Learning curve and surgical time predictors in robot-assisted transvaginal natural orifice transluminal endoscopic surgery: a risk-adjusted cumulative sum analysis.

To identify factors associated with operative time and evaluate the learning curve of robot-assisted transvaginal natural orifice transluminal endoscopic surgery (RA-vNOTES) using conventional and risk-adjusted cumulative sum (CUSUM) analyses. In this single-surgeon retrospective study, a total of 116 patients who underwent RA-vNOTES for benign gynecologic disease between December 2021 and August 2024 were included. To evaluate the learning curve, both conventional CUSUM and RA-CUSUM analyses were performed. Expected operative time for each case was estimated using a multivariable linear regression model including body mass index (BMI), uterine weight (per 100 g), and parity. RA-CUSUM was calculated as the cumulative sum of observed minus expected operative time. Multivariable linear regression models identified higher BMI (p = 0.001) and greater uterine weight (p < 0.001) as independent predictors of prolonged total operative time. No significant predictors were identified for docking time or console time. Transition plots and conventional CUSUM demonstrated progressive procedural improvement over time. RA-CUSUM analysis identified peak learning points at approximately 31 cases for docking time, 32 cases for console time, and 30 cases for total operative time, indicating stabilization after approximately 30-32 cases. Operative time stabilization in RA-vNOTES was achieved after approximately 30 to 32 cases in this single-surgeon experience. Higher BMI and greater uterine weight are significant predictors of longer total operative time. RA-CUSUM provides a more accurate assessment of the learning curve by accounting for patient-related variability, which may inform surgical training strategies and case selection during the implementation of RA-vNOTES.