Robotic versus open pancreatoduodenectomy after completion of the learning curve: a propensity score–matched and CUSUM analyses of perioperative outcomes

BACKGROUND: Robotic pancreatoduodenectomy (R-PD) has emerged as a minimally invasive alternative to open pancreatoduodenectomy (O-PD), although its role remains debated due to technical complexity and heterogeneous outcomes. This study compared perioperative outcomes between R-PD and O-PD in a high-volume center while minimizing the impact of the learning curve, and additionally evaluated the progressive improvement in surgical outcomes between the early and late phases of the robotic experience through comparative analyses and CUSUM assessment. METHODS: A single-center propensity score-matched retrospective cohort study was conducted using a prospectively maintained database of patients undergoing PD between January 2020 and December 2025. The most recent 50 consecutive R-PDs (late group, LR-PD) were compared with O-PD after 2:1 matching (matched open group, MO-PD) to balance baseline characteristics. Perioperative, postoperative, and pathological outcomes were compared, and multivariable analyses were performed to identify independent predictors of key outcomes. A subgroup analysis comparing the first 50 R-PDs (early group, ER-PD) with the subsequent 50 cases (LR-PD) was also performed, along with CUSUM analyses of outcomes showing differences between the two phases of the robotic experience. RESULTS: After matching, 100 MO-PD patients were compared with 50 LR-PD patients. LR-PD was associated with significantly lower estimated blood loss (EBL) (200 vs 300 mL; p = 0.0012) and shorter length of hospital stay (LOS) (12 vs 14 days; p = 0.038), but longer operative time (OT) (582 vs 362 minutes; p < 0.001). Rates of major complications (Clavien-Dindo ≥ III) and pancreas-specific complications were comparable between groups. A significantly lower incidence of postoperative lymphatic fistula was observed in the LR-PD (4 vs 15%; p = 0.045). At multivariable analysis, the LR-PD was independently associated with lower EBL (β - 116.80; p = 0.023), shorter LOS (β - 1.48; p = 0.041), and a reduced risk of postoperative lymphatic fistula (OR 0.30; p = 0.047), while remaining an independent predictor of longer OT (β 99.75; p < 0.001). Within the R-PD, OT was significantly shorter in the LR-PD compared with the ER-PD (582 vs 705 minutes; p < 0.001). CUSUM analyses identified inflection points at cases 51, 55, and 53 for OT, major complications, and clinically relevant PPH, respectively. CONCLUSIONS: In a high-volume setting after the learning curve, R-PD is a potentially safe and feasible alternative to O-PD, offering reduced EBL, shorter LOS, and lower incidence of lymphatic fistula, despite longer OT. Further prospective studies are warranted to confirm these findings.