Safe and effective use of active robotics for TKA: Early results of a multicenter study

Abstract

Background

A novel active robotic system for total knee arthroplasty (TKA) performs automated milling of bone surfaces. Study objectives were to assess system safety and effectiveness in a US population.

Methods

A multicenter clinical trial was conducted, following 115 patients for at least 6-months. A pre-defined list of robot-related adverse events was used to evaluate safety. Efficacy was assessed radiographically comparing planned versus achieved coronal limb alignment.

Results

No pre-defined adverse events occurred and postoperative limb alignment more than ±3° from plan occurred in 11.2 % of cases.

Conclusion

Active robotics for TKA is safe and effective as demonstrated in this trial.

1. Introduction

Primary total knee arthroplasty (TKA) is a successful operation for the treatment of symptomatic degenerative knee arthritis. Most patients experience decreased pain, improved function, and increased quality of life after surgery with greater than 95 % implant survival after ten years.¹ Accurate limb and implant alignment are critical factors for a successful procedure.^2,3 Despite advancements, lower limb malalignment and instability, which may be related to component position, remain common causes of failure leading to revision surgery.⁴

Literature is controversial regarding the effect of malalignment on patient outcomes, with studies showing suboptimal component placement to be associated with an increased rate of polyethylene wear, an increased rate of revision arthroplasty, and lower knee function.^4,5 Aseptic loosening, instability, and stiffness account for over 50 % of revisions within the first two years post TKA,⁶ but it is unclear how many of these complications are related to subtle differences in three-dimensional (3D) component position and fit. A possible reason for the disagreement on the role of malalignment on patient outcomes may be the implant design, where the kinematics of particular knee designs may perform poorly in classic mechanical alignment,⁷ as potentially postulated for single radius cruciate retaining knee designs. It is safe to say that there is no gold standard for alignment. Irrespective of the alignment target, however, it remains important to hit the chosen target as accurately as possible. Robotic assistive systems aim to reduce errors in surgical technique by allowing the surgeon to plan the implant positions and sizes based on a 3D model of the knee, and by assisting the surgeon in bone cutting.

An active robotic system (ARoS) for TKA automatically executes planned bone cuts without human guidance of the cutting tool during bone preparation. Active robotics for TKA has been available since 2000 with the introduction of the ROBODOC®, which has been used in more than 8000 interventions. An updated version of the ROBODOC®, the TSolution One (TS1, THINK Surgical Inc., Fremont, CA), has recently become available after receiving European Conformity (CE) clearance in 2018 and Food and Drug Administration (FDA) clearance in 2019.

A multicenter investigational device exemption (IDE) clinical trial was conducted to evaluate the safety and effectiveness of the TS1 robot for TKA, to allow for its use in the US. The study had two primary objectives: 1) To determine the safety of the system by comparing the incidence of the composite adverse events (AEs) to literature control values for manual TKA, and 2) To determine the effectiveness of the system by comparing lower limb malalignment rate in the coronal plane to literature control values for manual TKA. The study also evaluated secondary measures of safety and effectiveness such as bleeding complications, femoral and tibial component positional accuracy in 3D, and patient reported outcome measures. This report details the first multicenter experience with TS1 for TKA.

2. Materials and methods

Eight surgeons from six US sites participated in this multicenter, prospective, non-randomized IDE trial. The study received Western Institutional Review Board approval and consent was obtained from all patients prior to participating in the study. Inclusion criteria for this study included patients eligible for primary unilateral TKA due to osteoarthritis defined radiographically by a Kellgren-Lawrence Grade of 3 or higher. Exclusion criteria for this study included previous open knee surgery in the operative knee; body mass index >40 kg/m²; coronal deformity greater than 20° or a sagittal flexion contracture greater than 15°; any type of metallic implant in the operative leg. All patients were enrolled from February 2017 through December 2018. The study, initially designed as a 3-month safety study, expanded to a 6-month safety and efficacy study due to regulatory requirements, thus requiring voluntary re-consent. Six-weeks, 3-month, and 6-month follow-up data was collected for all consented patients, and data collection was discontinued upon the last patient's 6-month visit. Twelve-month follow-up data was collected for patients reaching that milestone prior to database closure. More details of the study methods can be found on www.clinicaltrails.gov. (NCT03017261).

2.1. Investigational device

The TS1 consists of a preoperative planning workstation and an active, independently driven robotic-arm with a high speed cutter, which mills the bone to the precise inner dimensions of the selected tibial and femoral implants (Fig. 1). The surgeon can interrupt/stop the milling using a pendant. For this open platform approach, a CT scan is used to develop the bone model and a library of implants is used to select and position tibial and femoral components to achieve a targeted alignment and 3D position. After surgeon evaluation and approval of the manufacturer's provided plan, a transfer media is formatted with the cut path information of the robotic arm to be transferred to the robot. Intraoperatively, the surgeon exposes the knee, docks the patient's lower extremity to the robot with external fixations pins, and confirms that the robot has a clear path to mill the bone. After milling, the robot is removed from the operative field, trialing and subsequent implantation of components are carried out, and the procedure is completed in the surgeon's routine manner.⁸

Fig. 1.

TSolution one® (TS1) Total Knee Application including TPLAN® and TCAT®.

2.2. Data collection

All TKA-related complications were collected by each site. An independent clinical event committee adjudicated and classified events for a degree of relatedness to the device (the robot) or to the TKA procedure. (i.e. possible, probable, or definitely related). The primary safety endpoint was assessed against a composite endpoint (7.6 %) of incidence rates of seven relevant AEs associated with manual TKA in literature⁹: medial collateral ligament injury (2.7 %),¹⁰ extensor mechanism disruption (2.1 %),¹¹ neural deficit (1.3 %),¹² periprosthetic failure (0.68 %),¹³ patellofemoral dislocation (0.5 %),¹⁴ tibiofemoral dislocation (0.2 %),¹⁵ and vascular injury (0.15 %)¹⁶. This composite value was compared to the percentage of patients that experienced at least one of the relevant AEs at any point during the study period. Blood transfusions and further surgical or non-surgical interventions were recorded and assessed separately as secondary safety endpoints.

The primary effectiveness endpoint was the malalignment rate, defined as the percentage of patients with lower limb alignment greater than ±3° from the planned alignment, determined by comparing the Hip-Knee-Ankle (HKA) angle from the 3-month postoperative standing radiograph to the planned HKA angle from the supine preoperative CT. The postoperative HKA angle measured by two independent radiologists was defined as the angle between the mechanical axes of the femur and tibia (Fig. 2A). The mechanical axes were defined by a line connecting the center of the femoral head to the apex of the intercondylar notch, and a line connecting the centers of the intercondylar notch to the center of the ankle. As the acquisition protocol was not rigidly enforced, several 3-month radiographs were of unacceptable quality. Radiographs were deemed unacceptable if they had sub-optimal image quality, obscured anatomy, or excessive internal-external rotation of the knee, assessed by determining if the anterior flange of the femoral component was not centered between the condylar arches (Fig. 2B). This could occur with improper positioning, residual flexion deformity, or other technical issues. If x-ray images were unacceptable, the sum of the femoral and tibial component varus-valgus errors measured using the pre- and postoperative CT scans was utilized. Most investigators targeted the neutral mechanical axis for limb alignment, while one investigator preferred kinematic alignment. The malalignment rate found in this study was compared to the 32 % malalignment rate for manual TKAs identified from literature.¹⁷

Fig. 2.

The postoperative Hip-Knee-Ankle (HKA) angle measured by two independent radiologists was defined as the angle between the mechanical axis of the femur (red) and mechanical axis of the tibia (blue) (A). Acceptable and unacceptable rotation of the knee (B). Orange arrows indicate the greatest medial and lateral margins of the femoral component. Green arrows indicate the most medial and lateral margins of the anterior flange of the femoral component. Images were deemed acceptable if the anterior flange of the femoral component was centered between the condylar arches. Images were deemed unacceptable if there was internal or external rotation of the femoral component such that the anterior flange was flush with the margin of the condylar arch profile, or the condylar arches were no longer aligned to the x-ray beam and were distinct from the flange.

The secondary effectiveness endpoint evaluated planned versus achieved component placement utilizing pre- and postoperative CT scans. The 3D bone models generated from the CT scans were superimposed in a common reference frame¹⁸ (Fig. 3) and then the difference in placement between planned and actual implants were computed in 3D space. Short-term (preliminary) clinical outcomes including the 2011 Knee Society Scores (KSS) and Short Form 12 (SF-12) Health Surveys, were completed at baseline, 6-week, 3-month, 6-month, and 12-month visits and compared to literature reports for manual TKA. The SF-12 Survey was assessed against the minimal clinically important difference (MCID) score of 2 at 6 weeks following procedure.¹⁹

Fig. 3.

A preoperative CT scan was performed on each patient before surgery and used to create the preoperative plan (A). After surgery, another CT scan was performed to generate 3D models of the femur and tibia, and of the implanted femoral and tibial components (B). Next, the 3D bone models were registered to each other to bring the planned and implanted components in a common coordinate system (C). The position and orientation of the surgically placed implants (purple) were then measured relative to the planned implant positions and orientations (gray) (D).

2.3. Statistical analysis

The sample size determination was based on the primary outcomes of safety and effectiveness. It was assumed that the composite safety event rate for patients undergoing primary TKA with robotic femoral and tibial preparation would be 20 % smaller than the historical control of 7.6 %. The effectiveness hypothesis was that the probability of HKA malalignment for patients undergoing robotic-assisted implantation would be 50 % smaller than the reference rate of 32 %. A one-sided type 1 error of α = 0.05 was assumed. Using industry standard software (nQuery Advisor 7.0, Module POT0x-1), the smallest sample size with power ≥ 80 % was 103. To accommodate an attrition of 10 %, the sample size was set at 115.

The primary safety hypothesis was tested using a one-sided Exact Binomial test with type 1 error rate set to α = 0.05 comparing against the reference proportion (0.076) plus the non-inferiority margin (0.06), i.e., a failure rate less than 0.136. The primary effectiveness hypothesis was tested using a one-sided Exact Binomial test comparing against a 50 % reduction from the reference proportion (0.32), i.e., a failure rate less than 0.16, and then a posterior probability was computed to assess the degree of belief (0 = no belief, 1 = absolute belief) on a reduction in malalignment rate (i.e., ARoS reduced the malalignment rate by at least 40 %).

Analyses were performed using SAS Version 9.4 (SAS Institute Inc., Cary, North Carolina) and R Version 3.6 (R Foundation for Statistical Computing, Vienna, Austria).

3. Results

A total of 115 patients (58 males, 57 females) were enrolled at six US centers. Eight surgeons participated in the study, with varying levels of experience in computer-assisted surgery.

A total of 115, 115, 106, and 59 patients were evaluated at 6-weeks and at 3-, 6-, and 12-months respectively. Nine patients that completed their 3-month visit did not reconsent to the 6-month visit, but two patients reconsented for the 12-month visit. Mean patient age at surgery was 65.9 years (range: 43.0–85.0) and mean body mass index was 30.7 kg/m² (range: 20.6–39.8). Details of the demographic, baseline of continuous variables, and history of comorbidities are listed in the Supplementary Appendix A.

Eight cases (6.95 %) were converted to conventional instrumentation: three prior to any cutting and 5 cases where the investigator finished the procedure manually. In the former group, two cases did not pass diagnostics tests and the machine was accidently turned off by the user prior to tibial cuts in one case. In the remaining cases, conventional instruments were used after the femur had been cut but prior to tibia cutting in three cases, and after the femur had been cut and the tibia was partially cut in two cases. Therefore, a total of 107 patients were included for analysis of malalignment, while all 115 patients were included in the primary safety analysis. No intraoperative complications or adverse events were reported for the 8 cases completed with manual instrumentation.

No patients (0 %) experienced any of the seven predefined adverse events. Therefore, the safety hypothesis was met (p < 0.001) (Table 1). Other adverse events included joint stiffness (3) resolved through manipulation under anesthesia, superficial surgical incision infection (1) resolved by local wound care, and a fixation/registration tack left in the knee (1). Only the latter was definitely related to the device. The patient remained asymptomatic and elected against further intervention. Therefore, no patient underwent further surgical intervention related to the TKA procedure. No blood transfusions were performed perioperatively and there were no prosthetic joint infections.

Table 1.

A table of the results of the study demonstrating the probable benefits of the TSolution One Total Knee Application. The study provides significant data to demonstrate with a high probably (Posterior Probability 0.987) that the TSolution One Total Knee Application had a 40 % reduction in malalignment rate. No Subject experienced any of the pre-defined adverse events that comprised the primary safety endpoint.

Type of Benefit	N of Subjects	N of Events	Magnitude of Benefit	Probability of Patients Experiencing Benefit
Safety: Absence of Composite Safety Events	115	0 (0 %)	100 % of Subjects	p-value <0.001
Effectiveness Malalignment Rate	107	12 (11.2 %)	40 % reduction in the malalignment rate	Posterior probability = 0.987

Thirty-two patients had unacceptable postoperative x-ray images. For these patients, HKA angles were measured using CT scans.¹⁸ One patient had an unacceptable postoperative CT scan that did not contain an adequate field of view, but the x-ray image was acceptable. Of the 107 cases completed robotically, 45.8 %, 65.4 %, and 88.8 % were within 1, 2, and 3° from the target alignment respectively, while 11.2 % (12/107 cases) were outside of ± 3-degrees from the target alignment goal (Table 1).

Of those with greater than a 3° difference, 5.6 % were between 3 and 4°, 3.7 % were between 4 and 5°, and 1.9 % were between 5 and 6°. The malalignment rates for the mechanically (11.4 %) and kinematically (10.8 %) aligned cases were similar. The malalignment rate was 9.4 % (10/106 cases) when calculated using postoperative CT measurements for all cases rather than postoperative x-ray images.

The root mean square errors (RMSEs) in the 3D component placement were within 1.5 mm and 1.5° (Table 2) in all directions. The largest RMSEs occurred in the proximal-distal and flexion-extension of the femoral component (1.5 mm and 1.5° respectively), and in the varus-valgus of the tibial component (1.4°). The planned component sizes matched the implanted femoral and tibial component sizes in 100 % and 77 % of cases, respectively. In 92 % of discrepant cases, the tibial implant was larger than planned.

Table 2.

3D difference between Postoperative and Planned Component Placement for DoF Controlled by the Robotic Device. - The errors are computed in terms of root mean square error (RMSE) and represent the difference between the postoperative implant placement and the planned implant placement.

RMSE	Medial-Lateral (mm)	Anterior-Posterior (mm)	Proximal-Distal (mm)	Flexion-Extension (deg)	Varus-Valgus (deg)	Internal-External (deg)
Femoral Component Error	X	1.1	1.5	1.5	1.2	1.0
Tibial Component Error	X	X	1.0	1.3	1.4	X

X - Note that the errors for the femoral and tibial component in medial-lateral translation and for the tibial component in the anterior-posterior translation and internal-external rotation were removed as they did not depend on the robotic device since the lugs and tibial keel were prepared with conventional instruments per the surgeon's choice.

Significant improvements from baseline were observed in the KSS Functional, KSS Objective, KSS Patient Satisfaction, and SF-12 Physical Composite scores and maintained through the later follow-up visits (Fig. 4, Fig. 5). The MCID of 2 points in the SF-12 Physical Composite score, as identified by Clement et al.,²⁰ was met at 6 weeks and beyond.

Fig. 4.

A comparison of Knee Society Scores (KSS) from the baseline visit through the 12-month follow-up visit, where * indicates a statistically significant difference (p < 0.05) from the baseline value based on a paired t-test. There was 100 % patient follow-up through 3 months, 92 % follow-up at 6 months, and 59 % completed their 12-month follow-up at the time of database closure.

Fig. 5.

A comparison of Short Form 12 (SF-12) scores from the baseline visit through the 12-month follow-up visit, where * indicates a statistically significant difference (p < 0.05) from the baseline value based on a paired t-test. There was 100 % patient follow-up through 3 months, 92 % follow-up at 6 months, and 59 % completed their 12-month follow-up at the time of database closure.

4. Discussion

This study investigated the safety and effectiveness of the ARoS for primary TKA. The main finding of this study is that no patient (0 %) experienced any of the pre-defined complications commonly associated with manual TKA, confirming the safety of the device. Likewise, the second important finding is that the malalignment rate found in this study is 65 % less compared to manual TKA (11.2 % vs 32 %, respectively), which demonstrates a 40 % statistically significant reduction based on the confidence intervals, demonstrating the effectiveness of the device.

Secondary findings for safety were that no patient experienced revisions, reoperations, and bleeding complications, despite an expected incidence ranging from 6 % to 36 % for the latter.^21,22 Only a limited number of complications were possibly related to the use of the ARoS or the TKA. Two out of three ARoS cases with joint stiffness were within ±3° from the planned alignment. One patient had a fixation tack left in the bone, which however has not impacted clinical performance out to 12 months. The risk of reoccurrence of this specific event was mitigated through training at the hospital and improved labeling.

Secondary findings for effectiveness were that all positions and orientations of the postoperative implants were within 1.5 mm or 1.5° from the planned implant placement. Specifically, femoral and tibial component placement errors in medial-lateral translation, tibial component in the anterior-posterior translation, and internal-external rotation did not depend on the system, as the robotic cutting of the finishing steps (i.e. femoral lugs and the tibial keel) was optional for this study.

There is a paucity of studies reporting on implant placement accuracy and malalignment rate with other robotic platforms,23, 24 with most of the research focusing on cadaveric studies. One study by Sires et al.²⁵ compared intraoperatively measured implant placement errors and limb alignment using the MAKO system to postoperative CT scans obtained in 29 patients at a mean time interval of 5.9 months. They found the mean absolute difference between the final implant placement and planned placement to be 1.90 ± 1.88° and 1.97° ± 1.41° for the femoral and tibial component rotations, which are higher but comparable to the current study (1.3 ± 0.9° and 1.0 ± 0.9°, respectively). However, translational errors were not reported by Sires. Additionally, they noted a 7 % malalignment rate with alignment differences ranging from 0° to 6°, which are both comparable to the results of this study. It needs to be noted that the methodologies used to determine the planned alignment with the active robotic system and the postoperative alignment measured in this study with X-rays are not identical. Alignment assessments can be impacted by the patient position during the X-ray image capture.²⁶ In the current study, the alignment rate is slightly higher (90.6 % vs 88.8 % of cases result aligned) when measured with postoperative CT scan alone instead of the X-ray measurement.

While this evaluation represents the first multicenter trial of the TS1 active robotic device for TKA, clinical results using the predicate device (ROBODOC®) have been available since 2000. In 2014, Liow and colleagues reported their findings in 27 patients²⁷ with no mechanical axis outliers. They aborted the robotic procedure in 7.4 % of cases. Subsequent 60 patient (31 robotic and 29 conventional) randomized control studies were conducted28, 29, 30 between May and December 2012. All cases were performed by a single experienced surgeon using the NexGen LPS Flex posterior stabilized implant. Their postoperative assessment included standing long leg radiographs, AP lateral, and skyline view radiographs. Limb alignment and joint line position were measured for all patients. Clinical assessment was performed utilizing standard scoring systems as outcome measures. Results suggested that while clinical outcome measures were similar, radiographic outliers and anterior femoral notching were more prevalent in the conventional group compared to the robotic group (19.4 % vs 0 %; 10.3 % vs 0 % respectively). Their study showed no significant differences in clinical follow-up at 6 months. They have since reported a 2 year follow up study of the same study population,²⁹ and believe their data now suggests a trend toward a higher score in SF-36 Quality of life (QOL) measures with significant differences in SF-36 Vitality (p = 0.03), role emotional (p = 0.02), and a greater proportion of patients achieving SF-36 vitality MCID (48.4 vs 13.8 %, p = 0.009). These evaluation metrics may prove beneficial for other robotic studies. However, retrospective studies such as those by Jeon et al.³¹ and Cho et al.,³² have not shown significant improvement in long-term clinical or radiological outcomes between groups.

Recently, Liow and colleagues shared their initial experiences and technique using the ROBODOC system.³³ They abandoned the robotic procedure in 10 % of their cases and noted the mean robotic surgery duration of 91 min. They noted up to a 3.2 % early revision rate related to technical errors in cases where the robotic procedure was abandoned and finished manually. They remain enthusiastic about the role of robotics, but voice concerns related to the cost and limited flexibility of current system. Our multicenter experience showed an “abort rate” of 6.95 % and no infections throughout the study period.

These prior studies suggest that active robotic technologies can be used safely and will produce alignment, positioning, and sizing approaches that may be more precise. The challenge to demonstrate meaningful outcomes compared to conventional techniques in the long-term remains. The study short-term patient reported outcome scores measured at 3, 6, and 12 months are statistically significant from baseline (Fig. 5).

While our patient-reported outcome measures appear in line with reported results, our study was not designed to find a difference but rather to demonstrate comparability. To demonstrate improvement in outcomes, a prospective and randomized study will be required, such as the recently reported early results from Khlopas et al.³⁴ These studies also point to the need to properly evaluate the learning curve, cost effectiveness, and workload efficiencies that can be achieved with robotic technologies.

Some observational learning curve information for the surgeon and the surgical team was collected in this study. Of the five investigators, one surgeon had extensive navigation and ARoS experience (BNS), two surgeons had prior navigation and haptic robotic experience (SK, RL), and two surgeons had limited to no navigation or robotic experience and this technology was new to all of the OR environments. For this study significant operative time decreases occurred following 10–12 cases (36.51 ± 7.4 min) when compared to the first three (49.1 ± 17 min) cases (p < 0.028). No differences were found in patient reported outcomes, complications, and alignment accuracy between cases completed before and after the learning curve was overcome.³⁵

This study was able to demonstrate the safety and efficacy of the active robotic technology for TKA. The study goals were limited and influenced by regulatory requirements and leave avenues for future research in the post market environment. Some suggestions include: 1) a prospective, randomized study comparing active robotic to manual TKA, in a US population, including longer term outcome measures, 2) cost effectiveness of robotic TKA, 3) evaluation of accurate implant sizing and placement on soft tissue releases.

In summary, standardization of the procedure and the ability to accurately position implants as planned can lead to innovative approaches to education, device design, ligament management, and many other aspects of primary TKA.

5. Conclusions

This multicenter clinical trial demonstrated safety and effectiveness of an active robotic device and defined features of the process that ensure safety. It demonstrated the ability to achieve accurate alignment, a low level of complications, and good clinical outcomes. The results of this initial multicenter trial of the TS1 system resulted in formal approval of the system by the FDA, which can now be used for TKA in the US.

Appendix A. Supplementary data

The following is/are the supplementary data to this article: