Construction of a nomogram for hypertriglyceridemic severe acute pancreatitis that includes metabolic indexes

Abstract

Background

Current scoring systems for hypertriglyceridaemia-induced acute pancreatitis (HTG-AP) severity are few and lack reliability. The present work focused on screening predicting factors for HTG-SAP, then constructing and validating the visualization model of HTG-AP severity by combining relevant metabolic indexes.

Methods

Between January 2020 and December 2024, retrospective clinical information for HTG-AP inpatients from Weifang People’s Hospital was examined. CT scans of included patients were evaluated for muscle and fat parameters. To identify independent predictors of HTG-SAP, univariate regression, least absolute contraction and selection operator (LASSO) regression, and multivariable logistic regression were conducted. Meanwhile, the nomogram was created for model visualization, and the model was verified for accuracy, consistency, stability, and utility by calibration, clinical decision curve (DCA), as well as receiver operating characteristic (ROC) analyses.

Results

Altogether 244 HTG-AP patients were enrolled, and they were categorized as a severe group (N = 44) or a non-severe group (N = 200) in line with Atlanta classification criteria. The analysis showed that lactate dehydrogenase (LDH), serum creatinine(Scr), visceral adipose tissue index (VATI), serum albumin(ALB), and triglyceride and glucose (TyG) index independently predicted the HTG-AP severity prediction model, and a nomogram was constructed for visualization, with an internally validated Harrell’s consistency index (c-index) of 0.966 (95% CI, 0.943–0.989), besides, calibration curves, ROC, and DCA all revealed that the nomogram had good predictive ability.

Conclusion

LDH, Scr, VATI, ALB, and TyG independently predict HTG-SAP, and our constructed prediction model has high sensitivity and specificity, which can early identify HTG-AP severity, with a view to giving appropriate interventions to the patients in time, delaying the progression of the patients’ conditions, and reducing the complications.

Introduction

Acute pancreatitis (AP) results from premature pancreatic enzyme activation, causing pancreatic tissue digestion and pancreatic or neighboring tissue inflammatory necrosis [1, 2]. Although most of the clinical manifestations of AP are not severe, which are successfully treated clinically, certain AP may cause organ failure (OF), local complications, and systemic inflammatory response), aggravating the disease or even death, triggering the poor prognosis. As living standards improve among individuals, AP caused by hypertriglyceridemia is increasing, and HTG-AP induced AP will be even severe and have the worse prognosis [3]. Despite the current improvements in the treatment of Severe AP (SAP), its morbidity and mortality remain high [4]. Therefore, early assessment of disease progression in patients with HTG-AP and timely and rational treatment are essential to reduce the risk of adverse disease outcomes and death.

Currently, various scoring systems, such as the Bedside Index for Severity in Acute Pancreatitis (BISAP) score, modified computed tomography severity index (MCTSI) score, and Ranson score, are related to AP severity [5–7]; however, the existing scoring systems remain complex and time-consuming, which have low significance in forecasting the HTG-AP severity and prognostic outcome, and these scores lack the assessment of the body’s metabolism [8]. Abnormal metabolic conditions, like insulin resistance (IR) and obesity, are tightly related to AP severity [9, 10]. IR, defined as the inability to maintain glucose homeostasis by receptors unresponsive or with reduced response to insulin, represents the chronic low-grade inflammation that exerts an important pathogenic effect on AP [11]. The hyperinsulinaemiceuglycaemic clamp remains the gold standard to diagnose IR, but it is time-consuming, costly and requires specialized personnel. Notably, the TyG index is suggested to be the creditable alternative marker of IR [12]. However, further investigation is still needed regarding the relation of TyG index with HTG-AP severity.

Obesity is previously suggested to be related to an increased mortality rate and the local and systemic complication rates among AP patients [13], which may be due to the presence of more visceral adipose tissue among obese individuals. Excess proinflammatory factors are activated and released from the visceral adipose tissue during the onset of AP, further aggravating the body’s inflammatory response while contributing to SAP development [10]. Body mass index (BMI) has been extensively applied in assessing obesity because of its non-invasiveness, rapidness and cost-effectiveness. However, differences in muscle and adipose compositions are not considered in BMI, and the same BMI may indicate totally distinct body compositions. Computed tomography (CT) has been frequently employed for diagnosing SAP and is utilized for analyzing body composition. Related studies have demonstrated the stronger relation of visceral adipose tissue (VAT) with AP severity [14]. However, the relation of VAT with HTG-AP severity is rarely explored.

The present work focused on screening and exploring independent clinical risk factors and related metabolic indicators associated with HTG-AP severity, for developing the prediction model to predict HTG-AP severity, which will help early grade HTG-AP severity, with a view to giving patients appropriate interventions in a timely manner, slowing down the progression of the patients’ disease and reducing complications.

Materials and methods

Study population

This study retrospectively included patients hospitalized for HTG-AP between January 2020 and December 2024 in Weifang People’s Hospital. Patients aged 18 to 70 years developing HTG-AP and hospitalized in 7 days after disease onset were included into the present work. AP was diagnosed and classified according to a modified version of the Atlanta Classification, and secondly, the patient had to have concomitant hypertriglyceridemia (serum triglyceride > 11.3 mmol/L or falling in 5.65–11.3 mmol/L and chylous serum). Exclusion criteria included: (a) Systemic treatment at another hospital prior to admission, like plasma exchange, and fluid resuscitation; (b) Severe comorbid diseases on admission, including chronic cardiac, hepatic, or renal failure; (c) A cancer history; (d) Pregnant or in labor; (e) Acute episodes of chronic pancreatitis; and (f) insufficient clinical information. This retrospective study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board (IRB) of Weifang People’s Hospital (Approval No. KYLL20250402-4). The requirement for informed consent was waived by the IRB due to the retrospective and anonymized nature of the data analysis. All patient identifiers were removed prior to analysis to ensure data confidentiality and anonymization.

Clinical and laboratory data

By reviewing electronic medical records, basic characteristics of the patients were extracted retrospectively, including age, gender, BMI, underlying medical history, Normal drinking history, and complications. Related fasting blood findings in 24 h post-admission were measured, like platelet count (PLT), hemoglobin (HB), blood glucose (GLU), white blood cell count (WBC), albumin (ALB), serum creatinine (Scr), total bilirubin (TBIL), lactate dehydrogenase (LDH), Alanine aminotransferase (ALT), aspartate aminotransfer­ase (AST), serum calcium (Ca²⁺), C-reactive protein(CRP), blood amylase (AMY), total cholesterol(TC), triglycerides (TG), and TyG (with TyG index being computed by: Ln [TG (mg/d L) * GLU (mg/dL)/2]) [15]. Besides, we collected severity scores, consisting of Ranson score (0–11), the BISAP score (0–5) and the MCTSI score (0–10) three scores.

Body composition parameters

The participants in the present work completed an abdominal CT scan, and abdominal CT scans obtained at 3rd lumbar vertebral level (L3) were examined with SliceOmatic software (version 5.0; Tomovision, Magog, Canada). Values between − 29 and 150 HU indicated skeletal muscle area (SMA), those between − 190 and − 30 HU indicated subcutaneous adipose tissue (SAT), whereas those between − 150 and 50 HU suggested VAT. These regions were standardized through dividing each region by height square (m²), yielding a subcutaneous adipose tissue index (SATI), a visceral adipose tissue index (VATI), as well as a skeletal muscle index (SMI). Two experienced radiologists with no knowledge of clinical information were responsible for CT measurements. Re-measurements were performed when measurements were inconsistent.

Severity assessment

All diagnostic CT scans were performed within 72 h of admission. We graded AP severity into mild, moderately severe, and severe following Atlanta classification criteria [16]. (1) MAP: MAP shows no local/systemic complications (2) MSAP: MSAP involves OF subsiding in 48 h (transient OF) and/or systemic/local complications with no persistent OF. (3) SAP: SAP presents with persistent OF(> 48 h), consisting of single or multiple OF. Later, all cases were classified as HTG-SAP or HTG-NSAP group. The definition of OF is a modified Marshall score of at least two for any one of the three systems. Local complications include acute peripancreatic fluid collection, acute necrotic fluid collection, the formation of pancreatic pseudocysts, walled-off necrosis, and infected pancreatic necrosis.

Rationality of sample size.

As a retrospective cohort study, our sample size was determined by the available eligible patients meeting strict inclusion/exclusion criteria during the study period. To evaluate the statistical reliability, we conducted a post-hoc power analysis for the primary outcome (e.g., the AUC of the prediction model). With the observed effect size (AUC = 0.966), α = 0.05, and n = 244 (44 HTG-SAP vs. 200 HTG-NSAP), the achieved power was 99%, far exceeding the conventional 80% threshold. This finding indicates that our sample was sufficient to detect this clinically significant effect.

Statistical analysis

R 4.1.2 and SPSS 25.0 were employed for statistical analyses. Normally-distributed data were represented by means ± standard deviation and analyzed by independent sample t-test. Whereas non-normally-distributed ones were represented by medians (lower quartile, upper quartile) [M (QL, QU)] and compared through Mann–Whitney rank-sum test. Counts were represented by case numbers and percentages, and later analyzed through chi-square test. P < 0.05 stood for statistical significance.

The best predictors of HTG-SAP severity were screened using univariate analysis and least absolute shrinkage and selection operator (LASSO) regression. We incorporated factors of p < 0.05 from univariate logistic regression into multivariable logistic regression with the forward LR (forward stepwise regression based on maximum likelihood estimation) method. Thereafter, independent predicting factors obtained for HTG-SAP were adopted for building a regression model. and visualized with Nomogram to predict independent factors of SAP in HTG-AP cases. To mitigate overfitting issues and internally validate performance, we employed repeated 10-fold cross-validation with 50 iterations. Nomogram was used for visualization to predict factors independently predicting HTG-AP patients. The marker and prediction model accuracy for predicting HTG-SAP was analyzed by receiver operator characteristic (ROC) curves. These models were also compared with each scoring system. Differences between the ideal model and Nomogram were analyzed by calibration curves, whereas model utility was evaluated by clinical decision curves.

Result

Between January 2020 and December 2024, there were 378 patients with HTG-AP. After excluding 134 cases, 244 cases were enrolled into this work and classified as HTG-SAP (N = 44) or HTG-NSAP (N = 200) group. Our study screening flowchart is presented in Fig. 1.

Fig. 1.

Flowchart. Retrospective selection process of patients

Basic patient characteristics

Basic features of 244 cases can be observed from Table 1. Age, gender, WBC, HB, PLT, TBIL, ALT, AST, Amylase, TG, TC, Normal drinking history, hypertension, coronary heart disease, SATI, and SMI revealed no obvious differences in both groups (P > 0.05; Table 1). Scr, LDH, GLU, Diabetes mellitus, BMI, CRP, TyG, fatty liver disease, ascites, VATI, Ranson scores, BISAP scores, MCTSI scores of HTG-SAP group increased, while Ca²⁺ and Alb decreased in HTG-NSAP group, with significant differences (P < 0.05; Table 1).

Table 1.

Clinicopathologic characteristics of 244 patients with HTG-AP

Characteristics	All patients (n = 244)	HTG-NSAP group (n = 200)	HTG-SAP group (n = 44)	P-value
Age (years)	38.66 ± 9.89	38.74 ± 9.78	38.3 ± 10.45	0.788
Sex, male	185(75.8%)	150(75%)	35(79.5%)	0.524
WBC(×109/L)	13.18 ± 4.76	13.08 ± 4.49	13.64 ± 5.88	0.477
HB (g/L)	153.33 ± 21.93	154.17 ± 21.08	149.52 ± 25.36	0.204
PLT (×109/L)	235(196,289.5)	235(197.25,291)	239(182.5,286)	0.463
Scr (µmol/L)	55(44,70.75)	52.50(42.00,64)	75(64,163.75)	< 0.001
TBIL (µmol/L)	14.35(8.6,20.38)	14.6(8.43,19.75)	13.45(9.15,24.83)	0.451
ALB (g/L)	42.87 ± 6.79	44.51 ± 5.23	35.42 ± 8.04	< 0.001
LDH (U/L)	268.5(201.3,408.3)	249.5(196.0,303.3)	496.5(381.3,615.3)	< 0.001
GLU (mmol/L)	10.65(6.8,14.68)	9.55(6.4,14.4)	13.15(9.35,16.8)	0.001
ALT (U/L)	26(16,41.75)	26(16.25,40.75)	28(16,45.75)	0.374
AST (U/L)	23.5(17,33.75)	23.5(17,32)	24(18.25,59.75)	0.104
Ca2+(mmol/L)	2.19 ± 0.26	2.26 ± 0.18	1.90 ± 0.38	< 0.001
Amylase (U/L)	220(85.25,408.75)	214.5(73.5,394)	224(127.5,568.75)	0.118
TG (mmol/L)	15.3(11.09,21.76)	15.36(11.02,21.43)	15.09(11.09,25.12)	0.401
TC (mmol/L)	9.45(6.73,13.28)	9.3(6.62,12.55)	10.6(7.7,14.9)	0.102
CRP (mg/L)	45.6(7.3,163.8)	25.60(4.75,123.15)	215.40(82.63,345.93)	< 0.001
TyG	11.78 ± 0.75	11.71 ± 0.71	12.10 ± 0.83	0.002
Normal drinking history	103(42.2%)	82(41%)	21(47.7%)	0.413
Diabetes mellitus, n (%)	57(23.4%)	41(20.5%)	16(36.4%)	0.024
Hypertension	52(21.3%)	41(20.05%)	11(25%)	0.509
CHD	5(2%)	4(2%)	1(2.3%)	0.908
Fatty liver disease	121(49.6%)	90(45%)	31(70.5%)	0.002
Ascites	59(24.2%)	37(18.5%)	22(50%)	< 0.001
MCTSI	2(2,4)	2(2,4)	6(2,7.5)	< 0.001
BISAP	1(0,2)	1(0,1)	2(1,3)	< 0.001
RANSON	2(1,3)	1(1,2)	4(3,5)	< 0.001
BMI (kg/m2)	28.51 ± 4.47	28.09 ± 4.12	30.39 ± 5.46	0.002
SATI (cm/m2)	63.46 ± 27.81	61.97 ± 24.88	70.20 ± 38.09	0.075
VATI (cm/m2)	63.87 ± 20.58	61.49 ± 18.24	74.69 ± 26.60	< 0.001
SMI (cm/m2)	57.13 ± 11.61	56.67 ± 10.39	59.22 ± 16.03	0.188

WBC White blood cell count, HB Hemoglobin, PLT Platelet count, Scr Serum creatinine, TBIL Total bilirubin, ALB Albumin, LDH Lactate dehydrogenase, GLU Glucose, ALT Alanine aminotransferase, AST Aspartate aminotransfer­ase, TG Triglycerides, TC Total cholesterol, CRP C-reactive protein, TyG Triglyceride and glucose, CHD Coronary heart disease, BMI Body mass index, SATI Subcutaneous adipose tissue index, VATI Visceral adipose tissue index, SMI Skeletal muscle index

Screening for independent predictors of HTG-SAP

Univariate analysis of clinical and laboratory indicators P < 0.05 were enrolled into LASSO regression for creating 1,000 models, and those having simple Lambda values, small errors, and relatively simple combinations were selected as references. Nine indices (Scr, LDH, CRP, TyG, fatty liver disease, ascites, VATI, Ca²⁺, ALB) were screened (Fig. 2) and subsequently incorporated for multivariable logistic regression, with LDH, SCR, VATI, ALB, and TyG as the independent predictors and LDH, Scr, VATI, and TyG were risk factors, and ALB was a protective factor. p-value obtained from Hosmer-Lemeshow test was 0.838, illustrating that this model was valid. Multivariable regression results identified VATI (odds ratio [OR], 1.050; 95% confidence interval [CI], 1.018 ~ 1.083; P = 0.002), Scr (OR, 1.062; 95% CI, 1.030 ~ 1.094; P < 0.001), LDH (OR, 1.004; 95% CI, 1.001 ~ 1.006; P = 0.004), and TyG (OR,2.721;95% CI,1.250 ~ 5.926; P = 0.012) as factors independently predicting HTG-SAP occurrence, whereas ALB (OR, 0.758; 95% CI, 0.684 ~ 0.840; P < 0.001) as independent protective factors for HTG-SAP Attack (Table 2).

Fig. 2.

Selection of predictive factors using the least absolute shrinkage and selection operator logistic regression algorithm. A LASSO coefficient profiles of the 12 candidate variables. B The best value was determined by the two dashed vertical lines drawn according to the minimum mean-square error criterion (left dashed line) and the standard error criterion (right dashed line). In the present study, nine predictors were selected according to the minimum mean-square error criterion (λ = 0.0113477)

Table 2.

Multivariable logistic regression analysis results of independent predictors

Predictive factor	Regression coefficient	Wald	P-value	OR	95%CI
VATI	0.048	9.501	0.002	1.050	1.018 ~ 1.083
LDH	0.004	8.483	0.004	1.004	1.001 ~ 1.006
Scr	0.060	14.872	< 0.001	1.062	1.030 ~ 1.094
TyG	1.001	6.357	0.012	2.721	1.250 ~ 5.926
ALB	−0.277	28.004	< 0.001	0.758	0.684 ~ 0.840

VATI Visceral adipose tissue index, LDH Lactate dehydrogenase, Scr Serum creatinine, TyG Triglyceride and glucose, ALB Albumin

Construction of a new HTG-SAP forecasting model

Based on multivariable logistic regression results, our HTG-SAP prediction model was constructed using LDH, Scr, VATI, ALB, and TyG; the model was visualized using the R software, and a nomogram was obtained (Fig. 3). Based on the constructed clinical prediction model nomogram, points were assigned to each patient based on the presence of these risk factors. The sum of these points (“total points”) was converted to the probability of HTG-SAP. One randomized column of patients was predicted to have a 92.7% probability of SAP. Patient condition could be critical clinically, with CT manifestations of pancreatic necrosis with renal insufficiency, and this patient was admitted to intensive care unit due to SAP (Fig. 4).

Fig. 3.

The nomogram of the prediction model for predicting a patient’s first HTG-SAP episode

Fig. 4.

Based on the predictive model Nomogram, the density distributions and predictive probabilities of the predictors and total scores of all patients in the database can be obtained. A straight line perpendicular to the corresponding axis is plotted for each risk factor until it reaches the top line labeled “POINTS”. The scores for all risk factors are summed and a descending line is drawn from the axis labeled “TOTAL POINTS” (total score) until the diagnostic probability is intercepted. Representative case example of a real patient predicted by our model. The patient was randomly selected from the validation cohort to illustrate the clinical applicability of the scoring system. This 60-year-old male patient scored 135 points (diagnosis probability: 92.7%), indicating high-risk hypertriglyceridemic severe acute pancreatitis (HTG-SAP). Clinically, the patient presented with critical conditions, including pancreatic necrosis and renal insufficiency on CT imaging, and required intensive care unit (ICU) admission. The case demonstrates the model’s ability to identify severe cases with high confidence

The AUCs for LDH, Scr, VATI, ALB, TyG, and the HTG-SAP prediction model were 0.825, 0.821, 0.637, 0.808, 0.635, and 0.966, respectively. These AUCs were further compared with the HTG-SAP prediction model, and the differences were significant (P < 0.01). ROC curves were used to calculate the cut-off value, sensitivity, specificity, PPV and NPV of each independent predictor and the HTG-SAP prediction model (Table 3; Fig. 5).

Table 3.

Comparison of independent predictors and predictive models for HTG-SAP

Index	AUC	Cut-off value	Sensitivity	Specificity	PPV	NPV	p-value
LDH	0.825	343	0.818	0.815	0.493	0.953	< 0.001
Scr	0.821	58.5	0.864	0.665	0.362	0.957	< 0.001
VATI	0.637	75.02	0.5	0.77	0.324	0.875	0.004
ALB	0.808	37.05	0.636	0.910	0.609	0.919	< 0.001
TyG	0.635	11.88	0.614	0.635	0.270	0.882	0.005
HTG-SAP model	0.966	0.163	0.886	0.910	0.684	0.973	< 0.001

LDH Lactate dehydrogenase, Scr Serum creatinine, VATI Visceral adipose tissue index, ALB Albumin, TyG Triglyceride and glucose, HTG-SAP Hypertriglyceridemia severe acute pancreatitis, AUC Area under the curve, PPV Positive predictive value, NPV Negative predictive value

Fig. 5.

Receiver operator characteristic curve of independent predictors and predictive models for HTG-SAP. The AUC of the HTG-SAP model, LDH, Scr, VATI, ALB, TyG, and the HTG-SAP prediction model were 0.966, 0.825, 0.821, 0.637, 0.808, 0.635

Nomogram internal validation

To ensure the robustness of the nomogram, rigorous internal validation was performed using repeated 10-fold cross-validation with 50 repetitions. This approach mitigates overfitting and provides a reliable estimate of the model’s generalizability. The validation results demonstrated strong predictive performance, with an average area under the receiver operating characteristic curve (AUC) of 0.891, indicating excellent discriminatory power. Additionally, the model achieved a Cohen’s kappa statistic of 0.558551, reflecting moderate to substantial agreement beyond chance between the predicted and observed outcomes. These metrics collectively affirm the model’s stability and clinical applicability for early severity stratification in HTG-AP patients.

Nomogram calibration verification

The good fit of our model was revealed by HL goodness-of-fit test (χ2 = 4.209, P = 0.838); and model calibration was assessed by the calibration curve. From Fig. 6, “apparent” stands for initial curve, “ideal” indicates an optimal standard curve, whereas “bias-corrected” is a calibration curve. For training group, its Brier score was 0.051, suggesting that our nomogram predicted HTG-SAP.

Fig. 6.

Calibration curve of hypertriglyceridemia severe acute pancreatitis (HTG-SAP) model. The x-axis represents the predicted probability of HTG-SAP calculated according to the model, while the y-axis exhibits the actual probability of HTG-SAP. The apparent calibration curve (dotted line) indicates the model performance in the original data, while the bias-corrected curve (solid line) represents the model performance after correction for optimism using 1000 bootstrap resamples. A perfect prediction would fall on the 45-degree (dashed) reference line

Model clinical utility assessment

DCA for HTG-SAP model were plotted, and at the threshold probability > 0, the model curve was above both extreme value lines, suggesting that timely clinical intervention is beneficial and of good clinical value if this model predicts a HTG-SAP risk in patients (Fig. 7).

Fig. 7.

Decision curve analysis (DCA) evaluating the clinical utility of the HTG-SAP prediction model. The y-axis represents net benefit, quantifying the clinical value of the model, while the x-axis shows threshold probabilities corresponding to clinical decision thresholds. The black horizontal Line represents the assumption that no patients with HTG-SAP require treatment, whereas the solid gray Line assumes that all patients require treatment. The red curve reflects the performance of our HTG-SAP model. When the threshold probability exceeds 0%, the model yields a higher net benefit than both extreme strategies (black/gray lines), demonstrating its clinical utility in guiding interventions for patients at risk of HTG-SAP

Comparison of clinical prediction models and scoring systems regarding the ROC

As revealed by ROC curves for HTG-SAP prediction model, MCTSI, BISAP, and Ranson scores, this model outperformed MCTSI, BISAP, and Ranson scores in predicting the HTG-SAP progression in cases. The area under curve (AUC) values for 4 models were 0.966 (95% CI, 0.9429–0.9885), 0.821 (95% CI, 0.7595–0.8832), 0.777 (95% CI, 0.7049–0.8494), and 0.838 (95% CI, 0.7779–0.8977), separately (Fig. 8).

Fig. 8.

ROC curves of HTG-SAP, MCTSI, BISAP, and Ranson’s score. The AUCs of the HTG-SAP model, MCTSI score, BISAP score, and Ranson’s score were 0.966, 0.777, 0.821, and 0.838

Discussion

Prior studies of metabolic markers for the prediction of AP severity—including the VATI model developed by Gu et al., which achieved an AUC of 0.878 for early severity prediction [17]; the TyG approach reported by Xinyu et al. (AUC = 0.882) [18]; and the BMI-enhanced BISAP score proposed by Guzman et al. with 92.3% accuracy [19]—have established a novel integrated model specifically for the prediction of HTG-AP severity. We identified VATI, TyG, Scr, ALB, and LDH as significant predictors of HTG-AP severity and prognostic outcomes and constructed a comprehensive prediction model that demonstrated superior discriminative capacity (AUC = 0.966). Through nomogram visualization and risk stratification, this tool enables timely classification of disease progression severity, facilitating individualized treatment for patients at distinct clinical stages.

Several studies have shown that increased obesity, especially visceral adiposity, plays a key role in predicting AP severity as well as prognosis, and our study demonstrated that VATI independently predicted HTG-SAP, which conforms to prior studies [14, 20, 21]. There may be several reasons why increased visceral adiposity aggravates the severity of HTG-AP. First, excessive hydrolysis of VAT in AP releases unsaturated fatty acids (UFAS) at large amounts, constituting most of lipid mediators of SAP, and the excess UFAS aggravate inflammatory response, OF and pancreatic necrosis, resulting in SAP [22]. Animal studies have shown that inhibition of lipid hydrolysis prevents organ damage and reduces mortality in obese mice with pancreatitis [23]. Second, VAT is metabolically active and secretes various adipokines, including leptin, Resistin and Adiponectin, which may be related to AP occurrence and development through regulating inflammation and oxidative stress [24, 25]. Previous studies have shown that the more VAT, the higher the serum IL-6, CRP and TNF-α levels, and the greater inflammatory response [22, 26, 27]. Finally, increased VAT causes increased abdominal pressure and compression of visceral tissues and organs, resulting in decreased perfusion of visceral tissues and organs, decreased cardiac output, respiratory disturbances, and decreased glomerular filtration rate, which can easily lead to multiorgan dysfunction [28]. Therefore, VATI is a favorable indicator for predicting HTG-AP severity.

Our study found that TyG independently predicted HTG-SAP. TyG, a substitute of IR, is involved in AP development [12, 29, 30]. IR, as the chronic low-grade inflammation, is featured by increased pro-inflammatory factor levels in circulation that can exacerbate the inflammatory response in AP, resulting in local complications, OF or systemic inflammatory response syndrome (SIRS) [31, 32]. Interestingly, free fatty acids can be increasingly produced by visceral adipose tissue, and excess free fatty acids enhance the expression of IKK and JNK pathways, increasing IR development [33]. Besides, IR enhances MCP-1 generation within adipocytes, thereby attracting monocytes while activating pro-inflammatory macrophages, exacerbating inflammation in adipose tissue [34, 35]. However, the relationship between VAT and TyG in HTG-SAP still needs to be further explored.

In the present study, Scr and LDH were found to be significantly increased during early HTG-SAP, and Scr detection is previously suggested as a key predictor for assessing AP severity [36]. In AP, inflammatory mediator release promotes vascular permeability, blood volume decreases, and consequently renal perfusion decreases, and Scr accumulates in the body along with other metabolic waste products [37, 38]. LDH is mainly distributed within the heart, skeletal muscle, liver and kidneys. SAP patients may develop cardiac, renal and pulmonary insufficiency [39]; therefore, their LDH levels are elevated. Studies have shown that early alterations of LDH contents, particularly within initial 24 h post-admission, are predictors of HTG-AP severity, conforming to our results [40].

ALB can be produced in the liver and catabolized within many organs. ALB is suggested to be related to AP severity and prognosis [41, 42], but few studies have used serum ALB to predict HTG-AP severity. We found that ALB contents showed negative correlation with HTG-AP severity. In the AP progression process, elastase and trypsin injure vascular endothelial cells, leading to elevated vascular permeability as well as the later ALB-rich plasma penetration into the tissue space [43]. Additionally, hypoperfusion reduces the liver’s ability to synthesize ALB, and when combined with fasting, this contributes to decreased ALB levels [44]. Furthermore, pancreatic lipase leakage may also lead to lipolysis and elevated levels of UFAS, which bind to ALB, and consequently, a decrease in ALB levels occurs [45]. Therefore, early serum ALB supplementation during HTG-SAP may reduce fatty acid toxicity and prevent disease progression.

Study strengths and limitations

The advantages of the predictive model we developed for HTG-AP severity are firstly that it includes clinical tests and metabolism-related indicators and requires completion of an abdominal CT examination, an approach routinely performed in hospitals, making it more popular to apply our HTG-SAP model in clinic. Second, few research has analyzed the relation of TyG index and VATI with AP. In this study, TyG index, VATI and HTG-AP severity were closely related to each other, which provided a new prediction direction for HTG-SAP prediction. Consequently, clinicians must focus on HTG-AP cases developing metabolic abnormality; health education and intervention are needed for preventing disease progression and evolution to HTG-SAP.

Certain limitations must be noted in this study; firstly, due to the retrospective nature, selection bias may be unavoidable. Secondly, the small sample size and low HTG-SAP incidence can limit our statistical power for analyzing indicators related to HTG-AP severity. Furthermore, while our nomogram integrates the CT-based visceral adiposity tissue index (VATI) to increase precision, we acknowledge that specialized quantification software (e.g., SliceOmatic) is not routinely available in all clinical settings. Importantly, doctors can still use our nomogram with basic CT measurements (e.g., visually estimating VAT or simple manual measurements) when advanced tools are unavailable. If future hospital systems integrate quantitative CT analysis, our model will automatically benefit from improved precision. Finally, the model was only analyzed in a single center, and its broad applicability is not yet known; therefore, additional prospective, multicenter, large-sample studies must be carried out for validating our predictive model and broader model applicability.

Conclusions

Collectively, LDH, Scr, VATI, TyG, ALB are independent predictors of HTG-AP severity, and establish an early visualization model of HTG-AP severity, which has high accuracy, sensitivity specificity, and good clinical utility. It can help to adopt appropriate pretreatment for patients with different conditions and reduce the waste of medical resources and the delay of interventions for critically ill patients.

Authors’ contributions

HQ and QT equally contributed to the writing – review & editing, conceptualization; HZ contributed to the data curation, supervision; JT contributed to the data curation, methodology; CW contributed to the supervision, visualization; XZ contributed to the investigation, resources; YL contributed to the project administration, writing – original draft, supervision; ZW contributed to the methodology, project administration, writing – original draft; All authors critically revised the manuscript, agree to be fully accountable for ensuring the integrity and accuracy of the work, and read and approved the final manuscript.

Funding

No funding was received.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.