High accuracy and learning curve improvement of augmented reality in total knee arthroplasty: A single-centre study on 157 patients.

Purpose: Augmented reality (AR) technology is emerging as a viable alternative for improving surgical precision in total knee arthroplasty (TKA). While previous studies have assessed AR accuracy, limited research has explored its learning curve and long-term performance. This study evaluates the accuracy of an AR-assisted system in achieving pre-planned osteotomies and investigates potential improvements over time with increased surgical experience. Methods: A total of 157 patients underwent TKA using an AR-based guidance system, with preoperative planning defining varus/valgus and posterior slope angles. The achieved angles were compared to planned values using two error assessment metrics. Additionally, the data from the first 76 procedures were compared to the last 81 to assess a learning curve effect. Surgical time and blood loss were also evaluated. Results: The AR system maintained high accuracy, with mean deviations of 0.47 ± 0.54° for varus/valgus and 0.76 ± 0.74° for posterior slope. Errors remained below 1° in the majority of cases (98.1% for varus/valgus and 86.0% for posterior slope). Over time, a learning effect was observed, with mean errors decreasing by up to 44.6% for varus/valgus and 8.6% for posterior slope. No significant differences were found in surgery duration or blood loss between the AR and control groups. Conclusions: This study confirms the reliability of AR-based guidance in TKA, demonstrating its ability to achieve precise osteotomies with minimal intraoperative variability. Additionally, the results suggest a learning curve effect, with improved accuracy over time. These findings support AR as a promising, cost-effective alternative to robotic navigation in knee surgery. Level of Evidence: Level II.