Predicting pancreatic fistula risk after splenectomy in cirrhotic splenomegaly: Surgical anatomy and implications

Abstract

Clinically relevant postoperative pancreatic fistula (CR-POPF) is an underrecognized complication after splenectomy, particularly in patients with cirrhosis and splenomegaly, where altered anatomy and limited risk assessment models increase operative complexity. A retrospective study published in World Journal of Hepatology by Huang et al, including 186 patients undergoing splenectomy for cirrhotic splenomegaly, identified body mass index and splenic thickness as independent predictors of CR-POPF and proposed a nomogram-based risk model. Among these patients, 21 developed biochemical leaks and six developed CR-POPF, including four grade B and two grade C fistulas. This editorial comments on these findings, briefly reviews CR-POPF mechanisms, diagnosis, and management, and emphasizes the role of splenic and vascular anatomical variations in cirrhotic splenomegaly, which may increase susceptibility to iatrogenic injury during splenic pedicle dissection, particularly when stapling devices are used. Incorporating anatomical risk awareness into surgical planning may improve the prevention and early management of CR-POPF in this high-risk population.

Key Words: Pancreatic fistula; Splenectomy; Accessory spleen; Liver cirrhosis; Splenomegaly; Body mass index; Splenic thickness; Nomogram; Surgical anatomy

Core Tip: Clinically relevant postoperative pancreatic fistula (CR-POPF) is an underrecognized complication after splenectomy in cirrhotic patients with splenomegaly. A retrospective study of 186 patients identified body mass index and splenic thickness as independent predictors of CR-POPF and proposed a nomogram-based model; 21 patients developed biochemical leaks, and six developed CR-POPF. This editorial highlights these findings, briefly reviews CR-POPF management, and emphasizes the role of splenic vascular anatomical variations that may increase susceptibility.

This editorial refers to “Predicting pancreatic fistula post-splenectomy in cirrhosis with splenomegaly: Risk factors and nomogram validation” by Huang L et al, 2026; https://www.wjgnet.com/1948-5182/full/v18/i3/115108.htm.

---

INTRODUCTION

Pancreatic fistulas are abnormal communications between the pancreatic ductal system and the skin or a visceral organ, most commonly arising after inflammation or surgery. Less frequent causes include trauma, malignancy, ductal stones, and strictures[1,2]. Post-pancreatitis fistulas result from partial or complete ductal disruption due to inflammation or vascular compromise, particularly in necrotizing pancreatitis[3-5].

The International Study Group on Pancreatic Surgery defines postoperative pancreatic fistula as drain fluid with amylase levels > 3 × the upper normal serum value plus clinical consequences. In the revised classification, former grade A is redefined as a biochemical leak, while grades B and C constitute clinically relevant postoperative pancreatic fistula (CR-POPF), requiring intervention or associated with organ failure or mortality[1]. Subsequent refinements addressed heterogeneity and interobserver variability, particularly after left pancreatectomy[6].

Risk prediction indicates that gland-related factors outweigh surgeon- or volume-related variables, with soft pancreatic texture and small duct diameter being the strongest independent predictors of CR-POPF[1,2]. Other risk factors include high exocrine output, ischemia, technical factors, high body mass index (BMI), male sex, blood loss, prolonged operative time, and non-ductal pathology[4,5]. Novel tools such as the Real Amylase Value and preoperative models like the Roberts Score may improve early risk stratification and morbidity prediction[7,8]. Although the incidence of CR-POPF remains stable, mortality has declined due to earlier detection and improved supportive care[9,10]. Percutaneous and endoscopic approaches-particularly endoscopic ultrasound-guided drainage with lumen-apposing metal stents-are now cornerstone therapies with high success and favorable safety profiles[11,12]. Adjunctive measures such as negative pressure wound therapy may aid complex cases, while somatostatin analogues (e.g., octreotide, pasireotide) are used therapeutically or prophylactically, with variable efficacy[13-16]. However, studies developing nomogram-based risk prediction models for CR-POPF complicating splenectomy in patients with cirrhotic splenomegaly and portal hypertension remain limited.

CR-POPF AFTER SPLENECTOMY FOR CIRRHOTIC SPLENOMEGALY: SURGICAL ANATOMY, RISK FACTORS AND NOMOGRAM PREDICTION

The paper from Huang et al[17] on the recent issue of World Journal of Hepatology, including 186 patients who underwent splenectomy for cirrhotic splenomegaly, identified BMI and splenic thickness as independent predictors of CR-POPF and proposed a nomogram-based risk prediction model. Using this approach, the authors reported that among the 186 patients, 21 developed biochemical leaks and 6 developed CR-POPF, including four grade B and two grade C fistulas. Although devascularization was performed in 34 cases, it does not appear to have been the primary indication for surgery. The statistical analysis presented is noteworthy. The authors identified BMI, prothrombin time, splenic thickness, splenic vein diameter, stapler division, and pancreatic texture as significant risk factors for CR-POPF. However, splenic thickness, splenic vein diameter, and pancreatic texture demonstrated significant collinearity. Notably, the number of independent risk factors exceeded the number of actual CR-POPF events. Only six patients (3.2%) developed clinically relevant CR-POPF, an impressively low rate for a procedure that is technically more demanding than standard splenectomy, particularly in cirrhotic patients with extensive high-pressure collateral vessels around the spleen.

This low event rate, however, limits the stability of standard regression models, as having more predictors than events increases the risk of overfitting. To address this, the authors appropriately employed least absolute shrinkage and selection operator (LASSO)-penalized logistic regression, which selects the most relevant independent variables while shrinking less informative ones. Using this approach, BMI [odds ratio (OR): 3.58] and splenic thickness (OR: 1.25) emerged as the principal predictors of CR-POPF. In fact, the use of LASSO regression represents an appropriate strategy to mitigate overfitting in small-event datasets.

By integrating stapler division, splenic thickness, prothrombin time, and BMI, Huang et al[17] developed a nomogram to graphically depict CR-POPF risk. Model performance was assessed by plotting sensitivity against 1-specificity, yielding a concordance index (equivalent to the area under the curve) of 0.816. This suggests an approximately 82% probability that the model can correctly distinguish between patients with and without CR-POPF, indicating good to very good predictive performance[18,19].

Nevertheless, the findings should be interpreted with appropriate caution. First, the study did not specifically address the presence of splenic collateral vessels, which are often associated with technically challenging dissections in cirrhotic patients. Although liver stiffness was mentioned as a potential surrogate marker for collateral severity[20], it does not appear to have been included in the regression analysis. Incorporating this parameter in future studies may further strengthen risk stratification. In addition, splenomegaly can increase surgical complexity by distorting peripancreatic anatomy and potentially increasing the risk of pancreatic injury during splenic pedicle dissection[21-23].

Second, anatomical variations such as a gastrosplenic trunk and accessory spleens near the splenic hilum-features not routinely incorporated into current risk models-may influence the risk of CR-POPF (Figure 1). The figure demonstrates these viscerovascular variations in a dissected female cadaver prepared by the corresponding author. Accessory spleens are detected in 10%-40% of autopsies and 45%-65% of splenectomized patients and are usually incidental imaging findings[24-26]. Awareness of these variations is important for accurate diagnosis and safe surgical management, particularly in patients with recurrent autoimmune thrombocytopenic purpura or cirrhotic splenomegaly, to reduce the risk of relapse after splenectomy. Such variations may also affect the blood supply to intrapancreatic accessory spleens and necessitate more extensive hilar dissection and manipulation of the pancreatic tail, especially in the setting of portal hypertension-related anatomical changes[17,19,23,24].

Figure 1 Accessory spleens (black arrows) and a gastrosplenic trunk (blue arrow) may complicate splenectomy and increase the risk of clinically relevant postoperative pancreatic fistula. Note that the stomach was transected at the cardiac end and reflected to expose the gastric bed (original figure from a dissected elderly female cadaver). Sp: Spleen; St: Stomach; P: Pancreas; LG: Left gastric artery; SA: Splenic artery.

It remains unclear whether Huang et al[17] systematically evaluated for accessory spleens at the splenic hilum or within the pancreatic parenchyma, which may influence pancreatic texture and postoperative outcomes. Further clarification on this aspect may enhance the interpretability of the findings and inform future investigations[24-26].

Third, LASSO regression is a valuable method for analyses with limited events and has been widely applied in similar settings. In this study, BMI and splenic thickness were identified as independent predictors in the multivariate LASSO analysis, while the final nomogram also incorporated prothrombin time and stapler use. Providing further insight into the rationale for including these variables, despite their lack of statistical significance in the multivariate model, may help enhance transparency and facilitate a fuller understanding of the model’s development and clinical applicability. Finally, the calculated lower confidence limit for the nomogram’s concordance index is 0.5; this value suggests limited predictive utility if true to model performance.

Limitations

The small number of CR-POPF events may limit model robustness, underscoring the need for external validation in larger, multicenter cohorts. This predictive model also requires further prospective validation and comprehensive performance assessment (e.g., calibration, discrimination, and decision-curve analysis) before widespread clinical adoption. In addition, pancreatic texture assessment by palpation, although clinically acceptable, remains subjective; incorporation of imaging-based or elastography-derived markers may further enhance anatomical risk assessment in future studies[27]. Moreover, variations in splenic anatomy and the celiac trunk may influence splenectomy outcomes, and the nomogram-based variables, including splenic thickness and stapler use, warrant further investigation in future studies[28-30].

CONCLUSION

Huang et al[17] made an important contribution to understanding CR-POPF after splenectomy for cirrhotic splenomegaly by integrating cirrhosis-specific surgical anatomy with predictive modeling. Their findings highlight the critical role of the pancreatic tail-splenic hilum relationship in this setting and provide a valuable framework for improving perioperative and intraoperative decision-making and patient outcomes. While predictive nomograms are valuable tools, the integration of anatomical awareness, surgical expertise, and rigorous external validation remains essential before translating statistical prediction into routine clinical practice.