SRS391 - Global outcomes after robotic left pancreatectomy: learning curve, morbidity, and oncological quality

Abstract Background Robotic left pancreatectomy (RLP) is increasingly adopted worldwide. Current guidelines suggest a learning curve of 21 cases, yet its impact on major morbidity and oncological quality remains uncertain. Methods A retrospective multicentre international study was undertaken across 10 centres in the UK, Spain, Germany, the USA, and Australia (2014–2025). Consecutive patients undergoing RLP were categorised into competency (first 21 cases per centre) or proficiency phase (>21). The primary outcome was major morbidity (Clavien–Dindo grade ≥ IIIa), and secondary outcome was the R1-margin rate. Mixed-effects logistic regression and risk-adjusted cumulative sum (RA-CUSUM) analyses were performed. Results 521 patients underwent RLP, including 200 (38.4%) in the competency phase and 321 (61.6%) in the proficiency phase. Median age was 63 years (IQR 54–73), and 56.4% were female. Rates of major morbidity were similar between phases (14.5% versus 14.0%, P = 0.981), as were R1 margin rates (21.6% versus 19.7%, P = 0.092). Independent predictors of major morbidity included male sex (OR 1.76, P = 0.037), preoperative pancreatitis (OR 2.23, P = 0.031), and multivisceral resection (OR 2.29, P = 0.034). Among patients with adenocarcinoma (n = 110), prolonged operative time (OR 1.39 per hour, P = 0.004) and nodal positivity (OR 2.62, P = 0.047) predicted R1 resection, while learning phase was not associated with margin status. RA-CUSUM demonstrated no excess morbidity over time in any centre. Conclusions In this first multicentre international analysis, the learning phase of RLP did not increase the risk of major morbidity or compromise oncological quality. These findings support the safe global adoption of RLP with appropriate training and case selection.