), NAR (OR = 1.44, 95%CI: 1.27–1.64, P < 0.001), and SIRI (OR = 1.29, 95%CI: 1.15–1.46, P < 0.001) were significantly associated with the risk of MSAP/SAP. In contrast, NLR (P = 0.11), PLR (P = 0.40), and SII (P = 0.091) did not show statistical significance after multivariate full adjustment. The associations between inflammatory markers and MSAP/SAP are shown in Table 2.

ROC curve analysis showed that CCR had the highest AUC for predicting MSAP/SAP (AUC = 0.768, 95%CI: 0.737–0.799), indicating better discriminatory performance than the other indices (Figure 2). Additionally, RCS curve analysis indicated that, except for NAR, RDW/Alb, and MLR, all other inflammatory indices exhibited significant nonlinear relationships with the risk of MSAP/SAP (P < 0.05). After performing internal validation using bootstrap resampling, CCR remained the highest-performing inflammatory index, demonstrating the highest optimism-corrected AUC. Pairwise comparisons using the DeLong test confirmed that the AUC of CCR was significantly higher than those of the other indices (all P < 0.05).

Given the evaluation of multiple inflammatory indices across several models, including spline and threshold analyses, CCR was selected as the primary index for further in-depth analyses due to its numerically higher ROC-AUC in the main models.

RCS analysis demonstrated that NLR, PLR, MLR, CAR, CALLY, CCR, CLR, SII, and SIRI were significantly associated with the risk of MSAP/SAP, showing pronounced nonlinear dose–response relationships. In contrast, LMR, RDW/Alb, and NAR exhibited significant overall associations but without meaningful nonlinear trends. The detailed dose–response curves are presented in Figure 3.

Further threshold effect analysis revealed a significant nonlinear association between CCR and the risk of MSAP/SAP (P < 0.001). The conventional logistic regression model suggested that for every one-unit increase in CCR, the risk of MSAP/SAP increased by 15.7% (OR = 1.157, 95%CI: 1.14–1.176, P < 0.001). The segmented logistic regression model an inflection point (change point) at CCR = 15. When CCR < 15, there was no significant association with MSAP/SAP risk (OR = 1.015, 95%CI: 0.965–1.068, P = 0.558); however, when CCR ≥ 15, the risk of MSAP/SAP significantly increased with increasing CCR (OR = 1.212, 95%CI: 1.182–1.245, P < 0.001). The likelihood ratio test further supported that the segmented logistic regression model provided a better fit than the conventional logistic regression model (P < 0.001).

To further evaluate the potential clinical triage utility of CCR, diagnostic performance metrics were calculated at the optimal cutoff value (CCR = 16.835). At this threshold, CCR demonstrated a sensitivity of 0.659 and a specificity of 0.785. The positive predictive value (PPV) was 0.730, and the negative predictive value (NPV) was 0.723. In addition, the positive likelihood ratio (LR+) was 3.069, and the negative likelihood ratio (LR−) was 0.434, indicating a moderate ability of CCR to discriminate patients at higher risk of MSAP/SAP at admission. The threshold effect analysis results are presented in Table 3.