Multifactorial analysis of the learning curve for robot-assisted laparoscopic biliopancreatic diversion with duodenal switch

Objective: To assess the impact of surgeon, patient, and case-specific factors on the learning curve of robot-assisted laparoscopic biliopancreatic diversion with duodenal switch (RA-LBPD/DS). Background: The BPD/DS has better resolution of diabetes and hypercholesterolemia, and the best long-term weight loss compared to the laparoscopic gastric band or the Roux-en-Y gastric bypass. Despite excellent results, the BPD/DS is least commonly performed because of greater malabsorption, longer operative duration, and higher technical complication rates. A reduction in technical complications and operative duration will enable the BPD/DS to be offered more frequently. Methods: Consecutive patients (N = 120) undergoing RA-LBPD/DS between October 2000 and August 2008 were analyzed using univariate and multivariate logistic regression to determine the influence of surgeon and patient factors on complications and operative duration. Independent variables were case number, age, gender, body mass index, American Society of Anesthesiologists (ASA) score, difficult anatomy, and need for extensive adhesiolysis. Dependent variables were complications (leaks, bleeding, and conversion) and operative duration. The best-fit model predicted the risk factors for complications, and a risk-adjusted cumulative sum analysis estimated the learning curve. Results: Operative duration decreased an average of 3 minutes with each successive case (P < 0.001, R = 0.63) and with patient's female gender. Adhesiolysis, difficult anatomy, liver biopsy, and higher ASA score increased operative duration. The incidence of high blood loss (13.3%), conversion (2.2%), and leaks (5.8%) were experienced by a total of 22 patients (18.3%). There was no mortality. Complications declined after 50 cases and were strongly predicted by increasing surgeon case number. Conclusions: The learning curve for the RA-LBPD/DS is 50 cases. Risk factors influencing outcomes were identified.