RÉSUMÉ
Background: Clinically relevant postoperative pancreatic fistula (CR-POPF) remains the most common cause of perioperative morbidity following pancreatico-duodenectomy (PD). Early and accurate prediction of CR-POPF can be helpful in postoperative drain management as well as stratifying patients for enhanced recovery protocol after surgery. Both fistula risk score (FRS) and postoperative drain amylase levels have been analyzed in past. However, currently there is no clear consensus regarding the ideal predictor. Present study sought to assess the utility of postoperative day 3 drain amylase (POD-3DA) level as a predictor of CR-POPF in comparison with FRS.Methods: A retrospective analysis was done on 57 patients who underwent PD at our institute between 2014 to 2018. POPF was defined and graded in accordance with ISGPF definition. Receiver operating characteristic (ROC) analysis predicted a threshold of POD3DA >486 IU/l associated with CR-POPF. Sensitivity, specificity and odds ratios with 95%CI calculated and ROC curves were plotted for POD3DA of ≥500 IU/l and FRS (negligible/low vs. moderate/ high) as predictors of CR-POPF.Results: Incidence of POPF and CR-POPF was 63% and 32% respectively. Sensitivity and specificity of POD3DA ≥500 and moderate/high FRS for predicting CR-POPF were 83%, 79% & 78%, 51% respectively. Difference between ROC area under the curve (AUC) for POD3DA ≥500 IU/l (0.868) and FRS (0.692) was significant (p=0.028). Combining FRS and POD3DA ≥500 IU/l improved specificity (87%) at the cost of sensitivity (67%). The negative predictive value of POD3DA <500 IU/l and negligible/low FRS were 91.2% and 83.3% respectively.Conclusions: POD3DA level greater than 5 times of upper normal range is more precise at predicting CR-POPF, hence clinically more reliable for drain and postoperative management.