A novel preoperative risk score based on computed tomography morphometrics for predicting clinically relevant postoperative pancreatic fistula after pancreatoduodenectomy.

BACKGROUND

Clinically relevant postoperative pancreatic fistula (CR-POPF) remains one of the most serious and frequent complications after pancreatoduodenectomy (PD), leading to prolonged hospital stay, increased costs, and higher mortality. Despite numerous intraoperative risk models, there is a lack of objective, reproducible, and fully preoperative tools for reliable risk stratification.

METHODS

This retrospective study included 234 patients who underwent PD. Preoperative multiphase contrast-enhanced computed tomography (including noncontrast native phase) was used to measure pancreatic parenchymal attenuation, ventrodorsal and craniocaudal duct-to-parenchyma (D/P) ratios, main pancreatic duct diameter, and body mass index (BMI). A 6-point scoring system was constructed based on multivariate logistic regression and receiver operating characteristic (ROC) curve analysis. Patients were classified into 4 risk groups.

RESULTS

Pancreatic attenuation of ≤30 Hounsfield units, D/P ratio of ≤0.2, duct diameter of <3 mm, and BMI of ≥25 kg/m² were significantly associated with CR-POPF. The combined 6-point score achieved an area under the ROC curve of 0.89, with a sensitivity of 81.8% and a specificity of 91.7%. POPF incidence ranged from 4.6% in low-risk to 82.5% in very-high-risk patients. The scoring system also correlated with the severity of the fistula. Baseline incidence of CR-POPF (International Study Group of Pancreatic Surgery grade B/C) was 18.8% in this cohort.

CONCLUSION

The proposed model enables fully preoperative, quantitative risk assessment of CR-POPF using objective and widely available parameters. Its implementation allows for individualized surgical planning and targeted prophylactic strategies, potentially improving postoperative outcomes and reducing complication rates after PD. The model is currently undergoing internal validation in a prospective cohort, whereas external validation remains necessary.