A decade-long prospective analysis of the learning curve after 435 robotic pancreaticoduodenectomies from a single tertiary institution.

INTRODUCTION

Robotic pancreaticoduodenectomy (PD) has emerged as a leading approach in complex pancreatic surgery, but the path to proficiency is steep. While early learning curve assessments are available for simpler robotic procedures, few studies quantify the extended time and caseload needed to achieve consistency and mastery in robotic PD.

METHODS

With IRB approval, we prospectively followed 435 patients who underwent robotic PD between 2013 and 2024. The learning curve was assessed using Cumulative Sum (CUSUM) analysis based on operative time. Data are presented as median (mean ± SD), and statistical significance was set at p ≤ 0.05.

RESULTS

CUSUM analysis identified the turning point for consistent procedural proficiency at 88 cases. Patient characteristics included a median age of 70 (68 ± 10.9) years, BMI of 27 (27 ± 5.1) kg/m, and a predominance of males (54%). A total of 253 patients (58%) had previous abdominal surgery, 36 (8%) had undergone neoadjuvant therapy, and average ASA class and Charlson Comorbidity Index were 3 (± 3) and 5 (± 1.5), respectively. Over time, operative efficiency translated into improved outcomes: unplanned conversions to open surgery declined significantly (p < 0.0001), estimated blood loss decreased (p = 0.04), and hospital stays shortened (p = 0.008). Interestingly, despite growing proficiency, operative times increased (p < 0.0001), reflecting a shift toward more technically demanding cases. A significant rise was also observed in the proportion of patients with prior abdominal surgeries treated after the learning curve milestone (p = 0.007).

CONCLUSIONS

True mastery in robotic PD demands extended experience, with significant procedural gains occurring after the 88-case mark. This extended learning curve should be acknowledged when designing training programs and evaluating surgical outcomes in robotic pancreatic surgery.