Learning Curve From 450 Cases of Robot-Assisted Pancreaticoduocectomy in a High-Volume Pancreatic Center

We aimed to describe our experience and the learning curve of 450 cases of robot-assisted pancreaticoduodenectomy (RPD) and optimize the surgical process so that our findings can be useful for surgeons starting to perform RPD.Robotic surgical systems were first introduced 20 years ago. Pancreaticoduodenectomy (PD) is a challenging surgery because of its technical difficulty. RPD may overcome some of these difficulties.The medical records of 450 patients who underwent RPD between May 2010 and December 2018 at the Shanghai Ruijin Hospital were retrospectively analyzed. Operative times and estimated blood loss (EBL) were analyzed and the learning curve was determined. A cumulative sum (CUSUM) analysis was used to identify the inflexion points. Other postoperative outcomes, postoperative complications, and long-term follow-up were also analyzed.Operative time improved graduallyovertimefrom405.4 ± 112.9 minutes (case 1-50) to 273.6 ± 70 minutes (case 301-350) (P < 0.001). EBL improved from 410 ± 563.5 mL (case 1-50) to 149.0 ± 103.3 mL (case 351-400) (P< 0.001). According to the CUSUM curve, there were 3 phases in the RPD learning curve. The inflexion points were around cases 100 and 250. The incidence of pancreatic leak in the last 350 cases was significantly lower than that in the first 100 cases (30.0% vs 15.1%, P = 0.003).RPD is safe and feasible for selected patients. Operative and oncologic outcomes were much improved after experience of 250 cases. Our optimization of the surgical process may have also contributed to this. Future prospective and randomized studies are needed to confirm our results.