Inherent Bias in ROSA^® Zimmer Biomet Pre-Op Planning Using 2D to 3D X-Atlas^® Coronal Knee Axis Measurement

Background: Robotic assistance is considered capable of improving precision and outcomes of total knee replacement. We assessed the inherent biases, pre-procedural planning accuracy using 2D to 3D X-Atlas^®, and final knee axis outcomes of the ROSA^® Knee System (Zimmer Biomet, Warsaw, IN, USA). Methods: A total of 55 patients who underwent robotic-assisted knee replacement using ROSA^® Knee System (Zimmer Biomet, Warsaw, IN, USA) at a single center were included. Pre-procedural measurements performed by ROSA were compared to those performed by senior consultants. Component sizes predicted by ROSA^® were compared to those implanted. A final axis measurement was taken during the procedure. Results: Femur components were exactly matched in (83.64%) cases, accurately matched in a further 8 (14.55%), and inaccurately matched for only 1 (1.82%). Tibial component sizes were exactly matched by the planning for 39 (70.91%), accurately for 12 (21.82%), and inaccurately for 4 (7.27%). ANOVA did not show statistically significant differences between the predicted and implanted femur (p = 0.96) nor the tibia components (p = 0.27). We show that ROSA^® pre-procedural planning has a statistically significant bias (p = 0.001), with a deviation of 0.83 degrees into varus, when assessing the knee axis in the coronal plane, compared to senior consultant measurements. The average of the final coronal knee axis was 0.37 degrees in varus (SD = 2.49). Conclusions: ROSA^® accurately predicts implanted component sizes. Despite the small and statistically significant varus bias in initial knee axis assessment, the system results lay within the ±3° of neutral knee axis, which is the widely accepted knee replacement standard.