Predict the failure of an enhanced recovery protocol after pancreaticoduodenectomy

Abstract

Introduction: Few is known about patients' compliance to enhanced recovery protocols (ERp) for pancreaticoduodenectomy (PD). We aim to evaluate the problem and develop a tool to identify patients who will fail to comply with the ERp. Methods: From 1/1/2014 to 01/01/2016 205 consecutive patients underwent a PD in our center and were managed accordingly to an ERp.Postoperative complications were graded using Clavien-Dindo classification (grade≥3=major).Perioperative data were prospectively collected. Items failed were define as: no active deambulation postoperative day (POD) 1; < 4 hours out of bed POD2; removal of NGT and bladder catheter after POD1 and 3 respectively; reintroduction/suspension of oral feeding after POD4;intravenous infusions after POD4. Univariate/multivariate analysis and regression models were conducted.Data are presented as percentage, mean (CI). Results: Observing the mean number of failed items in patients without or with postoperative complications (1.13(0.88-1.37) vs 2.53(2.25-2.81);p< 0,001), major complications (1.66(1.446-1.88) vs 3.15(2.64-3.67);p=0,001) and prolonged length of stay, we define the failure of the ERp as the failure of two of more items. 56.6% failed the ERp. We create a predictive model considering age,BMI,operative time and the consistency of the pancreatic stump (OR 1.031,p =0.012;OR 1.111.p=0.031;OR 1.004,p=0.022;OR 2.892,p=0.002 respectively). We attributed a score from regression coefficients to each variable.Overall patient's score predict the failure of the ERp (AURC 0,7476(0,678-0,816)). The score showed a good agreement between predicted and observed proportions of failures (Hosmer-Lemeshow test;p=0.7127). Conclusion: Despite the advantages of ERps,still half of patients after PD are not able to comply with an ERp.ERp failure seems to be predictable.Can high risk patients benefit from a specific ERp?