Platform-specific learning curve and early tibial radiographic agreement of ROSA®-assisted medial unicompartmental knee arthroplasty in an experienced high-volume surgeon: a consecutive case series

Purpose To evaluate the platform-specific learning curve and early tibial radiographic agreement of imageless ROSA ® -assisted medial unicompartmental knee arthroplasty (UKA), with short-term clinical findings assessed exploratorily. Methods This retrospective consecutive case series included 30 patients who underwent ROSA ® -assisted medial UKA with the Persona ® Partial Knee implant between January and April 2026 at a high-volume arthroplasty referral centre. A continuous learning-curve cumulative summation (LC-CUSUM) analysis of total surgical duration was performed against an independently defined surgeon-specific historical benchmark derived from institutional operative-register skin-to-skin times for manual UKA performed by the same surgeon (20 ± 4 min). Adequate performance was defined as a mean duration of 25 min, inadequate performance as 30 min, and the reference standard deviation as 4 min. Type I and type II error targets were 0.05 and 0.10, respectively; the decision boundary was calibrated by Monte Carlo simulation over a 30-case monitoring horizon. Robotic time and a supportive comparison of cases 1–5 versus 6–30 were secondary efficiency analyses. Postoperative radiographic measurements were performed by an observer blinded to the planned and robot-validated ROSA ® values. Agreement was quantified using Bland–Altman bias and 95% limits of agreement. Results Mean robotic time was 14.5 ± 2.4 min, mean non-robotic time was 8.9 ± 3.1 min (range, 4.1–15.6), and mean total surgical duration was 23.4 ± 2.3 min. The continuous LC-CUSUM score crossed the calibrated decision boundary (h = 4.5895) at case 6. Robotic time was 16.0 ± 3.8 min in cases 1–5 and 14.2 ± 2.0 min in cases 6–30 ( p = 0.787); corresponding total durations were 24.8 ± 3.8 and 23.1 ± 1.9 min ( p = 0.385). Bland–Altman bias for postoperative minus planned measurements was + 0.13° for tibial coronal alignment and + 0.10° for tibial slope; bias for postoperative minus robot-validated HKA was + 0.34°. All 30 postoperative radiographic measurements were within ± 2° of the planned target for both tibial coronal alignment and slope (100%; exact 95% CI, 88.43%–100%). Conclusions The predefined platform-specific temporal-performance standard was reached after six cases by an experienced high-volume UKA surgeon. Tibial planned-to-postoperative deviations were small in this consecutive series, while the findings remain conditional on the specified benchmark and require external validation.