Equivalent accuracy of image‐based and imageless robotic platforms in achieving planned femoral rotation in primary total knee arthroplasty: A CT‐based study

Purpose Robotic‐assisted total knee arthroplasty (TKA) is increasingly adopted worldwide, yet comparative data on the rotational accuracy of image‐based and imageless platforms remain limited. The aim of this study was to compare the accuracy of image‐based and imageless robotic platforms in achieving planned femoral rotation in primary TKA. We hypothesized that the two platforms would be equivalent, with any difference within ±1°. Methods This single‐surgeon retrospective cohort study included patients undergoing image‐based and imageless robotic‐assisted TKA for primary osteoarthritis. All patients underwent pre‐ and postoperative computed tomography (CT) scans. Native femoral rotation was defined using the transepicondylar axis (TEA)–posterior condylar axis (PCA) angle. Femoral rotation was planned intraoperatively relative to the PCA, and postoperative TEA–PCA was measured on CT. Accuracy was defined as the absolute difference between achieved and intended rotation. Measurements were independently performed by two blinded fellowship‐trained knee surgeons. Statistical analysis included two one‐sided tests (TOST) equivalence testing and analysis of covariance (ANCOVA). Significance was set at p < 0.05. Results A total of 150 TKAs were included (imageless 79, image‐based 71). Patients in the image‐based group were younger (67.7 vs. 71.3 years, p = 0.007), with no other baseline differences. The mean rotational error was 1.3 ± 1.0° for imageless and 1.2 ± 1.2° for image‐based TKA ( p = 0.346; effect size −0.09). Equivalence testing confirmed statistical equivalence, with the 90% confidence interval (–0.23 to 0.31) lying within the predefined ±1° margin ( p < 0.01). ANCOVA demonstrated a small adjusted difference of 0.51° ( p = 0.045), with minimal explained variance (adjusted R 2 = 0.01). Intra‐observer reliability was excellent (intraclass correlation coefficient [ICC] 0.91), while inter‐observer reliability was good (ICC 0.82). No differences were observed in patient‐reported outcomes, except lower pain scores in the image‐based group at 12 months. Conclusion Image‐based and imageless robotic platforms demonstrated equivalent accuracy in achieving planned femoral rotation in primary TKA. Any observed differences were not clinically meaningful. Level of Evidence Level III, retrospective comparative cohort study.