Hypertriglyceridemic acute pancreatitis with diabetes mellitus: Investigation of factors and predictive models for severity progression

Abstract

BACKGROUND

In recent years, the incidence of acute pancreatitis (AP) caused by hypertriglyceridemia has been on the rise. In China, it has surpassed alcohol to become the second leading cause of AP. Patients with hypertriglyceridemia-induced AP combined with diabetes mellitus (HTG-AP-DM) have a higher probability of developing severe disease.

AIM

To investigate the risk factors for progression to severe disease in HTG-AP-DM and to develop and validate a predictive model.

METHODS

This retrospective study, with a sample size of 133 patients, analyzed clinical data from patients with HTG-AP-DM admitted to the First Hospital of Jilin University between January 2019 and December 2024. Disease severity was classified according to the Revised Atlanta Classification, with mild and moderately severe cases categorized as the non-severe group and severe cases as the severe group. Potential predictors were initially screened using univariate logistic regression, followed by multivariate logistic regression to identify independent risk factors for progression to severe disease in patients with HTG-AP-DM. A predictive model was subsequently developed based on these factors. Model performance was evaluated using receiver operating characteristic curve analysis to quantify discriminative ability using the area under the curve (AUC), calibration curves to visualize agreement between predicted probabilities and observed outcomes, and decision curve analysis to assess clinical utility by calculating net benefit across threshold probabilities.

RESULTS

A retrospective analysis of 133 patients with HTG-AP-DM identified several independent risk factors for disease progression through univariate and multivariate logistic regression: Heart rate [odds ratio (OR) = 1.03, 95% confidence interval (CI): 1.01-1.05, P = 0.027], albumin (OR = 0.91, 95%CI: 0.84-0.98, P = 0.019), platelets (OR = 0.99, 95%CI: 0.98-0.99, P < 0.001), serum lipase (OR = 1.01, 95%CI: 1.01-1.01, P = 0.014), random plasma glucose (OR = 1.22, 95%CI: 1.07-1.39, P = 0.003), and pleural effusion (OR = 3.83, 95%CI: 1.30-11.25, P = 0.015). The predictive model incorporating these variables demonstrated excellent discriminative power (AUC = 0.894, 95%CI: 0.840-0.949) and good calibration between predicted and observed probabilities, whereas decision curve analysis confirmed its clinical utility across a wide range of probability thresholds, suggesting robust potential for clinical application.

CONCLUSION

Heart rate, serum lipase, random plasma glucose, and pleural effusion were identified as independent risk factors for disease progression to a severe stage, whereas albumin and platelets were independent protective factors. The predictive model developed based on these variables can rapidly assess whether a patient will progress to severe illness within 24 hours of admission, with an AUC of 0.894, demonstrating excellent prognostic performance.

Key Words: Hypertriglyceridemic acute pancreatitis; Diabetes mellitus; Severity; Clinical prediction model; Hyperlipidemia; Albumin and platelets; Protective factors

Core Tip: In a retrospective cohort of 133 patients with hypertriglyceridemic acute pancreatitis combined with diabetes mellitus, multivariate analysis revealed that heart rate, serum lipase, random plasma glucose, and pleural effusion were independent risk factors for severe disease, while albumin and platelets were protective. The predictive model incorporating these variables exhibited excellent discrimination (area under the curve = 0.894), good calibration, and net clinical benefit, aiding early risk stratification.

INTRODUCTION

Acute pancreatitis (AP) is a clinically heterogeneous disorder triggered by diverse etiological factors, characterized pathologically by premature activation of pancreatic enzymes leading to localized inflammatory responses in peri-pancreatic tissues, which may or may not be accompanied by systemic organ dysfunction. This condition arises from multiple pathogenic mechanisms, ultimately resulting in acute pancreatic tissue injury manifesting as edema, hemorrhage, and necrosis[1]. Biliary disorders, particularly choledocholithiasis, are the most common cause of AP. Chronic excessive alcohol consumption constitutes another major causative factor, typically ranking as the second most common cause. Additional triggers include specific medications (e.g., azathioprine, thiazide diuretics), surgical or nonsurgical trauma, infectious conditions, and immune-mediated diseases. Notably, the incidence of hypertriglyceridemia (HTG)-induced AP (HTG-AP) has demonstrated a significant upward trend in recent years. In China, HTG has now surpassed alcohol to become the second leading cause of AP[2].

Globally, HTG-AP accounts for 2%-5% of all AP cases. However, its proportion rises significantly to 10%-20% among cases of severe AP (SAP)[3]. HTG is associated with multiple phenotypic syndromes, with considerable overlap in their genetic determinants. The most common form is caused by polymorphisms that increase very-low-density lipoprotein production, leading to a moderate elevation in triglyceride (TG) levels (typically < 10 mmol/L), while total cholesterol and apolipoprotein B remain normal. Multifactorial chylomicronemia is a heterogeneous syndrome resulting from polygenic effects on chylomicrons and impaired clearance of very-low-density lipoprotein remnants, often leading to variable severity of HTG[4]. Familial chylomicronemia is a rare disorder (affecting approximately 1 in 300000 individuals), usually caused by pathogenic variants in the lipoprotein lipase (LPS) gene, resulting in persistent severe HTG and recurrent pancreatitis[5]. SAP can lead to local complications, including pancreatic tissue necrosis, pseudocyst formation, and peripancreatic abscesses. Systemically, patients often develop critical conditions such as multiple organ dysfunction syndrome, septic shock, and metabolic disturbances, with notably high rates of sepsis and mortality. A case report suggests that HTG-AP can progress to necrotizing pancreatitis, which may lead to organ failure and even death[6]. A Chinese epidemiological study revealed that among 475 patients with moderately severe or SAP, 108 (22.7%) were HTG-AP cases, with annual hospitalization rates increasing steadily from 14.3% to 35.5%. The analysis indicates a progressive annual rise in HTG-AP prevalence in China, a trend strongly associated with contemporary lifestyle changes, increased incidence of metabolic disorders, and chronic nutritional excess in the general population[7]. After following 400 patients with AP, the researchers found that the HTG group showed a trend toward younger age and a male-dominated gender composition, accompanied by higher rates of obesity and diabetes, compared to the non-HTG group[8].

Diabetes mellitus (DM) and HTG exhibit a bidirectional pathological relationship. On one hand, diabetic patients frequently present with concomitant lipid metabolism disorders such as HTG, which significantly elevates the risk of developing HTG-AP. On the other hand, the acute inflammatory response and pancreatic injury in HTG-AP may exacerbate insulin resistance, further worsening glycemic control, thereby creating a vicious cycle. Epidemiological data estimate that approximately 129.8 million adult patients with diabetes currently live in mainland China[9]. Relevant studies indicate that patients with coexisting HTG-AP and DM demonstrate significantly higher rates of disease recurrence and in-hospital mortality compared to non-diabetic HTG-AP patients[10]. Therefore, early risk assessment for critical illness in patients with HTG-AP and diabetes is particularly important within the first 24 hours of admission. This timeframe is considered a critical window for identifying the risk of complications or death and for implementing proactive preventive measures[11].

Currently, established scoring systems for AP include the Bedside Index for Severity in AP (BISAP), Acute Physiology and Chronic Health Evaluation II (APACHE II), Sequential Organ Failure Assessment (SOFA), and computed tomography (CT) severity index (CTSI). Although these systems are widely used in clinical practice, each has distinct advantages and limitations. The BISAP score is simple and rapid, making it suitable for initial triage in the emergency department; however, it lacks predictive power for local complications. The APACHE II score is comprehensive and well-suited for dynamic monitoring of intensive care unit patients, but its complexity limits its use in primary care settings. The SOFA score demonstrates exceptional performance, particularly in predicting disease severity and mortality risk, although its calculation process is relatively complex. The CTSI offers unique advantages in evaluating local complications, although it relies on imaging studies. Importantly, no clinical multifactorial scoring system currently exists that is specifically designed for patients with HTG-AP-DM.

Based on readily available clinical, demographic, laboratory, and imaging data, this study developed the first 24-hour prognostic model for patients with high-severity HTG-AP-DM. This model serves as a quantitative tool for assessing the risk of early deterioration in HTG-AP-DM patients in emergency departments and intensive care units, providing a basis for implementing stratified management and targeted interventions. Furthermore, it supports critical decision-making in establishing early goal-directed therapy strategies in this high-risk population.

MATERIALS AND METHODS

Diagnostic criteria

AP: (1) Clinical manifestations: Characteristic acute onset of severe epigastric pain, persistent and potentially radiating to the back; (2) Laboratory findings: Serum amylase (AMY) and/or LPS levels exceeding three times the upper limit of the normal reference range; and (3) Imaging findings: Typical features of AP, such as pancreatic parenchymal edema or peripancreatic exudative changes. Fulfilling any two of the three criteria listed above is sufficient to establish the diagnosis of AP.

HTG-AP: (1) Meets diagnostic criteria for AP; and (2) Serum TG level ≥ 11.30 mmol/L, or serum TG level between 5.65 mmol/L and 11.30 mmol/L with lactescent serum, after exclusion of other etiologies of AP (e.g., biliary disease, alcohol, drugs, trauma, etc.).

DM: (1) Fasting blood glucose ≥ 7.0 mmol/L; (2) 2-hour blood glucose level ≥ 11.1 mmol/L on an oral glucose tolerance test (OGTT); (3) Random blood glucose ≥ 11.1 mmol/L, accompanied by typical symptoms of diabetes (such as increased thirst, increased appetite, increased urination, and weight loss); and (4) Glycated hemoglobin ≥ 6.5%. Fulfilling any single one of the above criteria is sufficient for the diagnosis of DM.

Revised Atlanta Classification: (1) Mild AP: Absence of organ failure and local/systemic complications; (2) Moderately SAP: Presence of transient organ failure (≤ 48 hours) and/or local or systemic complications; and (3) SAP: Presence of persistent organ failure lasting > 48 hours. In this study, mild and moderately severe cases were classified as the non-severe group, whereas severe cases were categorized as the severe group.

Modified Marshall Score: The scoring system evaluates three key parameters with progressive severity grades (0-4 points): (1) Partial pressure of arterial oxygen/fraction of inspired oxygen (PaO₂/FiO₂) ratio (mmHg): > 400 = 0, 301-400 = 1, 201-300 = 2, 101-200 = 3, ≤ 100 = 4; (2) Systolic blood pressure (mmHg): > 90 = 0, < 90 with fluid-responsive = 1, < 90 with fluid-unresponsive = 2, < 90 with pH < 7.3 = 3, < 90 with pH < 7.2 = 4; and (3) Serum creatinine (Cr) (μmol/L): ≤ 134 = 0, 134-169 = 1, 170-310 = 2, 311-439 = 3, > 439 = 4. Organ function was evaluated in patients with HTG-AP-DM using the Modified Marshall Score; a score of ≥ 2 defined organ failure.

Study design and participant selection

This retrospective study systematically collected clinical data from hospitalized patients diagnosed with HTG-AP-DM at the First Hospital of Jilin University between January 2019 and December 2024. The study was approved by the hospital’s Ethics Committee, approval No. 2024-865.

Inclusion criteria: (1) Met the diagnostic criteria for HTG-AP; (2) Pre-existing diagnosis of DM prior to hospitalization; (3) Age ≥ 18 years, regardless of sex; and (4) Hospital admission within 48 hours of symptom onset.

Exclusion criteria: (1) Incomplete clinical data; (2) Comorbid malignancies, autoimmune disorders, or other severe systemic diseases (cardiac, pulmonary, hepatic, or renal); and (3) Pregnancy or lactation.

This study enrolled 354 patients with HTG-AP who met the inclusion criteria at our institution. After excluding 23 cases with incomplete clinical data and 7 cases with severe comorbidities or pregnancy, 324 patients with HTG-AP remained, of whom 133 with DM were screened for inclusion in the analysis, including 60 cases in the severe group and 73 cases in the non-severe group. The process is shown in Figure 1.

Figure 1 Participant selection flowchart. HTG-AP: Hypertriglyceridemia-induced acute pancreatitis.

Data collection

The collected data encompassed both demographic characteristics and clinical parameters of the patients. Demographic and baseline characteristics included age, sex, height, weight, admission time, present illness, past medical history, and smoking or alcohol use history. The collected clinical data comprised: (1) Initial vital signs upon admission [heart rate (HR) and respiratory rate]; and (2) Worst laboratory values within the first 24 hours, including complete blood count [white blood cells, neutrophils, lymphocytes, hemoglobin, red cell distribution width, platelets (PLT)], coagulation profile [prothrombin time (PT)], liver function tests [alanine aminotransferase, albumin (ALB)], inflammatory markers (C-reactive protein, procalcitonin), renal/electrolyte parameters [Cr, serum calcium (Ca²⁺), sodium (Na⁺)], metabolic panel (glycosylated hemoglobin, cholesterol, TG), pancreatic enzymes (AMY, serum LPS), glucose levels [fasting plasma glucose and random plasma glucose (RPG)], arterial blood gases, and imaging-documented pleural/abdominal effusion, along with the calculated neutrophil-to-lymphocyte ratio and Modified Marshall Score.

Statistical analysis

Statistical analysis was performed using SPSS 27.0 and R 4.4.2, and data visualization was conducted using GraphPad Prism 10 and R 4.4.2. Variables with missing data < 10% were handled by multiple imputation, whereas those with missing data > 10% were excluded. Normally distributed continuous variables were expressed as the mean ± SD and compared using independent t-tests; non-normally distributed variables were reported as the median with interquartile range and compared using Mann-Whitney U tests. Categorical variables were presented as frequencies, n (%) and analyzed by χ² tests. Univariate logistic regression identified statistically significant variables, which were subsequently incorporated into multivariate logistic regression to determine independent predictors. Receiver operating characteristic (ROC) curves were generated to calculate the area under the curve (AUC), and a clinical nomogram was subsequently constructed. Internal validation was performed using bootstrap resampling (1000 iterations) to estimate the concordance index for model performance. The model underwent Hosmer-Lemeshow goodness-of-fit testing with calibration curves to verify accuracy. Decision curve analysis (DCA) assessed clinical utility and net benefit. A two-tailed P value < 0.05 was considered statistically significant.

RESULTS

Baseline characteristics of study participants

The study cohort comprised 133 patients, of whom 60 (45.11%) were classified as severe and 73 (54.89%) as non-severe. Comparisons of the baseline characteristics, vital signs, and laboratory test results are shown in Tables 1 and 2. Significant intergroup differences (P < 0.05) were observed in multiple parameters: Body mass index (BMI), HR, respiratory rate, neutrophil-to-lymphocyte ratio, PLT, PT, ALB, C-reactive protein, procalcitonin, Cr, Ca²⁺, TG, AMY, LPS, fasting plasma glucose, RPG, and pleural effusion. These results are shown in Figure 2A-C.

Figure 2 Graphic illustrations of clinical data in patients with hypertriglyceridemia-induced acute pancreatitis combined with diabetes mellitus. P < 0.05. A: Bar chart of statistically significant normally distributed variables. The X-axis represents the variable names, and the Y-axis represents the variable values (units not shown); B: Bar chart of statistically significant categorical variables. The X-axis represents the variable names, and the Y-axis represents the proportions of each variable (%); C: Boxplot of statistically significant nonnormally distributed variables. The X-axis represents the variable names, and the Y-axis represents the common logarithm (log₁₀) of the variable values (units not shown). HR: Heart rate; PT: Prothrombin time; ALB: Albumin; CRP: C-reactive protein; FPG: Fasting plasma glucose; RR: Respiratory rate; NLR: Neutrophil-to-lymphocyte ratio; PLT: Platelets; PCT: Procalcitonin; Cr: Creatinine; TG: Triglyceride; AMY: Serum amylase; LPS: Serum lipase; RPG: Random plasma glucose.

Table 1 Baseline characteristic comparisons between severe and non-severe groups, n (%).

Baseline characteristics	Total (n = 133)	Severe group (n = 60)	Non-severe group (n = 73)	P value
Age, years, median (IQR)	37 (32, 42)	37 (32, 41)	37 (32, 43)	0.435
BMI, kg/m2, median (IQR)	27.30 (24.70, 30.00)	28.48 (25.91, 30.56)	26.40 (23.90, 28.80)	0.011a
Gender	-	-	-	0.386
Male	71 (53.38)	44 (73.33)	27 (36.99)	-
Female	61 (45.86)	16 (26.67)	46 (63.01)	-
Underlying diseases
Hypertension	41 (30.83)	20 (33.33)	21 (28.77)	0.570
Heart disease	2 (1.50)	1 (1.67)	1 (1.37)	1.000
Fatty liver disease	22 (16.54)	6 (10.00)	16 (21.92)	0.066
Social history
Smoking history	48 (36.09)	19 (31.67)	29 (39.73)	0.336
Alcohol use history	51 (38.35)	19 (31.67)	32 (43.84)	0.151

^aIndicate that results are statistically significant.

IQR: Interquartile range; BMI: Body mass index.

Table 2 Vital sign and laboratory test result comparisons between severe and non-severe groups, n (%)/mean ± SD.

Variables	Severe group (n = 60)	Non-severe group (n = 73)	t/Z/χ2	P value
HR, bpm	119.20 ± 24.70	98.44 ± 21.05	-5.231	< 0.001a
RR, bpm	22.00 (18.00, 32.00)	18.00 (18.00, 20.00)	-3.272	< 0.001a
WBC, × 109/L	13.20 ± 7.94	11.48 ± 6.24	-1.401	0.163
NLR, × 109/L	9.33 (5.62, 14.17)	6.83 (4.17, 9.78)	-2.442	0.015a
Hb, g/L	156.97 ± 31.26	156.74 ± 26.89	-0.051	0.964
RDW, %	13.35 (12.88, 14.03)	13.20 (12.80, 13.50)	-1.732	0.083
PLT, × 109/L	190.00 (159.75, 239.50)	240.00 (198.00, 292.00)	-3.922	< 0.001a
PT, second	12.60 ± 1.41	11.92 ± 1.13	-3.081	0.003a
ALT, U/L	19.50 (11.90, 31.88)	21.00 (15.80, 31.00)	-0.922	0.357
ALB, g/L	34.49 ± 7.37	39.28 ± 7.25	3.761	< 0.001a
CRP, mg/L	269.50 ± 147.01	203.66 ± 133.49	-2.701	0.008a
PCT, ng/mL	4.26 (0.97, 11.32)	0.65 (0.16, 2.07)	-4.322	< 0.001a
Cr, μmol/L	105.65 (54.22, 161.22)	62.30 (49.90, 81.30)	-3.952	< 0.001a
Ca2+, mmol/L	1.84 (1.49, 2.02)	2.17 (2.00, 2.32)	-5.102	< 0.001a
Na+, mmol/L	134.31 ± 5.99	135.12 ± 5.06	0.841	0.401
HbA1c, %	9.83 ± 2.30	9.58 ± 1.83	-0.691	0.494
Chol, mmol/L	4.10 (2.77, 10.52)	4.75 (1.80, 8.81)	-0.512	0.611
TG, mmol/L	19.06 (14.02, 27.02)	16.21 (10.28, 24.59)	-2.722	0.006a
AMY, U/L	442.70 (235.95, 777.73)	159.00 (83.00, 403.10)	-3.502	< 0.001a
LPS, U/L	734.12 (342.25, 1371.66)	299.19 (97.00, 781.00)	-3.752	< 0.001a
FPG, mmol/L	16.53 ± 5.04	14.34 ± 3.89	-2.761	0.007a
RPG, mmol/L	16.65 (13.47, 22.05)	13.80 (11.90, 16.10)	-3.752	< 0.001a
pH	7.36 (7.29, 7.40)	7.37 (7.33, 7.41)	-1.322	0.188
Pleural effusion	44 (73.33)	27 (36.99)	17.483	< 0.001a
Abdominal effusion	39 (65.00)	37 (50.68)	2.763	0.097

^aIndicate that results are statistically significant.

¹t-test.

²Mann-Whitney U test.

³χ² test.

HR: Heart rate; RR: Respiratory rate; WBC: White blood cell; NLR: Neutrophil-to-lymphocyte ratio; Hb: Hemoglobin; RDW: Red blood cell distribution width; PLT: Platelet; PT: Prothrombin time; ALT: Alanine aminotransferase; ALB: Albumin; CRP: C-reactive protein; PCT: Procalcitonin; Cr: Creatinine; HbA1c: Glycosylated hemoglobin; TG: Triglyceride; AMY: Serum amylase; LPS: Serum lipase; FPG: Fasting plasma glucose; RPG: Random plasma glucose.

Factors associated with disease progression to severe stages

Univariate logistic regression was performed on the statistically significant variables identified above. Variables demonstrating statistical significance were subsequently entered into multivariate logistic regression analysis using backward stepwise elimination. The results revealed that HR, BMI, PT, ALB, Ca²⁺, LPS, RPG, and pleural effusion were associated with disease progression to severe stages. As BMI, PT, and Ca²⁺ did not show statistical significance (P ≥ 0.05), the remaining variables were used to construct the predictive model (Table 3).

Table 3 Analysis of risk factors for progression to severe disease in patients.

Variables	Univariate logistic regression			Multivariate logistic regression
	OR	95%CI	P value	OR	95%CI	P value
HR	1.04	1.02-1.06	< 0.001a	1.03	1.01-1.05	0.027a
RR	1.07	1.01-1.12	0.012a	-	-	-
BMI	1.12	1.02-1.23	0.023a	1.14	0.97-1.33	0.106
NLR	1.08	1.01-1.16	0.034a	-	-	-
PLT	0.99	0.98-0.99	< 0.001a	0.99	0.98-0.99	< 0.001a
PT	1.54	1.15-2.07	0.004a	0.69	0.43-1.1	0.120
ALB	0.91	0.87-0.96	< 0.001a	0.91	0.84-0.98	0.019a
CRP	1.01	1.01-1.01	0.009a	-	-	-
PCT	1.16	1.06-1.26	< 0.001a	-	-	-
Cr	1.02	1.01-1.03	< 0.001a	-	-	-
Ca2+	0.08	0.03-0.25	< 0.001a	0.26	0.06-1.02	0.053
TG	1.02	0.99-1.05	0.144	-	-	-
AMY	1.00	1.00-1.00	0.386	-	-	-
LPS	1.01	1.01-1.01	0.004a	1.01	1.01-1.01	0.014a
FPG	1.12	1.03-1.21	0.007a	-	-	-
RPG	1.20	1.09-1.31	< 0.001a	1.22	1.07-1.39	0.003a
Pleural effusion	4.69	2.23-9.86	< 0.001a	3.83	1.30-11.25	0.015a

^aIndicate that results are statistically significant.

OR: Odds ratio; CI: Confidence interval; HR: Heart rate; RR: Respiratory rate; BMI: Body mass index; NLR: Neutrophil-to-lymphocyte ratio; PLT: Platelet; PT: Prothrombin time; ALB: Albumin; CRP: C-reactive protein; PCT: Procalcitonin; Cr: Creatinine; TG: Triglyceride; AMY: Serum amylase; LPS: Serum lipase; FPG: Fasting plasma glucose; RPG: Random plasma glucose.

Nomogram prediction model construction and verification

The variance inflation factor was calculated for the independent variables (HR, ALB, PLT, LPS, RPG, and pleural effusion) included in the model, confirming that there was no significant multicollinearity among them (Table 4). The predictive performance of HR, ALB, PLT, LPS, RPG, and pleural effusion was evaluated using ROC curve analysis. The results demonstrated that these indicators had significant diagnostic value for the progression to severe disease in patients with HTG-AP-DM. The AUCs were as follows: 0.738 (95%CI: 0.653-0.823), 0.700 (95%CI: 0.608-0.792), 0.698 (95%CI: 0.610-0.787), 0.690 (95%CI: 0.599-0.780), 0.689 (95%CI: 0.597-0.782), and 0.682 (95%CI: 0.602-0.761), respectively (Figure 3).

Figure 3 Receiver operating characteristic curve of independent predictors among patients. AUC: Area under curve; HR: Heart rate; ALB: Albumin; PLT: Platelets; LPS: Serum lipase; RPG: Random plasma glucose.

Table 4 Multicollinearity analysis of variables in the prediction model.

Variables	VIF	Tolerance
HR	1.131	0.884
PLT	1.078	0.927
ALB	1.181	0.847
LPS	1.097	0.912
RPG	1.102	0.908
Pleural effusion	1.268	0.789

The extent to which the variance of a regression coefficient estimate is inflated due to linear correlations among the independent variables. A variance inflation factor close to 1 indicates no collinearity, while a variance inflation factor ≥ 10 suggests severe collinearity. Tolerance: Measures the degree to which an independent variable cannot be explained by other independent variables, i.e., the ability of that independent variable to uniquely explain the variance in the dependent variable. A tolerance value of 1 indicates no collinearity, whereas a tolerance ≤ 0.05 suggests severe collinearity. VIF: Variance inflation factor; HR: Heart rate; PLT: Platelet; ALB: Albumin; LPS: Serum lipase; RPG: Random plasma glucose.

A nomogram prediction model was constructed using the six variables described above, with the results presented in Figure 4. The AUC of the model was 0.894 (95%CI: 0.840–0.949), indicating high diagnostic value (Figure 5). The Hosmer-Lemeshow goodness-of-fit test yielded a P value of 0.988 (> 0.05), suggesting no significant deviation between predicted and observed outcomes and confirming good model fit.

Figure 4 Nomogram for predicting the severity risk in patients with hypertriglyceridemia-induced acute pancreatitis combined with diabetes mellitus. The nomogram is used by summing all points identified on the scale for each variable. The total points projected on the bottom scales indicated the likelihood of disease severity. HR: Heart rate; ALB: Albumin; PLT: Platelets; LPS: Serum lipase; RPG: Random plasma glucose.

Figure 5 Receiver operating characteristics curves of the predictive model for severity prediction. ROC: Receiver operating characteristic; AUC: Area under curve.

Given the relatively small sample size of patients with HTG-AP-DM, bootstrap resampling (n = 1000 iterations) was performed for internal validation. The corrected concordance index was 0.876, further supporting the robustness of the model. Calibration curve analysis demonstrated excellent agreement between predicted and actual probabilities (Figure 6). Additionally, DCA revealed that this model provided significant clinical net benefit in the HTG-AP-DM cohort (Figure 7).

Figure 6 Correction curves for predicting severity in hypertriglyceridemia-induced acute pancreatitis combined with diabetes mellitus patients. The nomogram model plots predicted severity on the X-axis and actual severity on the Y-axis.

Figure 7 Decision curve analysis of nomogram prediction models. The X-axis shows the threshold probability, and the Y-axis is the net benefit. The red line indicates that no patients progress to severity, and the black line indicates that all patients progress to severity.

Comparison of the prediction model with other scoring methods

The severity progression model developed in this study for patients with hypertriglyceridemic AP complicated by diabetes was named “HTG-AP-DM”. To further validate the clinical applicability of this model, it was compared with commonly used clinical prognostic scores (BISAP, APACHE II, SOFA, and CTSI), and the predictive capacity of each score and model for progression to severe disease in patients with HTG-AP-DM was evaluated using ROC curves. The results demonstrated that the BISAP, APACHE II, SOFA, and CTSI scores, as well as the HTG-AP-DM model, all exhibited significant diagnostic value in predicting progression to severe disease in patients with hypertriglyceridemic AP complicated by diabetes. The AUCs were as follows: 0.770 (95%CI: 0.695-0.8440), 0.755 (95%CI: 0.673-0.837), 0.748 (95%CI: 0.669-0.828), 0.639 (95%CI: 0.559-0.719), and 0.894 (95%CI: 0.840-0.949), respectively (Figure 8). These findings indicate that the HTG-AP-DM model developed in this study outperforms existing clinical scoring systems in terms of predictive efficacy and offers distinct clinical advantages for assessing progression to severe disease in these patients.

Figure 8 Receiver operating characteristic curves of the prediction model and commonly used clinical scoring systems. The receiver operating characteristic curve demonstrates the discriminative efficacy of the predictive model developed in this study compared to existing scoring systems for predicting outcomes in hypertriglyceridemia-induced acute pancreatitis combined with diabetes mellitus patients. The area under the curve is used to evaluate the predictive value of each factor, with higher area under the curve values indicating greater predictive power. AUC: Area under curve; HTG-AP-DM: Hypertriglyceridemia-induced acute pancreatitis combined with diabetes mellitus; BISAP: Bedside Index for Severity in Acute Pancreatitis; APACHEII: Acute Physiology and Chronic Health Evaluation II; SOFA: Sequential Organ Failure Assessment; CTSI: Computed tomography severity index.

Development of an interactive visualization tool for the medical predictive model

In this study, we interactively visualized the constructed HTG-AP-DM predictive model for progression to severe disease, achieving both an intuitive presentation of the model and individualized risk calculation. The specific interface of the visualized model is shown in Figure 9. The relevant platform can be accessed at the following URL: https://htg-ap-dm.shinyapps.io/dynnomapp/. By inputting patient-specific clinical indicator values, the model enables rapid, individualized risk assessment of disease progression to a severe stage, thereby increasing the clinical applicability of the predictive model.

Figure 9 Online dynamic nomogram prediction tool. HTG-AP-DM: Hypertriglyceridemia-induced acute pancreatitis combined with diabetes mellitus; HR: Heart rate; ALB: Albumin; PLT: Platelets; LPS: Serum lipase; RPG: Random plasma glucose.

DISCUSSION

Recently, due to marked shifts in dietary patterns and lifestyle, dyslipidemia-related factors have overtaken alcohol as the second most common risk factor for AP. The occurrence of AP is typically strongly associated with abnormally elevated serum TG levels[12]. In addition to the typical symptoms of AP, HTG-AP is characterized by markedly elevated TG levels, unremarkable elevation of AMY, pseudohyponatremia, frequent comorbidities, a tendency toward severe disease, insidious triggers, and a younger age of onset[13]. A study involving 4387 patients with AP across five hospitals in China from January 2013 to December 2023 demonstrated that, compared with other types of AP, patients with HTG-AP have a significantly increased risk of developing SAP and organ failure[14]. It has also been reported in the literature that SAP can lead to infectious, coagulation, and metabolic complications[15], and some patients may even develop neurological complications[16]. A retrospective study indicated that DM is independently associated with SAP[17]. A prospective observational cohort study also highlighted the importance of admission glucose levels in predicting the severity of AP[18]. Given the increasing incidence of HTG-AP in China, early identification of the disease and assessment of its prognosis are particularly important. Diabetes is the most common comorbidity in this patient population, making prognostic assessment particularly important for these individuals.

This study retrospectively analyzed clinical data from patients with HTG-AP-DM to identify independent factors influencing disease progression to severe forms and to construct a predictive model. Following the principles of simplicity, rapidity, and convenience, the study included only variables readily available at the early clinical stage, including baseline demographic data, laboratory test results, and imaging findings. The study results indicated that disease severity in these patients was associated with HR, ALB, PLT, LPS, RPG, and the presence of pleural effusion. There are currently few studies on prognostic predictors for patients with HTG-AP-DM. Therefore, in depth investigation and analysis of prognostic factors in these patients can enable early assessment and prediction of disease progression, thereby facilitating the delivery of more individualized and precise treatment strategies.

Increased HR is a compensatory response reflecting systemic inflammation and cardiovascular status. During an inflammatory response, the body is in a state of intense stress, which leads to sympathetic excitation and the subsequent release of large amounts of catecholamines and other substances, thereby accelerating HR. The possible mechanism underlying this association in patients with HTG-AP-DM is that hyperglycemia exacerbates intracellular dehydration and electrolyte disturbances, thereby affecting the normal function of cardiomyocytes and leading to an increased HR[19]. A recent retrospective study on HTG-AP during pregnancy found that the decrease in HR following dialysis treatment was statistically significant, which is usually indicative of clinical improvement[20]. Therefore, as a cost-effective and readily available parameter, HR can serve as a prognostic predictor.

The ALB level can serve as an important predictor for evaluating clinical outcomes in patients with various systemic diseases and traumatic injuries. Hypoalbuminemia is often closely associated with adverse patient outcomes, manifesting as a significantly increased risk of mortality, prolonged intensive care unit length of stay, and accelerated disease progression. As a sensitive indicator of nutritional status and the inflammatory response, the degree of hypoalbuminemia is significantly and positively correlated with disease severity[21]. In patients with HTG-AP-DM, early disease-related symptoms such as abdominal pain, nausea, and vomiting lead to reduced food intake, resulting in inadequate protein intake. Additionally, the release of inflammatory mediators caused by the disease disrupts vascular permeability, allowing ALB to leak from the intravascular space into the interstitial space, thereby inducing hypoalbuminemia. Therefore, a decrease in ALB level often reflects the severity of the disease. The study has indicated that ALB can serve as a novel biomarker for the severity of HTG-AP[22], which is consistent with the findings of the present study.

Under disease conditions, PLT are abnormally activated and release various inflammatory mediators, leading to a consumptive reduction in PLT. Patients with DM inherently have vascular endothelial dysfunction and a hypercoagulable state, which makes them more prone to developing thrombocytopenia[23]. Currently, there are few studies on PLT as a prognostic predictor in patients with HTGAP-DM. However, PLT is also one of the laboratory parameters included in several clinical scoring systems, such as the SOFA, which reflects the severity of illness and organ function status in critically ill patients[24]. In this study, PLT can serve as a predictive factor.

During the onset of AP, a storm of pro-inflammatory cytokines [e.g., interleukin (IL)-6, tumor necrosis factor alpha] triggers necroptosis and disruption of cell membrane integrity, leading to the massive release of zymogen granules from acinar cells. Concurrently, abnormal activation of lysosomal hydrolases impairs the autophagic degradation of trypsinogen, further promoting the entry of serum LPS into the systemic circulation through damaged acinar cell membranes. Because the clearance rate of LPS is significantly slower than that of AMY (half-life approximately 7-14 hours), its persistently elevated serum concentration exhibits a wide time window kinetic profile (rising 4-8 hours after disease onset, peaking at 24 hours, and persisting for 5-7 days). Consequently, LPS has been established as a core biomarker for the entire course of AP, including early prediction of severity and late monitoring of complications. Current studies have also shown that elevated levels of this marker are closely associated with the severity of pancreatic inflammation, consistent with the findings of the present study[25-27]. There are few studies on the association between elevated LPS levels and prognosis in patients with HTG AP DM. The predictive model ultimately constructed in this study can serve as a reference for prognostic assessment in this patient population.

RPG is the plasma glucose level measured in a venous blood sample at any time of day, regardless of the timing of the last meal. Generally, the presence of diabetic symptoms together with an RPG level of ≥ 11.1 mmol/L supports a diagnosis of DM. The present study indicates that RPG can serve as a predictive indicator for disease severity in patients with HTGAP-DM. Possible reasons include damage to pancreatic tissue that may impair insulin secretion and insulin action, leading to elevated blood glucose levels. Therefore, the degree of blood glucose elevation may serve as an indicator of the severity of pancreatic injury and metabolic disturbance, thereby reflecting the overall disease severity.

The presence of pleural effusion indicates the spread of the inflammatory response and the occurrence of systemic complications. In patients with HTG-AP-DM, poor vascular conditions impair the inflammatory response and reduce reparative capacity, making them more susceptible to developing pleural effusions. Large pleural effusions may lead to atelectasis, pulmonary infection, and an increased risk of respiratory failure. A meta-analysis concluded that pleural effusion is a common complication of AP, with a high overall incidence, and may be associated with increased mortality in patients with AP[28]. The presence of pleural effusion can be diagnosed by imaging modalities such as color Doppler ultrasound of the chest and CT. In addition to being consistent with previous studies, the independent factors identified in this study include both blood biomarkers and imaging findings, which were integrated into a single model, thereby enhancing the model's reliability and stability.

The predictive model constructed in this study was comprehensively validated using ROC, calibration, and DCA curves. These results demonstrate that the predictive model achieves high accuracy and provides significant guidance for treatment selection and prognostic assessment. Moreover, the model allows rapid assessment within 24 hours of admission. Particularly in emergency and intensive care settings, the prompt application of this model facilitates early identification of high risk patients and optimization of treatment strategies, thereby improving clinical outcomes.

The study has several limitations. First, as a retrospective study, certain factors could not be investigated because patients at our hospital do not routinely undergo testing for inflammatory markers such as IL-1, IL-6, and tumor necrosis factor alpha upon admission. The inclusion of pleural effusion as a variable requires prompt imaging examinations (e.g., chest/abdominal CT or chest/abdominal ultrasound) at admission. All patients included in our study met this condition; however, if other hospitals or institutions do not routinely perform such examinations, this prognostic model cannot be applied. Second, this is a single center study with only internal validation. Third, the retrospective design precluded accurate classification of diabetes subtypes (type 1 vs type 2) and reliable differentiation between first-onset and recurrent AP across all cases; information on the duration of diabetes and details of medications such as beta-blockers were also missing because these data were not uniformly documented in emergency records. However, given that the model focuses on early physiological responses to the current episode within 24 hours, we believe this omission has a limited impact on its predictive capacity for in-hospital severity. Nonetheless, future prospective multicenter studies are warranted to incorporate these variables, further refine risk stratification, and provide more precise support for clinical decision-making.

Severe HTG-AP has significant clinical consequences, including a high risk of complications and mortality, and imposes a heavy burden on healthcare resources and family finances. This study aimed to construct a predictive model for disease severity in patients with HTG-AP-DM, thereby providing a basis for clinical treatment decisions and prognostic assessment.

CONCLUSION

HR, serum LPS, RPG, and pleural effusion were identified as independent risk factors for disease progression to a severe stage, whereas ALB and PLT were independent protective factors. The predictive model developed based on these variables can rapidly assess whether a patient will progress to severe illness within 24 hours of admission, with an AUC of 0.894, demonstrating excellent prognostic performance.