Objective:

To develop and externally validate a clinically applicable risk-scoring model to predict late post-pancreatectomy hemorrhage (PPH) after pancreatoduodenectomy (PD).

Approach:

• Patient Selection: Patients who underwent curative-intent PD were included, with a training cohort of 300 patients and an external validation cohort of 105 patients.
• Model Development: The risk-scoring model was developed using multivariate logistic analysis to identify independent risk factors for late PPH.
• Validation: The model's performance was validated in an external cohort, assessing its discriminative capacity by categorizing predicted probabilities into low-risk and high-risk groups.

Key Findings:

• Late PPH occurred in 7.7% of the training cohort and 7.6% of the validation cohort.
• Median survival was significantly lower in patients with late PPH (20.7 months) compared to those without (35.2 months, P = 0.009).
• Independent risk factors for late PPH included BMI, preoperative total bilirubin, preoperative prothrombin time, and clinically relevant postoperative pancreatic fistula.
• The predictive model demonstrated concordance indices of 0.863 in the training cohort and 0.825 in the validation cohort.
• The most frequent site of hemorrhage was the common hepatic artery, accounting for 22.6% of cases.

Interpretation:

Limitations:

• The study is retrospective and conducted at high-volume centers, which may limit generalizability.
• The model requires further validation in diverse clinical settings.

Conclusion:

A simple risk-scoring model for late PPH after PD was developed and validated, facilitating risk prediction and potential preventive measures.