Risk-adjusted learning curve for robotic colorectal surgery in an educational setting: a single-center consecutive series.

Learning-curve assessments for robotic colorectal surgery can be biased by temporal changes in case-mix and by increasing trainee participation. We evaluated a risk-adjusted learning curve using residual-based cumulative sum (CUSUM). We retrospectively analyzed 270 consecutive robotic colorectal resections performed between July 2019 and December 2025 at a single center. Operative time was modeled using multivariable linear regression including age, body mass index, ASA class, clinical T/N/M category, neoadjuvant chemotherapy, selective lateral pelvic lymph node dissection (LPLND), multivisceral resection, console switch, and trainee involvement. Risk-adjusted CUSUM was calculated as the cumulative residuals (observed minus predicted operative time), and change points in the CUSUM trajectory defined proficiency phases. Median operative time was 279 [238–334] min. Overall complications (Clavien–Dindo > 0) occurred in 49/270 (18.1%), major complications (CD ≥ III) in 20/270 (7.4%), anastomotic leak in 13/270 (4.8%), conversion in 1/270 (0.4%), and transfusion in 7/270 (2.6%). Risk-adjusted CUSUM identified a change point at case 72, defining two phases. In this consecutive series, risk-adjusted CUSUM identified a phase transition in operative efficiency while short-term outcomes remained acceptable in the context of increasing trainee participation. Risk-adjusted learning-curve assessment may support benchmarking and evaluation of training programs in real-world practice.