Institutional learning curve for robotic pancreatoduodenectomy: a consecutive 100-case analysis

Robotic pancreatoduodenectomy (RPD) is a feasible but technically demanding procedure. Although surgeon-specific learning curves are well characterized, institutional curves remain underreported. This study aimed to evaluate the institutional learning curve for RPD using cumulative sum (CUSUM) and risk-adjusted CUSUM (RA-CUSUM) analyses.

We retrospectively reviewed 100 consecutive RPDs performed between September 2022 and November 2024 at a high-volume tertiary center. Clinical outcomes of patients who underwent RPD and open pancreaticoduodenectomy (OPD) were compared. Surgical failure was defined as conversion or Clavien-Dindo grade ≥ III complication. CUSUM and RA-CUSUM analyses were applied to identify learning curve thresholds.

Of the 508 PDs performed, 100 were robotic (84 hybrid, 16 totally robotic). Compared with OPD, RPD was associated with a longer operative time and fewer retrieved lymph nodes, but demonstrated similar R0 resection rates, incidence of major complications, and rates of clinically relevant postoperative pancreatic fistula, along with a shorter hospital stay. Surgical failure occurred in 32 cases, and independent risk factors included pylorus-resecting PD, high blood loss, and transfusion. The RA-CUSUM identified the institutional learning curve at 44 cases, with stabilization achieved by 76 cases. Beyond this threshold, the operative time and the incidence of major complications decreased significantly, and overall outcomes became comparable to those of OPD.

The institutional learning curve for RPD was achieved in approximately 44 cases. Despite a prolonged operative time, RPD exhibited comparable safety and oncologic adequacy to OPD, with a reduced hospital stay. These findings support the feasibility and safety of institutional adoption of RPD in high-volume centers.