Development and validation of an early prediction model for hypertriglyceridaemic severe acute pancreatitis: a retrospective study

Introduction

Acute pancreatitis (AP) is a critical inflammatory condition of the pancreas characterized by sudden onset and intense abdominal pain, and is the most prevalent gastrointestinal condition leading to hospitalization (Peery et al., 2012). AP occurs when pancreatic enzymes are activated prematurely, resulting in the digestion of pancreatic tissue and inflammation or necrosis in the pancreas and nearby tissues (Szatmary et al., 2022). According to the Atlanta classification, AP is categorized into mild, moderately severe, and severe acute pancreatitis (SAP) (Banks et al., 2013). Although most patients with AP have a mild and self-limiting form that can be managed with simple fluid replacement therapy, approximately 12–20% develop SAP, which progresses to local and/or systemic complications, including systemic inflammatory response syndrome (SIRS) and organ failure lasting more than 48 h (Banks et al., 2013). The mortality rate for patients with AP is approximately 2% (Gapp et al., 2019; Lee & Papachristou, 2019); however, this rate may increase to 28.3–52% in patients with SAP (Petrov et al., 2010; Párniczky et al., 2016; Schepers et al., 2019).

In recent years, the incidence of hypertriglyceridaemic acute pancreatitis (HTG-AP) has increased significantly (Lai et al., 2024). A retrospective study from China revealed that the incidence of HTG-AP increased approximately 2.6-fold over a decade, increasing from 8.4% in April 2012–March 2013 to 22.3% in April 2020–March 2021 (Lin et al., 2022). Several studies have suggested that patients with HTG-AP tend to experience more severe disease progression and have a greater risk of persistent organ failure than those with other aetiologies do (Lai et al., 2024; Zhou et al., 2024). The mechanism of HTG-AP is characteristically different from those of other aetiologies; the excess free fatty acids (FFAs) caused by HTG and elevated chylomicrons increase plasma viscosity, resulting in ischaemia in pancreatic tissue and inflammation in the pancreas (Kiss et al., 2023). Pancreatic steatosis (PS) is defined as excessive lipid deposition in the pancreas, resulting from diverse causes, including aging and metabolic syndrome-related factors such as diabetes, dyslipidaemia, and obesity (Wagner et al., 2022). Therefore, PS might be a good predictor of hypertriglyceridaemic severe acute pancreatitis (HTG-SAP). In a matched case–control study of 50 AP patients, Ko et al. (2022) reported that an 1% increase in PS was significantly associated with a 30% greater chance of experiencing a first attack of AP. A recent prospective cohort study of 42,599 participants revealed that PS was associated with a greater risk of AP (HR 3.982, 95% CI [2.192–7.234]) and was an independent risk factor for AP (P = 0.001) (Dong et al., 2024).

Currently, AP scores are frequently used to predict severity and guide early management in clinical settings, primarily including the Sequential Organ Failure Assessment (SOFA), the Modified CT Severity Index (MCTSI), and the Bedside Index of Severity in Acute Pancreatitis (BISAP), among others (Colvin et al., 2020). Nonetheless, a perfect multifactorial scoring system or biochemical marker for the early evaluation of AP severity has not yet been established (Silva-Vaz et al., 2020). Combining multiple biomarkers for prediction is a straightforward and effective method, especially when new indicators are introduced. However, accurate prediction of HTG-SAP remains challenging (Gliem et al., 2021), and few studies in the literature have explored the relationship between PS and HTG-AP severity. Therefore, we aimed to retrospectively analyse the admission data of patients with HTG-AP within 24 h to further investigate the significance of PS and establish a visual model for the early prediction of HTG-SAP, which can support clinical decision-making to reduce morbidity and mortality.

Methods

Patients

The inclusion criterion was patients aged ≥18 years with a first-time diagnosis of HTG-AP. The exclusion criteria were as follows: age less than 18 years (n = 0), hospital stay less than 48 h (n = 10), multiple admissions (n = 44), serious chronic organ failure (n = 28), biliary pancreatitis (n = 327), alcoholic pancreatitis and chronic pancreatitis (n = 256), other causes of pancreatitis (including pancreatic cancer) (n = 185), and missing information (n = 92). Overall, 346 patients with HTG-AP were included in the study. These patients were classified into two groups: those with HTG-SAP (n = 94) and those with hypertriglyceridaemic nonsevere acute pancreatitis (HTG-NSAP; n = 252). The HTG-NSAP group included patients with mild and moderately severe AP (Banks et al., 2013) (Fig. 1).

Results

Predictors of HTG-SAP and nomogram development

Twenty-seven variables with significant differences between the two groups in the univariate analysis were selected, which included T, HR, RR, De, CRP level, HCT, APTT, PT, INR, FIB level, D-D level, LP level S, AMY level, PCT level, TB level, BUN level, Cr level, GLU level, LAC level, L count, ALB level, Ca²⁺ level, pH, PaO2/FiO2, MCT_P, and the presence of PN, PE, AE and PS (all P < 0.05). Ultimately, eight variables, including RR, D-D level, BUN level, Ca²⁺ level, pH level, and the presence of PN, PE and PS, were identified as independent risk factors for HTG-SAP by multivariate binary logistic regression analysis. Among these variables, RR, D-D level, BUN level, and the presence of PN, PE, and PS were independent risk factors, whereas the pH and Ca²⁺ level were independent protective factors (Table 3).

Table 3:

Results of the multivariate logistic regression analysis of the independent predictors.

Eight predictors were identified as independent risk factors for HTG-SAP by multivariate binary logistic regression analysis (P < 0.05).

Predictor	β	Wald	P value	OR	95% CI
RR	0.163	6.384	0.012	1.177	1.037∼1.336
D-D	0.517	13.432	<0.001	1.677	1.272∼2.212
BUN	00.065	6.503	0.011	1.067	1.015∼1.122
Ca2+	−0.921	3.955	0.047	0.398	0.161∼0.987
PH	−6.943	8.646	0.003	0.001	0.000∼0.099
PN	1.840	18.415	<0.001	6.298	2.718∼14.596
PE	1.670	17.571	<0.001	5.313	2.433∼11.600
PS	1.546	14.188	<0.001	4.693	2.099∼10.491
Constant	45.130	6.717	0.010	–	–

DOI: 10.7717/peerj.20607/table-3

Notes:

Abbreviations

β

Regression coefficient

RR

Respiratory rate

D-D

D-dimer

BUN

Blood urea nitrogen

Ca²⁺

Serum calcium

PH

Potential of hydrogen

PN

pancreatic necrosis

PE

Pleural effusion

PS

Pancreatic steatosis

In accordance with the results of the multivariate analysis, the HTG-SAP prediction model was established as a nomogram based on RR, D-D level, BUN level, Ca²⁺ level, pH, and the presence of PN, PE and PS, as shown in Fig. 2.

Validation of the HTG-SAP prediction model

The ROC curves revealed that the ability of the new model to predict the progression of patients to HTG-SAP was better than those of the MCTSI, SOFA, and BISAP scores. The AUC values of the MCTSI, SOFA, BISAP scores and the HTG-SAP prediction model were 0.832, 0.784, 0.782 and 0.937, respectively, and the differences were all significant (P < 0.001) (Table 4, Fig. 3). The results of the ROC analyses indicated that the nomogram strongly predicted patients with HTG-SAP in the training (AUC = 0.937) and validation (AUC = 0.922) cohorts (Fig. 4). The calibration curve demonstrated that the predicted probabilities closely matched the observed outcomes in both the training and validation cohorts, indicating successful calibration (Fig. 5). According to the DCA, the nomogram provided a superior overall net benefit across a broad range of threshold probabilities, highlighting its strong potential for clinical utility (Fig. 5).

Table 4:

Comparison of the AUCs between the Pre-HTG-SAP and prediction systems.

Model	P value (Model VS. pre-HTG-SAP)
MCTSI	<0.001
SOFA	<0.001
BISAP	<0.001

DOI: 10.7717/peerj.20607/table-4

Notes:

We used a web tool to perform the comparison between two AUCs (nonparametric Wilcoxon statistic) http://vassarstats.net/roc_comp.html).

Abbreviations

AUC

Area under the curve

MCTSI

Modified CT Severity Index

BISAP

Bedside Index for Severity in Acute Pancreatitis

SOFA

Sequential Organ Failure Assessment

Pre-HTG-SAP

Predictive model for hypertriglyceridaemic severe acute pancreatitis

Discussion

In this study, we found that the incidence of the first occurrence of HTG-AP accounted for 26.9% of all cases of AP, which is higher than that of alcohol-induced AP, previously reported as the second leading cause of AP (Yu et al., 2019; Lin et al., 2022). A recent retrospective study including 10,071 cases of AP reported that hyperlipidaemia (25.3%) was the second most common aetiology after cholelithiasis (Lai et al., 2024). In Taiwan area, the incidence of HTG-AP was reported to range from 6.3–12.3% (Olesen et al., 2021). Additionally, the incidence rate of HTG-SAP among patients with HTG-AP was reported to be 27.2%, which is higher than the previously reported incidence rate of SAP among all patients with AP (10–20%) (Banks et al., 2013). This may be related to the unique pathogenesis and pathophysiological processes of HTG-AP. We also revealed that PS is an independent predictor of HTG-SAP, and the incidence of PS in patients with HTG-AP is 44.2%, which is higher than the rates reported in the general population (16–35%) (Wong et al., 2014; Lesmana et al., 2015; Koç & Taydaş, 2020). Furthermore, we developed a simplified and practical nomogram model to predict HTG-SAP with high accuracy (AUC= 0.937) based on independent predictors, including the RR, D-D level, BUN level, Ca²⁺ level, pH, and the presence of PN, PE and PS, within 24 h of admission. To validate the consistency and clinical applicability of our HTG-SAP prediction model, we compared our model with three well-established AP prognostic scoring systems. The results confirmed the accuracy of the model in predicting the occurrence of HTG-SAP.

The pathogenesis of HTG-AP is currently recognized as follows: (1) Excessive TGs in the blood are broken down into FFAs by pancreatic lipase. When the concentration of FFAs exceeds the binding capacity of ALB, protein kinases are activated, resulting in autodigestion of pancreatic cells. (2) FFAs trigger inflammatory responses, causing cytotoxicity and damage to pancreatic capillaries. (3) As blood viscosity increases, serum lipoprotein particles accumulate, potentially leading to embolization of pancreatic vessels. This exacerbates pancreatic ischaemia and acidosis, further enhancing the lipotoxic effects of FFAs (Kiss et al., 2023). Our study revealed that the presence of PS is a novel independent predictor of HTG-SAP. PS indicates the accumulation of pancreatic fat cells (Kiss et al., 2023). Lipid accumulation triggers cellular stress responses, lipotoxic effects, and dysfunctional intercellular signalling, leading to inflammation within the pancreatic microenvironment (Petrov, 2023), which forms the foundation for the onset and progression of HTG-AP. PS acts as a mechanistic pathological driver for most nongenetic pancreatic disorders. Dysfunction of intrapancreatic adipocytes and associated lipotoxicity can surpass an individual’s threshold, triggering inflammation within the pancreatic microenvironment and leading to cellular injury. The resulting inflammatory milieu is intensified by various forms of cellular injury and death, such as ferroptosis and pyroptosis. A critical mechanism in initiating acinar cell injury is the sustained aberrant interplay between lipid droplets and the endoplasmic reticulum. This interaction leads to the release of nonesterified fatty acids, cytosolic and mitochondrial calcium overload, mitochondrial dysfunction, and a compromised autophagic response, which collectively play major roles in the onset of acute pancreatitis (Petrov, 2023). The notably higher severity observed in HTG-AP can be primarily attributed to the lipotoxicity mediated by excessive FFAs. This concept is robustly supported by experimental evidence showing that the hydrolysis of triglycerides within the pancreas leads to a direct, dose-dependent toxic insult, resulting in oedema, haemorrhage and profound acinar cell injury (Saharia et al., 1977; Durgampudi et al., 2014; De Oliveira et al., 2019). This primary lipotoxic assault is further compounded by a secondary phenomenon of plasma hyperviscosity (Saharia et al., 1977). The elevated triglyceride levels increase blood viscosity, impairing pancreatic microcirculation and leading to ischaemia. The resulting cellular acidosis then creates a permissive environment for the intracellular activation of trypsinogen, thereby synergistically amplifying the initial inflammatory and autodigestive processes (Sakorafas, Tsiotos & Sarr, 2000). Furthermore, PS, which is characterized by the accumulation of intrapancreatic adipocytes, is closely linked to more severe presentations of acute pancreatitis (Petrov et al., 2010). The pathophysiological role of PS is likely mediated through adipocyte dysfunction and consequent lipotoxicity, driving a more intense inflammatory response and tissue damage. Collectively, the mechanisms of lipotoxicity and hyperviscosity in HTG-AP, often exacerbated by the presence of PS, provide a compelling explanation for the aggressive clinical course observed in these patients. A Mendelian randomization study revealed that genetically predicted PS was significantly associated with AP (OR per 1-SD increase: 1.40 95% CI [1.12–1.76], p = 0.0032), supporting a causal role of PS in pancreatitis (Yamazaki et al., 2024). Another study by Xie et al. (2019) which included 246 patients with AP, further demonstrated that PS was independently correlated with severity, mortality, and systemic complications in patients with AP, which is similar to our results. However, to date, no studies have investigated the relationship between PS and the severity of HTG-AP in clinical practice. Our research addresses this gap in the literature and indicates that PS is an independent predictor of HTG-SAP.

According to the RAC (Banks et al., 2013), organ failure is determined by evaluating three organ systems during the progression of AP: the respiratory, cardiovascular, and renal systems. In our study, RR, BUN level and presence of PE were confirmed to be independent predictors of HTG-SAP. The lipotoxicity from FFAs generated from TGs is among the main theories explaining how HTG-AP is aggravated (Kiss et al., 2023). These toxic structures from FFAs can damage platelets, the vascular endothelium, and acinar cells, leading to ischaemia and acidosis; acidosis further enhances FFA toxicity by activating trypsinogen. This direct damage can lead to pancreatic duct rupture and increased capillary permeability throughout the body (Kiss et al., 2023). Additionally, FFAs may directly stimulate the production of inflammatory mediators, including TNF-alpha, IL-6, and IL-10, thereby intensifying the local inflammatory response and initiating the inflammatory cascade (De Pretis, Amodio & Frulloni, 2018). These inflammatory mediators travel through the lymphatic and systemic circulation, where they reach the liver, lungs, heart, kidneys, and gastrointestinal tract, resulting in SIRS (Szatmary et al., 2022). Based on the above pathophysiological processes, patients with HTG-AP may develop lung injury, kidney injury, hypoproteinaemia and third-space fluid accumulation. These processes are manifested by an increased RR, an elevated BUN level, and the presence of PE. A retrospective study of 326 elderly patients indicated that RR is an important predictor of the survival of patients with SAP (Zhu et al., 2024). The increase in BUN levels in patients with HTG-AP can be explained by two main mechanisms of acute kidney injury (Nassar & Qunibi, 2019): (1) the reduction in intravascular volume due to interstitial fluid leakage and (2) the direct impact of activated pancreatic enzymes, along with inflammatory mediators and cytokines. A prospective observational study of 410 patients with AP revealed that an increase in BUN was an accurate predictor of mortality (AUC: 0.842) and persistent multiorgan failure (AUC: 0.828) (Pando et al., 2021). However, Lin et al. (2017) reported that determining BUN levels 24 h after hospital admission is highly accurate for predicting SAP. In contrast, our study indicated that the initial BUN level within 24 h after admission is a good predictor of HTG-SAP. The difference in these outcomes may be attributed to the pathogenesis of HTG-AP, which needs to be confirmed by further clinical studies. PE is recognized as a strong individual predictor of SAP according to the RAC (Banks et al., 2013). In a retrospective study of 465 patients with AP from three acute pancreatitis centres, Yan et al. (2021) reported that PE volume can be a reliable predictor of SAP (p < 0.05), which is consistent with our results. Consequently, the RR, BUN level, and presence of PE are important for predicting the occurrence of HTG-SAP.

Early local complications, including acute peripancreatic fluid collection (APFC) and acute necrotic collection (ANC), are also components used to assess the severity of AP (Banks et al., 2013). Our study revealed that D-D level, pH, the presence of PN, and Ca²⁺ level are individual predictors of HTG-SAP. Another potential mechanism underlying how HTG-AP is aggravated involves microcirculatory disturbances, which can result in local ischaemia and damage to acinar cells (Kiss et al., 2023). An elevated concentration of chylomicrons can lead to an excess of FFAs, and capillary leakage can directly and significantly increase plasma viscosity; this elevation in viscosity may result in impaired microcirculation, capillary blockage, ischaemia, haemostatic dysfunction, and acidosis, ultimately inducing ischaemic necrosis of pancreatic tissue (Kiss et al., 2023). Dysfunction in the coagulation system may result in microcirculatory collapse and failure across multiple organs, both within and beyond the pancreas, thereby increasing the mortality rate associated with AP (Gui et al., 2023). A retrospective study of 189 patients with HTG-AP revealed that the AUC of the D-D level for the prediction of HTG-SAP was 0.753 95% CI [0.680–0.826]; P < 0.001); therefor, the D-D level may serve as a sensitive early prognostic indicator for HTG-SAP (Gou, Yao & Cao, 2023). Metabolic acidosis (plasma pH less than 7.35) can occur in patients with AP due to several factors, including the buildup of lactic acid caused by shock, kidney dysfunction, or, in the later stages of the disease, the loss of bicarbonate-rich pancreatic secretions due to pancreatic duct disruption. A meta-analysis including thirteen studies with 2,311 patients with AP indicated that patients with AP who exhibited reduced blood pH or base excess experienced significantly greater mortality rates and increased severity scores (Rumbus et al., 2018). In a prospective study, Sharma et al. (2014) reported that patients with AP patients with a pH less than 7.35 presented a greater incidence of organ failure and required more frequent interventions, indicating that pH is a useful early marker for predicting adverse outcomes in patients with AP. PN is a consistent prognostic factor in AP, serving as an indicator of local complications according to the RAC (Banks et al., 2013). In a retrospective study, Fu et al. (2021) demonstrated that the extrapancreatic necrosis volume could be a valuable indicator for predicting poor outcomes in patients with AP. The necrosis of pancreatic and peripancreatic fat can lead to the autodigestion of mesenteric fat. During this process, Ca²⁺ binds to the digested fat to form calcium soaps, resulting in hypocalcaemia (Kiss et al., 2023). Yu et al. (2019) reported that the Ca²⁺ levels in patients with HTG-SAP were significantly lower than those in patients with SAP of other aetiologies, with values of 1.69 (95% CI [1.46–1.91]) versus 2.1 (95% CI [1.93–2.23]) (P < 0.001), confirming that a lower calcium level is correlated with an increased risk of HTG-SAP Additionally, established prognostic system models for AP, such as the Ranson score and the Glasgow score (JSS), also include Ca²⁺ levels in their scoring systems. Therefore, the use of Ca²⁺ as a prognostic indicator in the HTG-SAP model is reasonable and logical. Overall, these four indicators reflect the pathophysiological process of pancreatic ischaemic necrosis and can effectively predict the severity of HTG-AP.

In this study, a simple and accurate nomogram model was developed and validated to predict the occurrence of HTG-SAP using variables that are easily accessible within the first 24 h of admission. The performance of nomogram was evaluated based on its discrimination, calibration, and clinical applicability, and satisfactory results were obtained upon verification. Additionally, this nomogram model provides important insights for guiding treatment decisions for high-risk patients with HTG-SAP. We compared the predictive performance of various scoring systems, including the BISAP, SOFA and MCTSI, with that of our newly developed nomogram. The AUC value of the new HTG-SAP model was 0.960 (95% CI [0.908–0.966], P < 0.000), which was greater than those of the BISAP, SOFA and MCTSI (AUCs: 0.782, 0.784, and 0.823, respectively). Another major strength of our prediction model is its rapid clinical applicability. Unlike established scores such as SOFA, which requires 24-hour data, or MCTSI, which may require imaging assessment at 48–72 h, our model utilizes variables (e.g., RR, D-dimer, BUN, Ca²⁺, pH, and initial CT) that are routinely available within hours of admission (Supplemental File 4). This allows for early risk stratification within the critical initial 24 h, enabling timely intervention for high-risk patients with HTG-SAP and potentially improving outcomes. The results demonstrate that our nomogram model has superior predictive ability and greater clinical utility compared with those of other established scoring systems.

Conclusion

In summary, our study confirmed that RR, D-D level, BUN level, Ca²⁺ level, pH, and the presence of PN, PE and PS are independent predictors of HTG-SAP. The prediction model developed based on these predictors has greater consistency and practicability in predicting HTG-SAP than the BISAP, MCTSI and SOFA scores do.

Supplemental Information