Dynamic joint balancing provides consistent gap prediction without a learning curve in robotic-assisted total knee arthroplasty.

Dynamic joint balancing (DJB) in robotic-assisted total knee arthroplasty (RATKA) allows surgeons to simulate implant positioning and predict soft tissue balance intraoperatively before bone resection. Although virtual gap (VG) estimation is integral to this process, its accuracy in predicting the final gap (FG) after implantation remains uncertain. We conducted a retrospective analysis of 77 knees in 61 patients undergoing RATKA with the MAKO system. VG was recorded at four positions (medial/lateral in 10° extension and 90° flexion) prior to bone resection. FG was measured post-implantation using a 9-mm trial insert. Correlations between VG and FG were assessed, along with the frequency of VG-FG discrepancies ≥ 3 mm. We further assessed relationships with coronal implant alignment and evaluated the presence of a learning curve. VG and FG were moderately to strongly correlated at all measurement points (r = 0.50-0.71). FG exceeded VG by 0.6-1.3 mm on average, with the largest discrepancies in flexion lateral gaps. Gap errors ≥ 3 mm were significantly more frequent in flexion (8.4%) than extension (1.3%) (p = 0.0036). No significant association was found between implant alignment error and VG-FG discrepancy. Logistic regression revealed no learning curve effect for either surgeon. DJB-based VG estimation provides a reliable, reproducible intraoperative reference in RATKA. However, consistent overestimation-particularly in flexion-suggests the need for improved modeling or measurement techniques. DJB's independence from a learning curve supports its value in standardizing soft tissue balancing across surgical experience levels.