Implementation of computer of technology in TKA is a constantly evolving process as CAS continues to develop in an effort to optimize surgical outcomes, patient safety, efficiency, and cost effectiveness [2,3,4,5,6,7,8,9]. Recent literature suggest that the addition of such mechanical assistance be it through haptic or passive guidance with robotic technology may improve these aspects of CAS. There has also been suggestion of reduced periarticular soft tissue damage [2, 3, 9], reduced postoperative pain, reduced bleeding, decreased opiate requirement, shorter length of stay, and greater patient satisfaction [2, 3, 9]. The ROSA has been reported to provide some of these improvements [10, 11]. Through a series of waypoint acquisitions to orient the computer to either image based from preoperative radiographs or non-image based, intraoperative stress data can be used to preplan resections. Using this surgeon driven selection of desired implant placement, the ROSA system can be programmed to place appropriate cutting guides in position. It does not however, aid in the guidance of the actual resection through haptic limitations of blade, bure, or a drill placement. As higher costs and time expenditure are associated with using CAS, orthopedic surgeons must validate worthy improvement to justify adoption of new technology , though recent literature suggest that these costs may be offset by cost savings from shorter hospital stays, reduced postoperative costs, and reduced healthcare resource utilization [2, 3, 9]. With proposed improvements in surgical accuracy and benefits in outcomes and cost efficacy, robotic-assisted surgery continues to be a worthwhile endeavor to explore. Additionally, wider adoption and experience with the technology may continue to improve outcomes and cost benefits associated with robotic surgery. The purpose of this study was to assess one specific aspect of accuracy of ROSA by comparing the desired mechanical axis calculated intraoperatively by the ROSA computer versus the actual mechanical axis measured post-operatively with long-length knee radiographs.
This study calls into question strictly the accuracy of a novel robotic technology, ROSA, and provides insight into how it can be further improved or adjusted to justify the costs associated with it. In the anterior–posterior resection planes (HKA, α, and β), accuracy was exceptional. 91.89% of patients were correct to within 2° and 100% within 3° of the α angle. 72.97% of patients were correct to within 2° whereas 89.19% were correct to within 3° of the HKA angle., and finally 75.68% were within 2° and 91.89% within 3° on the β angle. However, the angles measured in the sagittal planes did not measure as consistently compared to the coronal plane resections. 51.43% of patients were within 2° on the γ angle, or sagittal femoral angle, whereas 77.14% of patients were within 3° of correct alignment. 57.14% of patients were correct to within 2° on the δ angle, or tibial slope and 74.29% were within 3° of correct alignment. These results differ from previous studies by Seidenstein et al. , who reported 92.9, 100, 71.4, 100, and 100% of planned HKA, α, β, δ, and γ knee angle respectively within 2° of the post-op angles, and 100%, 92.9%, 100%, 100%, 100% within 3° respectively. Another similar study by Parratte et al.  reported 97%, 100%, 100%, 87%, 100% of planned HKA, α, β, δ, and γ within 2° of the post-op angles, and 100%, 100%, 100%, 97%, 100% within 3°. These previous studies were performed on cadavers using the same CAS system, ORTHOsoft Total Knee Navigation system (Zimmer Biomet, Warsaw, IN). The ORTHOsoft Total Knee Navigation system is an infrared camera-based imageless navigation system that has reported accuracy of less than 1 mm and 1° . Given this accuracy and our results in the sagittal plane, there appears to be a disconnect in the transfer of data from the navigation system to the cuts. Whether this disconnect is based of the navigation system itself, the act of calibrating anatomic landmarks, or the movement of the cutting jig to the cutting surface of the knee, is difficult to ascertain. However, further research is required to evaluate these variables and larger studies are required to determine the repeatability of these results.
The HKA or mechanical angle is more often reported in outcomes studies as the litmus test for accuracy and likewise survivability and functional outcomes. Inaccuracies derived on the sagittal plane may likely be more tolerated. From a cosmetic standpoint, such as a hyperflexed femoral component or altered tibial slope resulting in a relatively normal function in short-term clinical outcomes. However, a singular malpositioned femoral or tibial component results in imbalance flexion/extension gap, resulting in motion loss, midflexion instability, and ultimately early failure. While the ROSA is an accurate option, with regards to the mechanical/HKA, α, and β angles, it fell short in accuracy which was evidenced by the decrease accuracy percentages of the δ and γ knee angles. Other studies on CAS system accuracies have been described and shown that they have improved accuracy in these planes using both CT, and CT free systems [13,14,15].
There were several short falls and limits of this study. The first limitation is the use of radiographs which fall short in capturing rotation abnormalities as it is only a 2-dimensional view to measure a 3-dimensional construct. Additionally, although our technicians are thoroughly and skillfully trained, the x-ray technician taking each radiograph was not controlled, leaving room for variability in the results. The gold standard for accuracy is spiral CT imagery which is extremely expensive, difficult to attain, and more importantly exposes the patients to excessive radiation. That being said, the bulk of studies still rely on 36 inch imagery which given the comparison and literature seems to be a fair assessment of this study against those of others.
Additionally, this study was a sequence non-rehearsed start-up study with the senior author having had 20 years of experience in computer-assisted surgery (CAS). However, the surgeon was not familiar with the ROSA system when he started in an attempt to get a real world experience of the typical surgical expected outcomes rather than correcting and making up for deficiencies and technical expertise and skills for instance, we noted during the placement of pins on the ROSA system a ‘pin drift’ occurs as a result of the potential inefficiencies in the haptics installed in the ROSA system to hold it steady. Even in the best of circumstances, in hard bone with geometric curvatures, a pin skid can potentially occur unbeknownst to the artificial intelligence and haptic technology of the ROSA system. The haptics that control the rigidity of the arm based off of the servos which are electronically controlled stabilizers did not appear to have complete resistance to this deficiency. This error then follows throughout the entire procedure and in the end reflects itself at the end in the inaccuracies of the proposed planned angles of the knee and what was actually obtained in post-operative radiographic imaging. While in the latter series of cases this pin skid was attempted to be corrected, there did not appear to be a significant difference in the accuracy obtained in the later cases over those of the first. Therefore the significance of this while trying to reconcile inaccuracies may or may not be the root cause of this downfall. Finally, 36 inch long films taken which are the standard for accuracy have their own inherent limits. While utilizing the best of radiographic techniques, rotation and inconsistencies exist in any type of radiographic follow-up inherent with this type of measurement. The fact that a number of patients were used as well as this study being compared to previous studies using the same techniques will tend to normalize this variance.