Validation of day 1 drain fluid amylase level for prediction of clinically relevant fistula after distal pancreatectomy using the NSQIP database

Abstract

Background: The role of postoperative day 1 drain fluid amylase level in predicting clinically relevant postoperative pancreatic fistula is under investigation. In a previous multicenter study conducted on 338 patients undergoing distal pancreatectomy, day 1 drain fluid amylase level has been correlated to the development of a clinically relevant pancreatic fistula and an amylase value of 2,000 U/L was found to be most predictive of the development of clinically relevant postoperative pancreatic fistula. Our objective was to validate the previously established cutoff level for drain fluid amylase on postoperative day 1 after distal pancreatectomy as a predictor for clinically relevant postoperative pancreatic fistula using a different patient population from the National Surgery Quality Improvement Program database. Methods: We studied all patients undergoing distal pancreatectomy from the National Surgery Quality Improvement Program pancreatectomy specific participant use file from 2014 to 2016. We applied the day 1 drain fluid amylase level of 2,000 U/L cutoff to divide patients into 2 groups and compared clinical outcomes in both groups. Among patients with a day 1 drain fluid amylase level < 2,000 U/L, we compared the patient characteristics of those who developed a clinically relevant postoperative pancreatic fistula to those who did not. Finally, to independently validate the previously defined day 1 drain fluid amylase level, we proceeded to determine the optimal cutoff value of day 1 drain fluid amylase level, which can be used as a predictor for the development of clinically relevant postoperative pancreatic fistula after distal pancreatectomy using a receiving operating characteristic curve. Results: A total of 1,007 patients underwent distal pancreatectomy. The mean day 1 drain fluid amylase level was 4,290.04 ± 8,492.35 U/L. Clinically relevant postoperative pancreatic fistula occurred in 203 patients (20.2%). Using bivariate analysis, patients with day 1 drain fluid amylase level ≥ 2,000 U/L were more likely to develop clinically relevant postoperative pancreatic fistula (32.5% vs 11.25%, P <.0001), to have a higher mean number of days before drain removal (8.83 vs 5.59, P <.0001), to have a drain 30 days postoperatively (12.59% vs 3.63%, P <.0001), and to undergo percutaneous drainage (13.75% vs 9.69%, P =.04). Among patients with a day 1 drain fluid amylase level < 2,000 U/L, 11% of patients went on to develop a clinically relevant postoperative pancreatic fistula. Analysis of this subgroup of patients did not identify any discernable preoperative characteristics that were predictive of this complication. Application of maximal Youden index calculated the day 1 drain fluid amylase level value at 2,000 U/L with a sensitivity of 67.98% and a specificity of 63.81% for clinically relevant postoperative pancreatic fistula, with a positive predictive value of 32.17%, a negative predictive value of 88.75%, and a Youden index of 0.32. Conclusion: Using a different population of patients and a different data set as well as an independent analysis, we successfully validated a day 1 drain fluid amylase level of 2,000 U/L as striking the best balance in terms of sensitivity and specificity for the detection of clinically relevant postoperative pancreatic fistula. The identified cutoff might be employed in the design of a trial of early drain removal in patients undergoing distal pancreatectomy. © 2018 Elsevier Inc.