Early postoperative differentiation between biochemical leak and clinically-relevant pancreatic fistula after pancreaticoduodenectomy: development of a predictive risk score

Abstract Background Several risk scores are available which predict pancreatic fistula after pancreaticoduodenectomy (PD), but do not differentiate between biochemical leak (BL) and clinically relevant pancreatic fistula (CR-POPF). The aim of this study was to identify factors that differentiate between BL and CR-POPF in the early postoperative period. Methods Consecutive patients diagnosed with BL and CR-POPF after PD were identified from a prospectively maintained database (2009-2019). Data were collected for demographics, intraoperative and laboratory parameters on the first five postoperative days (PODs), including drain fluid amylase (DFA), C-reactive protein (CRP) and albumin. Independent predictors of CR-POPF were identified using a multivariable binary logistic regression model, which was subsequently converted to a risk score Results 187 patients consisted of 99 BL and 88 CR-POPF. In those with CR-POPF, the leak became clinically relevant a median of 9 days after surgery; these patients had a significantly higher length of hospital stay than those with BL (median: 24 vs. 10 days, p < 0.001). On multivariable analysis, male gender (p = 0.002), higher DFA (p < 0.001) or CRP (p < 0.001) on POD3, lower albumin (p = 0.028) on POD3 were all found to be independent predictors of CR-POPF. A risk score based on these factors returned an area under the ROC curve of 0.78. Conclusions In patients with a confirmed pancreatic fistula it may be possible to differentiate between BL and CR-POPF using early postoperative variables, particularly DFA, serum albumin and CRP. Early identification of CR-POPF may allow earlier intervention to improve clinical outcomes.