A Novel Prediction Model of Clinically Relevant Postoperative Pancreatic Fistula After Pancreaticoduodenectomy: A Retrospective Clinical Study.

INTRODUCTION

Clinically relevant postoperative pancreatic fistula (CR-POPF) after pancreaticoduodenectomy (PD) is one of the most life-threatening complications. Early drain removal after PD is recommended in the clinical setting; however, the risk stratification for CR-POPF to guide drain removal is not widely recognized.

METHODS

This single-center retrospective study included patients who underwent PD between 2013 and 2024. In total, 280 patients were divided into training (n = 196) and validation (n = 84) cohorts by random sampling. The predictive impact of CR-POPF was evaluated using receiver operating characteristic curves of perioperative factors in the training cohort. Based on a probability formula, a predictive model of CR-POPF was established.

RESULTS

CR-POPF occurred in 29.1% of 196 patients in the training cohort. Postoperative day 5 C-reactive protein and drain fluid amylase levels were the strongest diagnostic values in patients with CR-POPF (area under the receiver operating characteristic curve, area under the ROC curve: 0.787 and 0.750, respectively). Logistic regression analysis provided the prediction index (10.5 × C-reactive protein [mg/dL] + 0.01 × drain fluid amylase [IU/L] ± 21.8 [add if the pancreas is soft; subtract if it is hard]) was defined as a new prognostic model for the probability of CR-POPF. A higher CR-POPF prediction index was an independent prognostic factor in the multivariate analysis. Finally, the clinical feasibility of the CR-POPF prediction index was confirmed in the validation cohort.

CONCLUSIONS

The CR-POPF index could guide safe early drain removal, which may help surgeons to implement appropriate drain management after PD.