Nationwide outcomes of 1000 robotic pancreatoduodenectomies across the four phases of the learning curve

Abstract Background Robotic pancreatoduodenectomy (RPD) is increasingly being implemented to enhance patient recovery, but it is unclear what happens to patient outcomes after the initial learning curve. The aim of this study was to evaluate the first 1000 consecutive RPD performed in the Netherlands. Methods A nationwide analysis of patients who underwent RPD in 13 centres (March 2016–August 2023) from the Dutch Pancreatic Cancer Audit was performed. Patients were grouped based on published learning curve cut-offs (phases 1–4): 1–15, 16–62, 63–84, and >84 RPD per centre respectively. Outcomes were compared between the four learning curve phases. Ideal Outcome rates were used to compare outcomes between centres. Results Overall, 1000 patients after RPD were included. The conversion rate was 10.1%, the rate of Clavien–Dindo complications of grade ≥III was 41.3%, the rate of postoperative pancreatic fistula of grade B/C was 24.4%, and the rate of in-hospital/30-day mortality was 3.9%. Of the patients, 71.1% had a high updated alternative fistula risk score. Improvements between the phases were found for five outcomes: median operating time (420, 360, 349, and 369 min respectively; P < 0.001), conversion rate (21.7%, 10.0%, 2.8%, and 7.5% respectively; P < 0.001), rate of delayed gastric emptying (DGE) of grade B/C (32.3%, 22.6%, 15.4%, and 20.2% respectively; P = 0.003), reoperation rate (9.9%, 11.3%, 9.8%, and 4.9% respectively; P = 0.026), and median duration of hospital stay (12, 11, 10, and 10 days respectively; P = 0.035). The rate of Clavien–Dindo complications of grade ≥III and the rate of in-hospital/30-day mortality remained stable. The Ideal Outcome rate (mean 47%) did not differ between centres. Conclusion Across four learning curve phases in a nationwide cohort, improvements were observed for operating time, conversion rate, rate of DGE of grade B/C, reoperation rate, and duration of hospital stay.