Multivariate logistic regression analysis for prediction of clinically relevant pancreatic fistula in the early phase after pancreaticoduodenectomy

Background Postoperative pancreatic fistula ( PF ) remains a major complication after pancreaticoduodenectomy ( PD ). We aimed to investigate the predictors of clinically relevant PF after PD . Methods We retrospectively analyzed the predictive factors of relevant grade B / C PF using logistic regression analysis of 100 consecutive patients who underwent PD . PF was defined in accordance with the International Study Group on PF ( ISGPF ). Results White blood cell count ( WBC ) of 73.6 × 10 2 /μl, C ‐reactive protein ( CRP ) of 9.3 mg/dl and amylase value in drains (d‐amylase) of 647 U / I on postoperative day ( POD ) 4 were proposed as the cut‐off values for predicting grade B / C PF with high accuracy by the receiver operating characteristic ( ROC ) curve. Multivariate logistic regression analysis revealed that the three factors as significant predictive factors and the predicted probability of detecting grade B / C PF was calculated by the following formula; P = 1/[1 + exp{‐(2.033 × WBC +3.269 × CRP +2.698 × d‐amylase‐4.122)}]. P > 0.5 indicates the prospective incidence of the PF . When the cut‐off values of the three significant predictors were substituted into the formula, P always showed above 0.5 if more than two predictors were above their cut‐off values, indicating a high probability of grade B / C PF . Conclusions White blood cell count, CRP and d‐amylase on POD4 were predictive factors for clinically relevant PF after PD . These findings indicate that our formula is useful for management of drain after PD .