• minimally invasive pancreatic surgery
• pancreas
• pancreatic surgery
• robot assisted pancreatic surgery
• robot assisted surgery

Introduction: to analyze the learning curve of robot-assisted pancreatoduodenectomy (r-PD) and the effects of training on short-term prognosis.
Method: the learning curve of r-PD performed by a single highly experienced surgeon in a high-volume center was determined using the cumulative sum method, based on operative time (OT) and length of hospitalization (LOH). Data were extracted from a prospectively maintained database and analyzed retrospectively considering all events occurring within 90 days of index operation. The outcomes were studied into three different groups based on three phases of CUSUM-LOH.
Results: We analyzed 298 r-PD. A three phases model was observed for both outcomes. Considering OT there was a drop-down after 37 procedures (median OT 550 min. vs. 505 min., p=0.0017) and an increase after 233 procedures (median OT 505 min. vs. 547 min., p=0.0001). Considering LOH, the plateau was achieved after 130 procedures and a drop-down after 183 procedures (median LOH: 18 days vs. 15 days, p= 0.02) without differences in OT and conversion rate; conversely the vein resection rate increased (6.9% vs. 7.6% vs. 14.8%, p=0.04). The three groups didn’t shown difference considering severe post-operative complications and mortality rate, clinically-relevant POPF rate and failure-to-rescue rate.
Conclusion: OT is initially reduced (37 procedures) and subsequently increased (233 procedures) due to the higher procedure’s difficulty (represented by the vascular resections rate). Proficiency regarding post-operative outcomes (LOH) was obtained later (183 procedures). Despite more challenging procedures the mastery by the surgeon and the team allows complications rate to remain unchanged and shorten the LOH.