Construction of a predictive model for postoperative pancreatic fistula following pancreaticoduodenectomy and an exploration of inflammatory biomarker associations

Introduction

Pancreaticoduodenectomy (PD) represents the requisite complex surgical intervention for the curative treatment of pancreatic head lesions and serves as a therapeutic modality for chronic pancreatitis and pathologies involving the bile duct and duodenum (1,2). Despite significant advancements in surgical techniques and perioperative care, including the adoption of various anastomosis techniques (3-5), the prophylactic administration of octreotide (3), the utilization of externalized stents (6), the deployment of abdominal drainage (7), and the application of biological glues (8), postoperative outcomes remain suboptimal (9-12). Clinically relevant postoperative pancreatic fistula (CR-POPF) is among the most recalcitrant complications following PD, with an incidence rate ranging from 10% to 20% (13,14). Furthermore, CR-POPF is significantly correlated with other adverse events—such as delayed gastric emptying, infection, haemorrhage, prolonged hospitalization, and unplanned readmissions—as well as mortality, thereby imposing a substantial burden on patients (13,15). Consequently, the development of effective strategies for the prevention and early prediction of POPF following PD is a matter of clinical urgency.

Historically, the International Study Group of Pancreatic Surgery (ISGPS) established a definition for pancreatic fistula (16), stratifying the condition on the basis of severity and clinical sequelae (17). Concurrently, the Fistula risk score (FRS) has been introduced as a tool for predicting CR-POPF occurrence. The weighted FRS (ranging from 0 to 10 points) incorporates four risk elements: pancreatic parenchyma texture, pathology, pancreatic duct diameter, and intraoperative blood loss (18), with higher scores denoting more severe adverse outcomes. Risk stratification based on FRS scores has facilitated the formulation of distinct strategies to mitigate complications (3,6,10,19), undoubtedly providing valuable guidance for clinical practice (20). Moreover, several studies have sought to refine the FRS further; for example, Shubert et al. and Mungroop et al. extended its application to minimally invasive techniques (21,22), whereas McMillan et al. focused on constructing predictive models specifically for grade C fistulas (23).

However, the efficacy of the FRS in clinically reducing the incidence of POPF has not fully met expectations. Multiple studies indicate that the predictive performance of the FRS for POPF is suboptimal (21). Further analysis suggests that the selection and weighting of certain indicators within this scoring system may compromise its accuracy (24). Thus, the pursuit of novel predictive methodologies and the optimization of intervention measures remain imperative. Examples include the exploration of new predictors by the 2012 National Surgical Quality Improvement Program (NSQIP) Pancreatic Demonstration Project (PDP) (25), the implementation of precision medicine via individualized risk assessment on the basis of various FRS element combinations (26), and the development of the alternative FRS (a-FRS) (24).

Notably, inflammatory factors have received insufficient attention within the context of FRS and existing POPF risk research. Such biomarkers have demonstrated predictive value across a spectrum of diseases (27-31), as supported by pancreatic-specific research: the neutrophil-to-lymphocyte ratio (NLR) and lymphocyte-to-monocyte ratio (LMR) predict prognosis in patients with pancreatic cancer (32); the platelet-to-lymphocyte ratio (PLR) serves as a significant prognostic indicator for pancreatic ductal adenocarcinoma (PDAC) (33); and the NLR combined with the platelet count (PLT) are primary indicators for monitoring complications post-PD (34). On the basis of this evidence, the present study attempts to integrate inflammatory factors with clinical variables to increase the predictive precision for POPF. We present this article in accordance with the TRIPOD reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0108/rc).

Methods

Patients and data collection

This retrospective cohort study included 1,026 patients who underwent PD between March 2019 and March 2023. All the data were sourced from West China Hospital, Sichuan University. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocol was reviewed and approved by the Institutional Review Board of West China Hospital (No. 2023-880), and individual consent for this retrospective analysis was waived. We retrospectively collated general vital signs, baseline comorbidities, preoperative and postoperative laboratory results, and pancreatic fistula status via the electronic medical records system. The exclusion criteria included incomplete data, non-PD surgical procedures, autoimmune diseases, and patients under the age of 18 years.

Surgical procedure and perioperative management

All PD procedures were performed by experienced surgeons utilizing either open or laparoscopic classic Whipple techniques at West China Hospital. Pancreaticojejunostomy is universally employed. Intraperitoneal drainage tubes were routinely placed intraoperatively to monitor for POPF. Prophylactic antibiotics were administered to all patients, with regimens adjusted postoperatively on the basis of patient recovery. Somatostatin or octreotide therapy was typically maintained from postoperative day (POD) 2 to 5. Patients initially received parenteral nutrition in the early postoperative phase and gradually transitioned to enteral nutrition. Chest and abdominal CT scans were routinely performed on POD 3, followed by periodic abdominal imaging. Upon improvement of their general condition (typically PODs 5–7), patients were transferred to a rehabilitation facility for subsequent care.

Outcome, predictors, and definitions

The primary endpoint of this study was the occurrence of grade B/C POPF within 30 days post-operatively, which was classified according to the 2016 ISGPS definition (16). Potential predictors evaluated included sex, age, body mass index (BMI), comorbidities (hypertension, diabetes, cardiovascular disease, cerebral disease, pulmonary disease), jaundice, surgical approach (laparoscopic), extended resection, pathology (type, texture), pancreatic duct diameter, American Society of Anaesthesiologists (ASA) physical status (35), the Geriatric Nutritional Risk Index (GNRI) (36), intraoperative blood loss, pancreatic texture, operation time, haemoglobin, albumin, platelets, neutrophils, lymphocytes, monocytes, fibrinogen, serum creatinine, and inflammatory indices [NLR, PLR, systemic inflammatory response index (SIRI), systemic immune-inflammation index (SII), fibrinogen-to-albumin ratio (FAR), prognostic nutritional index (PNI), aspartate aminotransferase (AST)-to-lymphocyte ratio index (ALI), and AST-to-platelet ratio index (APRI)]. Pathological types were confirmed via postoperative pathology reports. The pancreatic texture (soft vs. hard) was extracted from operative notes. The pancreatic duct diameter was measured directly at the transection site intraoperatively. ASA classification was dichotomized with grade II as the threshold (37). Inflammatory indices were calculated as follows: NLR = neutrophil/lymphocyte, PLR = platelet/lymphocyte, SIRI = neutrophil × monocyte/lymphocyte, SII = platelet × neutrophil/lymphocyte, FAR, PNI, ALI, and APRI (27-31,38,39). The GNRI was calculated via the Lorentz formula for ideal weight, with a cut-off value of 98 for dichotomization (40). Laparoscopic PD was defined as laparoscopic resection combined with laparoscopic reconstruction (41).

Statistical analysis

All analyses were conducted via R version 4.5.1 and Python. Patients were randomly partitioned into a training dataset (70%) and an internal validation dataset (30%) based on the outcome (presence or absence of POPF), and the fixed random seed was used in this process. The validation dataset was strictly isolated during the previous model training phase. Missing values were assessed via the “naniar” R package; missing values under 5% were deemed negligible (Figure 1). We then used the “mice” R package to perform multiple imputation by chained equations under the missing at random (MAR) assumption, generating five parallel imputed datasets. The Kolmogorov-Smirnov test was used to assess normality; normally distributed variables are presented as the mean ± standard deviation (SD), whereas nonnormally distributed continuous data are presented as the median (Q1, Q3). The “tableone” R package facilitated group comparisons between POPF and non-POPF cohorts via Pearson’s χ² test or Fisher’s exact test for categorical variables and the Wilcoxon rank-sum test for nonparametric continuous variables.

Figure 1 Visualization of missing data proportions. ALI, AST-to-lymphocyte ratio index; ALT, alanine transaminase; APRI, AST-to-platelet ratio index; ASA, American Society of Anaesthesiologists; AST, aspartate aminotransferase; BMI, body mass index; FAR, fibrinogen-to-albumin ratio; GNRI, Geriatric Nutritional Risk Index; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; PNI, prognostic nutritional index; SII, systemic immune-inflammation index; SIRI, systemic inflammatory response index.

We employed the “glmnet” R package for 10–30% least absolute shrinkage and selection operator (LASSO) regression, which uses 10-fold cross-validation to determine the minimum λ for feature selection. This effectively eliminates irrelevant or multicollinear independent variables to reduce high-dimensional data (42). The selected features were subjected to multivariable logistic regression to identify independent risk factors. To avoid information loss and potential classification bias caused by arbitrary dichotomization, predictors such as BMI and operation time were retained as continuous variables. Prior to multivariable logistic regression modelling, the fundamental assumption of a linear relationship between continuous predictors and the log-odds (logit) of CR-POPF was rigorously evaluated using restricted cubic splines (RCS) with 3 knots. To maintain the simplicity and clinical interpretability of the final nomogram, these variables were subsequently modelled as linear continuous terms without complex non-linear transformations. A nomogram prediction model was then constructed via the “rms” R package. Calibration curves were generated via 1,000 bootstrap resamples, and model fit was assessed via the Hosmer-Lemeshow test. Receiver operating characteristic (ROC) curves were plotted via Python to calculate the area under the curve (AUC), which was used to evaluate model performance (43). To ensure absolute comparability with previously published literature, the traditional FRS score for each patient in this study cohort was strictly calculated according to the original 10-point scale established by Callery et al. in 2013 (18). A P value <0.05 was considered statistically significant.

Results

Clinical characteristics of the overall cohort

This study presents the data collection (Figure 1) and characteristics of the 1,026 enrolled patients, stratified by postoperative POPF status (Table 1). The proportion of missing data for all variables was less than 0.6%. In terms of general demographics and comorbidities, the POPF group presented a significantly lower incidence of diabetes (11.64% vs. 20.91%, P=0.005) and a higher BMI (median 23.50 vs. 21.80 kg/m², P<0.001); no other significant differences were observed. Notably, regarding pancreatic pathology and surgical metrics, the POPF group had a greater proportion of patients with a soft pancreatic texture (50.79% vs. 32.62%, P<0.001), a smaller pancreatic duct diameter (≤3 cm: 41.80% vs. 25.57%, P<0.001), and a greater incidence of pancreatic cancer or pancreatitis (55.03% vs. 43.01%, P=0.004). Additionally, the operation duration was prolonged in the POPF group (median 300 vs. 270 minutes, P<0.001). The APRI values were significantly lower in the POPF group (median 0.46 vs. 0.59, P=0.004).

Our study presents the clinical characteristics of the 719 patients included in the training set (70% random sample), comprising the non-POPF group (n=586) and POPF group (n=133) (Table 2). The results mirrored those of the overall cohort: the POPF group had a lower diabetes incidence (12.03% vs. 21.50%, P=0.02), higher BMI (median 24.14 vs. 21.85 kg/m², P<0.001), longer operation time (median 309.00 vs. 264.50 minutes, P<0.001), and greater proportion of patients with blood loss >1,000 mL (15.79% vs. 8.02%, P=0.009). In terms of pancreatic features, the POPF group presented a softer texture (51.88% vs. 32.76%, P<0.001), a smaller duct diameter (45.11% vs. 24.74%, P<0.001), and a greater prevalence of specific pathologies (56.39% vs. 43.00%, P=0.007). The APRI values were similarly lower (median 0.44 vs. 0.56, P=0.01).

The study shows no significant differences (P>0.05) in the baseline characteristics (sex, age, comorbidities, and pancreatic pathology) between the training set (n=719) and the validation set (n=307), indicating robust data homogeneity (Table 3).

LASSO regression analysis

Variables identified via univariate analysis were subjected to LASSO regression. Cross-validation was used to determine the optimal λ value for feature selection (Figure 2). The minimum λ was 0.01467322, identifying nine feature variables: hypertension, diabetes, pancreatic texture, pancreatic duct diameter, BMI, operation time, haemoglobin, fibrinogen, and the APRI.

Figure 2 LASSO regression analysis. (A) Selection of the optimal parameter (λ) in the LASSO model via tenfold cross-validation via the minimum criteria method. The optimal λ value is indicated by the vertical dashed line (λ=0.01467322). (B) LASSO coefficient profiles of the 34 clinical features. The plot is generated via a logarithmic scale of lambda values. The vertical line indicates the λ value selected via tenfold cross-validation, resulting in the identification of 10 features with nonzero coefficients. LASSO, least absolute shrinkage and selection operator.

Multivariate logistic regression analysis

The nine features selected via LASSO were incorporated into a multivariable logistic regression analysis (Table 4). The independent risk factors identified were diabetes, pancreatic texture, pancreatic duct diameter, BMI, and operation time (P<0.05). The resulting prediction model formula is as follows: logit(P) = −4.834 + (−0.575) × (pancreatic duct diameter) + 0.167 × (BMI) + 0.004 × (operation time) + (−0.688) × (diabetes) + (−0.506) × (pancreatic texture).

Probability prediction formula: P = 1/{1 + e^[−logit(P)]}.

Development of an individualized prediction model

The RCS analysis (Figure S1) confirmed a strictly linear relationship for BMI. For operation time, an approximately linear relationship was observed within the core clinical duration window (100–300 minutes). A nomogram was constructed by integrating these critical clinical features to visualize the relative contribution of each factor to the risk of POPF. As illustrated in the nomogram (Figure 3), soft pancreatic texture and small pancreatic duct diameter emerged as the strongest predictors, followed by elevated BMI, prolonged operation time, and the absence of diabetes. Conversely, the nomogram highlights protective factors: hard pancreatic texture, large duct diameter, the presence of diabetes, low BMI, and shorter operative duration.

Figure 3 A dynamic nomogram for predicting the risk of POPF. This dynamic nomogram integrates the standard nomogram with illustrative examples (red dashed lines and points). For the pancreatic texture, 1 represents hard, and 0 represents soft. For the pancreatic duct diameter, 1 represents a diameter >3 cm, and 0 represents a diameter ≤3 cm. For diabetes, 1 represents presence, and 0 represents absence. BMI, body mass index; POPF, postoperative pancreatic fistula.

Predictive model validation

In the training cohort, the model demonstrated robust discriminatory ability for POPF, with an AUC of 0.770 [95% confidence interval (CI): 0.778–0.811] (Figure 4A). This performance was sustained in the independent validation cohort, yielding an AUC of 0.759 (95% CI: 0.716–0.801) (Figure 4B). By comparison, when established scoring systems were applied to the same validation cohort, the a-FRS yielded an AUC of only 0.579 (95% CI: 0.500–0.665) (Figure 4C), and the FRS yielded an AUC of 0.529 (95% CI: 0.481–0.518) (Figure 4D).

Figure 4 Comparative analysis of ROC curves across distinct patient cohorts. (A) The proposed model in the training set. (B) The proposed model in the validation set. (C) The a-FRS model in the training set. (D) The FRS model in the training set. a-FRS, alternative FRS; AUC, area under the curve; CI, confidence interval; FRS, Fistula risk score; ROC, receiver operating characteristic.

Calibration curves (Figure 5) were used to assess the model’s accuracy. In both the training (Figure 5A) and validation (Figure 5B) cohorts, the bias-corrected calibration curve (blue solid line) closely aligned with the ideal curve (gray dashed line), indicating excellent agreement between the predicted and observed probabilities. The Hosmer-Lemeshow test (χ²=7.371, P=0.50) confirmed that there was no significant difference between the predicted and actual outcomes. Decision curve analysis (DCA) for the validation set (Figure 6) demonstrated that the model provides superior net benefit across a broad range of clinically relevant threshold probabilities (approximately 0.05–0.42). The incidence of POPF in the validation set was 18.24% (56/307), which was statistically comparable to the 18.50% (133/719) observed in the training set.

Figure 5 Calibration curve of the multivariate logistic regression model for predicting POPF. (A) Calibration on the training set. (B) Calibration in the validation set. POPF, postoperative pancreatic fistula.

Figure 6 DCA for the POPF prediction model. DCA, decision curve analysis; POPF, postoperative pancreatic fistula.

Discussion

Although surgical techniques for PD have evolved from traditional open approaches to minimally invasive laparoscopic and robotic-assisted methods and perioperative management has been optimized through strategies such as prophylactic somatostatin administration and refined drainage care (3,44,45), CR-POPF remains one of the most intractable complications, with an incidence of 10–20% (46). This complication not only precipitates secondary adverse events—including delayed gastric emptying, intra-abdominal infection, and postoperative haemorrhage—but also extends hospitalization duration by 3–5-fold, increases unplanned readmission rates by over 40%, and significantly increases medical costs and perioperative mortality risk (13,15). Consequently, the construction of a precise POPF risk prediction tool to facilitate a “high-risk warning and stratified intervention” management model remains a core imperative in pancreatic surgery.

Our final nomogram, which integrates the pancreatic texture, duct diameter, BMI, operation time, and diabetes status, achieved AUCs of 0.770 (training) and 0.759 (validation), with calibration curves showing high concordance (Hosmer-Lemeshow P=0.50). DCA confirmed a significant net benefit within the 0.05–0.42 threshold range.

The traditional FRS, the first widely validated POPF prediction tool, achieves risk stratification via four indicators: pancreatic texture, duct diameter, pathology, and intraoperative blood loss (18). However, limitations in its clinical utility have emerged across multiple studies. Our analysis of a single-centre cohort of 1,026 PD patients revealed that the traditional FRS yielded an AUC of only 0.529 in the training set, falling well below the threshold for clinical utility (AUC >0.7). This aligns with findings by Shubert et al. (21) and Mungroop et al. (24) reported insufficient predictive efficacy of the FRS in certain cohorts. Direct comparison of the present model with the FRS and a-FRS systems reveals the following: (I) the traditional FRS is highly dependent on intraoperative blood loss, which accounts for a significant proportion of the total score. However, recent external validation studies have indicated that intraoperative blood loss is no longer a significant predictor of POPF (24). Furthermore, with advancements in surgical instrumentation and the widespread adoption of minimally invasive techniques, massive intraoperative hemorrhage has become exceedingly rare, rendering this variable of limited value for risk stratification. Additionally, the intraoperative blood loss indicator in the FRS employs fixed weights and fails to account for differences between surgical approaches. For instance, in laparoscopic surgery, even minor bleeding (<500 mL) may indicate extreme anatomical difficulty, whereas the clinical significance of this threshold differs in open surgery (21). Moreover, the traditional FRS includes pathological type as a risk factor, which is largely an external manifestation of pancreatic texture and fibrosis (e.g., PDAC often leads to proximal ductal dilation and a firm, fibrotic texture, while ampullary tumours are frequently associated with a soft pancreas) (47). In our study, although pathological type was statistically significant in univariate analysis, it was entirely eliminated during the rigorous LASSO regression and multivariate feature selection process. This suggests that the crude classification of pathological type is statistically suboptimal. (II) Comparison with the a-FRS: the a-FRS improved upon the traditional FRS by excluding intraoperative blood loss and incorporating BMI to overcome limitations. However, its performance in our internal validation cohort remained unsatisfactory (AUC =0.589). Furthermore, BMI is a crude metric: it only reflects systemic obesity and fails to distinguish between subcutaneous fat and local pancreatic fat deposition (fatty pancreas), thereby ignoring the complex variations in the local microenvironment. Recent evidence suggests that pancreatic steatosis, rather than systemic obesity, is the key factor leading to increased tissue fragility and impaired healing capacity (48). Consequently, the a-FRS may underestimate the risk in patients with the Metabolically Obese Normal Weight (MONW) phenotype, which is particularly common in Asian populations. Additionally, the a-FRS remains highly dependent on subjective intraoperative assessment of pancreatic texture, a metric prone to significant inter-observer variability.

In contrast, the present model introduces operative time as a superior continuous variable, which objectively reflects the extent of local tissue damage, prolonged organ traction, and ischemia-reperfusion injury. Prolonged operative time is directly associated with increased tissue ischemia and an elevated risk of stress injury to the anastomosis (49). Furthermore, the simultaneous inclusion of BMI and diabetes status provides a critical pathological distinction: a high BMI promotes pancreatic fatty infiltration, rendering the gland soft and vulnerable to lipotoxic injury (50,51). Conversely, chronic diabetes triggers extensive pancreatic fibrosis and glandular atrophy. Fibrosis results in a firmer pancreatic texture, which enhances the efficacy of suture fixation and reduces the secretion of destructive exocrine enzymes, thereby serving as a protective factor against POPF (52,53). While the a-FRS system fails to account for these factors, our model captures these two opposing pathophysiological states, achieving a more granular and in-depth risk stratification.

Therefore, the clinical value of this model is threefold. First, diabetes status was a protective factor—the POPF incidence was significantly lower in diabetic patients [11.64% vs. 20.91%, P=0.005; odds ratio (OR) =0.51, P=0.003]. This aligns with mechanisms whereby long-term diabetes induces pancreatic fibrosis and reduces exocrine secretion. Second, the model relies on readily accessible clinical metrics, facilitating broad application without additional testing costs. Third, it enables precise “quantified risk-stratified intervention” management. Low-risk patients (<10%) may qualify for accelerated recovery pathways with early drain removal (POD 3–5) (7,44,54); moderate-risk patients (10–30%) may require intensified amylase monitoring and imaging; and high-risk patients (>30%) may benefit from pancreaticogastrostomy (4,5), external stents (6,55), or extended somatostatin therapy, balancing the risks of overtreatment and undertreatment.

Notably, although our final predictive model did not retain inflammatory factors as independent predictors, this does not negate their role in POPF pathogenesis; rather, it suggests a “strong clinical factor masking weak inflammatory factor effect”. In essence, CR-POPF is an acute local complication driven by mechanical healing failure and local enzymatic autodigestion (i.e., post-pancreatectomy acute pancreatitis). In the presence of significant local risk factors—such as a soft, fatty pancreas and a narrow main pancreatic duct—the patient’s baseline systemic inflammatory signals are completely masked. The initial analysis included seven classic inflammatory/immune indices (e.g., NLR, PLR, SIRI, and SII), which have predictive value in pancreatic cancer prognosis and complication monitoring (31-33). However, LASSO screening retained only fibrinogen and the APRI, which were subsequently excluded from multivariate analysis because of collinearity with pancreatic texture or nonsignificant ORs. These findings indicate that in the presence of potent predictors such as pancreatic texture (OR =2.89, P<0.001) and duct diameter (OR =2.57, P<0.001), the predictive contribution of inflammatory factors is overshadowed. Inflammation likely plays a “mediating” role in POPF development. For example, patients with a soft pancreatic texture exhibit greater pancreatic juice secretion (42), which may exacerbate anastomotic injury by stimulating local inflammation. Dynamic postoperative inflammatory markers (e.g., NLR changes on PODs 1–3) might reflect this pathophysiological process better than static preoperative metrics do. Recent research on glucocorticoids suggests that preoperative administration may reduce complications by suppressing excessive postoperative pancreatic inflammation (PPAP), thereby interrupting key upstream pathways leading to POPF (56). Future research should focus on combining “preoperative and dynamic postoperative inflammatory markers” or reverifying their value in scenarios lacking intraoperative metrics.

The limitations of this study include its single-centre, retrospective design, which introduces potential selection bias. Certain confounders (e.g., specific anastomosis techniques and antibiotic duration) were not analysed. Additionally, the model lacks external validation; future multicentre prospective studies are needed to confirm its robustness across different regions and surgical proficiencies. Finally, detailed diabetes data [duration, haemoglobin A1c (HbA1c)] were missing, preventing deeper analysis of the protective mechanism.

In conclusion, the nomogram model developed in this study refines the predictive metrics beyond conventional frameworks such as the FRS, significantly enhancing both predictive accuracy and clinical utility for post-pancreatectomy POPF. This tool offers clinicians an intuitive, individualized risk assessment mechanism, thereby supporting preoperative surgical planning, intraoperative technique selection, and optimization of postoperative management strategies. Ultimately, its application may contribute to a reduction in the incidence of POPF and alleviate the associated disease burden on patients.

Conclusions

In conclusion, based on large-scale real-world data and through the re-validation and optimization of existing risk factor systems, we successfully developed and validated a nomogram model. This model demonstrates good calibration and significant clinical utility in predicting POPF after pancreatoduodenectomy. As an individualized risk assessment tool, it empowers surgeons to identify high-risk patients perioperatively, facilitating the formulation of targeted prevention and management strategies and ultimately improving patient prognosis.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0108/rc

Data Sharing Statement: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0108/dss

Peer Review File: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0108/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0108/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of West China Hospital, Sichuan University (No. 2023-880), and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Clancy TE, Ashley SW. Pancreaticoduodenectomy (Whipple operation). Surg Oncol Clin N Am 2005;14:533-52. vii. [Crossref] [PubMed]
2. Bassi C, Dervenis C, Butturini G, et al. Postoperative pancreatic fistula: an international study group (ISGPF) definition. Surgery 2005;138:8-13. [Crossref] [PubMed]
3. Andrianello S, Marchegiani G, Malleo G, et al. Pancreaticojejunostomy With Externalized Stent vs Pancreaticogastrostomy With Externalized Stent for Patients With High-Risk Pancreatic Anastomosis: A Single-Center, Phase 3, Randomized Clinical Trial. JAMA Surg 2020;155:313-21. [Crossref] [PubMed]
4. Fernández-Cruz L, Cosa R, Blanco L, et al. Pancreatogastrostomy with gastric partition after pylorus-preserving pancreatoduodenectomy versus conventional pancreatojejunostomy: a prospective randomized study. Ann Surg 2008;248:930-8. [Crossref] [PubMed]
5. Topal B, Fieuws S, Aerts R, et al. Pancreaticojejunostomy versus pancreaticogastrostomy reconstruction after pancreaticoduodenectomy for pancreatic or periampullary tumours: a multicentre randomised trial. Lancet Oncol 2013;14:655-62. [Crossref] [PubMed]
6. McMillan MT, Ecker BL, Behrman SW, et al. Externalized Stents for Pancreatoduodenectomy Provide Value Only in High-Risk Scenarios. J Gastrointest Surg 2016;20:2052-62. [Crossref] [PubMed]
7. McMillan MT, Malleo G, Bassi C, et al. Multicenter, Prospective Trial of Selective Drain Management for Pancreatoduodenectomy Using Risk Stratification. Ann Surg 2017;265:1209-18. [Crossref] [PubMed]
8. Ochiai T, Sonoyama T, Soga K, et al. Application of polyethylene glycolic acid felt with fibrin sealant to prevent postoperative pancreatic fistula in pancreatic surgery. J Gastrointest Surg 2010;14:884-90. [Crossref] [PubMed]
9. El Nakeeb A, ElGawalby A, A, Ali M, et al. Efficacy of octreotide in the prevention of complications after pancreaticoduodenectomy in patients with soft pancreas and non-dilated pancreatic duct: A prospective randomized trial. Hepatobiliary Pancreat Dis Int 2018;17:59-63. [Crossref] [PubMed]
10. Ecker BL, McMillan MT, Asbun HJ, et al. Characterization and Optimal Management of High-risk Pancreatic Anastomoses During Pancreatoduodenectomy. Ann Surg 2018;267:608-16. [Crossref] [PubMed]
11. McMillan MT, Christein JD, Callery MP, et al. Prophylactic octreotide for pancreatoduodenectomy: more harm than good? HPB (Oxford) 2014;16:954-62. [Crossref] [PubMed]
12. Young S, Sung ML, Lee JA, et al. Pasireotide is not effective in reducing the development of postoperative pancreatic fistula. HPB (Oxford) 2018;20:834-40. [Crossref] [PubMed]
13. Casciani F, Bassi C, Vollmer CM Jr. Decision points in pancreatoduodenectomy: Insights from the contemporary experts on prevention, mitigation, and management of postoperative pancreatic fistula. Surgery 2021;170:889-909. [Crossref] [PubMed]
14. Vollmer CM Jr, Sanchez N, Gondek S, et al. A root-cause analysis of mortality following major pancreatectomy. J Gastrointest Surg 2012;16:89-102; discussion 102-3. [Crossref] [PubMed]
15. Abbott DE, Tzeng CW, McMillan MT, et al. Pancreas fistula risk prediction: implications for hospital costs and payments. HPB (Oxford) 2017;19:140-6. [Crossref] [PubMed]
16. Bassi C, Marchegiani G, Dervenis C, et al. The 2016 update of the International Study Group (ISGPS) definition and grading of postoperative pancreatic fistula: 11 Years After. Surgery 2017;161:584-91. [Crossref] [PubMed]
17. Pratt WB, Maithel SK, Vanounou T, et al. Clinical and economic validation of the International Study Group of Pancreatic Fistula (ISGPF) classification scheme. Ann Surg 2007;245:443-51. [Crossref] [PubMed]
18. Callery MP, Pratt WB, Kent TS, et al. A prospectively validated clinical risk score accurately predicts pancreatic fistula after pancreatoduodenectomy. J Am Coll Surg 2013;216:1-14. [Crossref] [PubMed]
19. Ecker BL, McMillan MT, Maggino L, et al. Pancreatogastrostomy Vs. Pancreatojejunostomy: a Risk-Stratified Analysis of 5316 Pancreatoduodenectomies. J Gastrointest Surg 2018;22:68-76.
20. Miller BC, Christein JD, Behrman SW, et al. A multi-institutional external validation of the fistula risk score for pancreatoduodenectomy. J Gastrointest Surg 2014;18:172-79; discussion 179-80. [Crossref] [PubMed]
21. Shubert CR, Wagie AE, Farnell MB, et al. Clinical Risk Score to Predict Pancreatic Fistula after Pancreatoduodenectomy: Independent External Validation for Open and Laparoscopic Approaches. J Am Coll Surg 2015;221:689-98. [Crossref] [PubMed]
22. Mungroop TH, Klompmaker S, Wellner UF, et al. Updated Alternative Fistula Risk Score (ua-FRS) to Include Minimally Invasive Pancreatoduodenectomy: Pan-European Validation. Ann Surg 2021;273:334-40. [Crossref] [PubMed]
23. McMillan MT, Vollmer CM Jr, Asbun HJ, et al. The Characterization and Prediction of ISGPF Grade C Fistulas Following Pancreatoduodenectomy. J Gastrointest Surg 2016;20:262-76. [Crossref] [PubMed]
24. Mungroop TH, van Rijssen LB, van Klaveren D, et al. Alternative Fistula Risk Score for Pancreatoduodenectomy (a-FRS): Design and International External Validation. Ann Surg 2019;269:937-43. [Crossref] [PubMed]
25. Kantor O, Talamonti MS, Pitt HA, et al. Using the NSQIP Pancreatic Demonstration Project to Derive a Modified Fistula Risk Score for Preoperative Risk Stratification in Patients Undergoing Pancreaticoduodenectomy. J Am Coll Surg 2017;224:816-25. [Crossref] [PubMed]
26. Trudeau MT, Casciani F, Ecker BL, et al. The Fistula Risk Score Catalog: Toward Precision Medicine for Pancreatic Fistula After Pancreatoduodenectomy. Ann Surg 2022;275:e463-72. [Crossref] [PubMed]
27. Meng F, Yan X, Qi J, et al. Association of neutrophil to lymphocyte ratio, platelet to lymphocyte ratio, and monocyte to lymphocyte ratio with depression: A cross-sectional analysis of the NHANES data. J Affect Disord 2022;315:168-73. [Crossref] [PubMed]
28. Liu B, Wang J, Li YY, et al. The association between systemic immune-inflammation index and rheumatoid arthritis: evidence from NHANES 1999-2018. Arthritis Res Ther 2023;25:34. [Crossref] [PubMed]
29. Song Y, Guo W, Li Z, et al. Systemic immune-inflammation index is associated with hepatic steatosis: Evidence from NHANES 2015-2018. Front Immunol 2022;13:1058779. [Crossref] [PubMed]
30. Wang RH, Wen WX, Jiang ZP, et al. The clinical value of neutrophil-to-lymphocyte ratio (NLR), systemic immune-inflammation index (SII), platelet-to-lymphocyte ratio (PLR) and systemic inflammation response index (SIRI) for predicting the occurrence and severity of pneumonia in patients with intracerebral hemorrhage. Front Immunol 2023;14:1115031. [Crossref] [PubMed]
31. Xu M, Chen R, Liu L, et al. Systemic immune-inflammation index and incident cardiovascular diseases among middle-aged and elderly Chinese adults: The Dongfeng-Tonji cohort study. Atherosclerosis 2021;323:20-9. [Crossref] [PubMed]
32. Sierzega M, Lenart M, Rutkowska M, et al. Preoperative Neutrophil-Lymphocyte and Lymphocyte-Monocyte Ratios Reflect Immune Cell Population Rearrangement in Resectable Pancreatic Cancer. Ann Surg Oncol 2017;24:808-15. [Crossref] [PubMed]
33. Shirai Y, Hamura R, Tanji Y, et al. The postoperative platelet-to-lymphocyte ratio predicts the outcome of patients undergoing pancreaticoduodenectomy for pancreatic head cancer. Surg Today 2024;54:247-57. [Crossref] [PubMed]
34. Ma M, Li G, Zhou B, et al. Comprehensive analysis of the association between inflammation indexes and complications in patients undergoing pancreaticoduodenectomy. Front Immunol 2023;14:1303283. [Crossref] [PubMed]
35. Hendrix JM, Garmon EH. American Society of Anesthesiologists Physical Status Classification System. 2026.
36. Grinstead C, Yoon SL. Geriatric Nutritional Risk Index (GNRI) and Survival in Pancreatic Cancer: A Retrospective Study. Nutrients 2025;17:509. [Crossref] [PubMed]
37. Ailhaud A, Girard E, Fournier J, et al. Completion pancreatectomy during pancreatoduodenectomy: A lifesaving solution in high-risk patients. World J Surg 2024;48:1123-31. [Crossref] [PubMed]
38. Xie QK, Chen P, Hu WM, et al. The systemic immune-inflammation index is an independent predictor of survival for metastatic colorectal cancer and its association with the lymphocytic response to the tumor. J Transl Med 2018;16:273. [Crossref] [PubMed]
39. Caputo D, Coppola A, Cascone C, et al. Preoperative systemic inflammatory biomarkers and postoperative day 1 drain amylase value predict grade C pancreatic fistula after pancreaticoduodenectomy. Ann Med Surg (Lond) 2020;57:56-61. [Crossref] [PubMed]
40. Bouillanne O, Morineau G, Dupont C, et al. Geriatric Nutritional Risk Index: a new index for evaluating at-risk elderly medical patients. Am J Clin Nutr 2005;82:777-83. [Crossref] [PubMed]
41. Montagnini AL, Røsok BI, Asbun HJ, et al. Standardizing terminology for minimally invasive pancreatic resection. HPB (Oxford) 2017;19:182-9. [Crossref] [PubMed]
42. Hu JY, Wang Y, Tong XM, et al. When to consider logistic LASSO regression in multivariate analysis? Eur J Surg Oncol 2021;47:2206. [Crossref] [PubMed]
43. Yu A, Li Y, Zhang H, et al. Development and validation of a preoperative nomogram for predicting the surgical difficulty of laparoscopic colectomy for right colon cancer: a retrospective analysis. Int J Surg 2023;109:870-8. [Crossref] [PubMed]
44. McMillan MT, Fisher WE, Van Buren G 2nd, et al. The value of drains as a fistula mitigation strategy for pancreatoduodenectomy: something for everyone? Results of a randomized prospective multi-institutional study. J Gastrointest Surg 2015;19:21-30; discussion 30-1.
45. Büchler M, Friess H, Klempa I, et al. Role of octreotide in the prevention of postoperative complications following pancreatic resection. Am J Surg 1992;163:125-30; discussion 130-1. [Crossref] [PubMed]
46. Hackert T, Werner J, Büchler MW. Postoperative pancreatic fistula. Surgeon 2011;9:211-7. [Crossref] [PubMed]
47. Czarnecka Z, Verhoeff K, Bigam D, et al. Impact of soft pancreas on pancreaticoduodenectomy outcomes and the development of the preoperative soft pancreas risk score. Ann Hepatobiliary Pancreat Surg 2025;29:62-71. [Crossref] [PubMed]
48. Zhou L, Xiao WM, Li CP, et al. Impact of Fatty Pancreas on Postoperative Pancreatic Fistulae: A Meta-Analysis. Front Oncol 2021;11:622282. [Crossref] [PubMed]
49. Su AP, Zhang Y, Ke NW, et al. Triple-layer duct-to-mucosa pancreaticojejunostomy with resection of jejunal serosa decreased pancreatic fistula after pancreaticoduodenectomy. J Surg Res 2014;186:184-91. [Crossref] [PubMed]
50. Liu QY, Zhang WZ, Xia HT, et al. Analysis of risk factors for postoperative pancreatic fistula following pancreaticoduodenectomy. World J Gastroenterol 2014;20:17491-7. [Crossref] [PubMed]
51. Mathur A, Pitt HA, Marine M, et al. Fatty pancreas: a factor in postoperative pancreatic fistula. Ann Surg 2007;246:1058-64. [Crossref] [PubMed]
52. Gaujoux S, Cortes A, Couvelard A, et al. Fatty pancreas and increased body mass index are risk factors of pancreatic fistula after pancreaticoduodenectomy. Surgery 2010;148:15-23. [Crossref] [PubMed]
53. Søreide K, Healey AJ, Mole DJ, et al. Pre-, peri- and post-operative factors for the development of pancreatic fistula after pancreatic surgery. HPB (Oxford) 2019;21:1621-31. [Crossref] [PubMed]
54. Van Buren G 2nd, Bloomston M, Hughes SJ, et al. A randomized prospective multicenter trial of pancreaticoduodenectomy with and without routine intraperitoneal drainage. Ann Surg 2014;259:605-12. [Crossref] [PubMed]
55. Motoi F, Egawa S, Rikiyama T, et al. Randomized clinical trial of external stent drainage of the pancreatic duct to reduce postoperative pancreatic fistula after pancreaticojejunostomy. Br J Surg 2012;99:524-31. [Crossref] [PubMed]
56. Chen H, Wang Y, Wang C, et al. The effect of perioperative of dexamethasone on postoperative complications after pancreaticoduodenectomy (PANDEX): a study protocol for a pragmatic multicenter randomized controlled trial. Trials 2023;24:569. [Crossref] [PubMed]