Robotic arm-assisted versus conventional unicompartmental knee arthroplasty

Trial design

The original trial was designed as a prospective parallel equally randomised single-centre study to compare alignment in two groups of patients undergoing unicondylar knee arthroplasty for treatment of osteoarthritis of the medial compartment of the knee. The Mako robotic arm-assisted system was used for one group and a manual UKA was performed using traditional surgical jigs in the other.

The present study reports a secondary exploratory analysis of whether the increased accuracy provided by robotic-arm technology influences early clinical outcomes.

Patients

A total of 139 patients were recruited from a single centre (Glasgow Royal Infirmary, Glasgow, United Kingdom) between October 2010 and December 2012 (Fig. 1) who were awaiting unicompartmental knee arthroplasty for medial compartment primary osteoarthritis. Enrolment was carried out by a research associate.

Fig. 1

Consolidated Standards Of Reporting Trials diagram showing the flow of participants through each stage of the randomised trial (UKA, unicompartmental knee arthroplasty; TKA, total knee arthroplasty).

Patients included in the study were those considered suitable for UKA by the surgical authors (MJGB, BJ, AM), who gave written informed consent and who could comply with the study follow-up regime.

Patients were excluded if they had any of the following: any contraindications detailed by the device manufacturer; any tibial deformity requiring tibial component augmentation; requirement for a total knee prosthesis; inflammatory polyarthritis; a disorder of the contralateral knee, feet, ankles, hips or spine causing significant abnormal gait or significant pain; a neurological condition affecting movement; or any other pre-existing condition that would, in the opinion of the investigator (MJGB), compromise their participation and follow-up in the study.

Randomisation and blinding

The randomisation was performed by the Robertson Centre for Biostatistics (University of Glasgow) via a bespoke web-based randomisation portal and S-Plus (TIBCO Software Inc, Palo Alto, California), with stratification by surgeon (MJGB, BJ, AM). The treatment team were blinded to the sequence, and patients and researchers were not informed of the outcome of the randomisation.

Treatment

Patients were randomised to receive either an Oxford Phase 3 UKA (Biomet, Warsaw, Indiana) implanted with the aid of traditional surgical Phase 3 jigs, or a Restoris MCK UKA (Mako Surgical Corporation) implanted with the aid of the Mako System, a tactile guided robotic arm-assisted system. Three specialist knee surgeons (MJGB, BJ, AM) with a minimum of five years independent practice as an orthopaedic surgeon performed the surgery. The clinical unit performs approximately 100 UKAs per annum.

Surgical technique: robotic arm-assisted UKA

A pre-operative CT scan was segmented by a trained technician to construct a 3D model of the patient's knee. This allows planning of individualised component positioning prior to surgery. The operating surgeon (MJGB, BJ, AM) defined the size and position of the femoral and tibial components in the pre-operative plan, optimising bone coverage, restoring joint anatomy and minimising bone resection. Implant alignment, therefore, was tailored to each patient. Using the pre-operative plan, the Mako system calculates the 3D volume of bone requiring resection, allowing the robotic arm to resect bone using a high speed, saline-cooled burr. Any burring outside of the predetermined zone is resisted by the robotic arm using tactile feedback and audio signals, with complete shutdown of the burr if the arm is forced outside of the zone.

The system uses optical motion capture technology (Polaris Spectra, NDI, Northern Digital Inc. Ontario, Canada) to track photoreflective marker arrays fixed to the femur and tibia through separate stab incisions. This technique allows dynamic referencing of the femur and tibia. Thus, the 3D bone resection volume moves with the limb in real time as the surgeon moves the limb. Visual feedback is given to the surgeon by the on-screen computer aided design images, and tactile feedback is provided by the robotic arm restricting the burr to stay within the resection volume.

The Restoris MCK implant consists of a cobalt chrome femoral component and a titanium tibial component with a fixed bearing polyethylene insert.

Surgical technique: conventional UKA

Conventional UKA operations were carried out using standard manual instrumentation and the Oxford Phase 3 UKA. Standard instrumentation involved pinning a tibial cutting guide to the tibia, providing a flat surface to guide manual resection of the bone using a handheld reciprocating saw for the vertical cut and an oscillating saw for the horizontal. This guide is aligned using visual and palpable anatomic landmarks. On the femoral side, an intramedullary rod is inserted into the distal femur to align the femoral cutting guide, again using visual landmarks. The standard instrumentation jigs and accompanying operating technique manual provide fixed target values for all patients, without the opportunity for tailoring of implant position to each patient’s anatomy.

The Oxford UKA consists of a cobalt chrome femoral and tibial implant and a fully congruent polyethylene mobile bearing.

Follow-up

Data were collected at three months and one year post-operatively. All trial data were collected by a blinded independent research nurse or research associate at the Glasgow Royal Infirmary, United Kingdom.

Power calculation

The primary outcome required 126 patients to detect a difference of 1° in tibial sagittal positioning with a power of 80% (a = 0.05). To allow for loss to follow-up, the total target of recruitment was 150 patients (75 in each group). Completion of recruitment was regarded as the primary stopping point for the surgical stage, and completion of follow-up at one year post-operatively for all patients was defined as the stopping point for this study of secondary outcomes.

Outcome measures

We report within this paper the secondary clinical outcomes from the randomised controlled trial. Outcome measures included the American Knee Society Score (AKSS), Oxford Knee Score (OKS), Forgotten Joint Score (FJS), Hospital Anxiety Depression (HAD) Scale, University of California, Los Angeles (UCLA) Activity Scale, Short Form-12 (SF-12), Pain Catastrophising Scale (PCS), somatic disease (Primary Care Evaluation Of Mental Disorders Score), pain visual analogue scale (VAS), analgesic use, patient satisfaction, complications relating to surgery, the requirement for revision surgery and a 90-day diary to catalogue early pain and functional recovery post-operatively.

Statistical analysis

Student’s t-test was used to compare continuous variables with a normal distribution of data. The Mann-Whitney U test was used to compare continuous variables without normal distribution. A chi-squared test or Fisher’s exact test was used to compare categorical data. These analyses were performed using GraphPad Prism 5.04 (GraphPad Software Inc., La Jolla, California).

Due to the exploratory nature of the study, a standard alpha level of 0.05 was adjusted for multiplicity to 0.005, given the ten secondary outcomes presented, following a Bonferroni correction. Given the exploratory approach of the study and the conservative nature of the Bonferroni approach, both levels were used to highlight key results.

The study was analysed on a per-protocol basis due to a lack of available data from which to perform an intention-to-treat analysis for those who had withdrawn or had been discharged, or those who were converted to TKA on the table or those who were revised.

Proposed additional secondary outcome analysis

A post hoc power calculation was carried out on the population means, sample deviations and sample numbers to determine the power ((1-β) × 100%) of the study at three months and one year to detect the minimally important clinical difference¹⁸ for the AKSS and OKS (supplementary material). This showed that the study was powered for OKS but otherwise underpowered for AKSS.

Factors predictive of excellent and poor clinical outcomes analysis

The ceiling limitations of the current standard patient-reported outcome measures are well recognised^19,20 and render them incapable of differentiating between degrees of excellence in clinical outcomes. For further analysis, outcome scores were dichotomised due to the non-linear nature of many of the scores used.

For the AKSS and OKS, the 90th centile was used to differentiate patients with an excellent outcome who may be limited by the ceiling limits of the respective scores; for the AKSS this value was 180 out of 200, and for the OKS it was 43 out of 48. For the FJS, a less stringent 80th percentile value was used as the ceiling effect of this score is minimal.²¹

The HAD score can be divided into an anxiety score and a depression score; a score of greater than eight is generally accepted as being indicative of depressive or anxious traits.²²

For the UCLA Activity Scale, a cutoff value of five was used. This represents patients who are able to undertake all of the basic daily activities of life such as housework, shopping and simple exercise.

The PCS cutoff level was set at 20 out of 54, which represents the cutoff point for the bottom tertile of the patient cohort, as previously been used by Riddle et al.²³

Patients who report more than three somatic-type symptoms on the Somatic Disease Scale, for which a cause cannot be identified, are considered to be at risk of somatic disease. Without access to general practitioner records, the ability to assess and verify all patient symptoms other than joint pain was restricted, and therefore a more stringent level of five reported symptoms was used. This level also allows for the fact that all patients in the cohort, by default of indication, report joint pain as a symptom.

Finally, for the pain VAS, a cutoff value of 70 out of 100 was used to denote severe pain. This value has previously been used by Kelly²⁴ to denote patients with severe pain.

Binary logistic regression modelling was used to predict factors that are important in the surgical outcomes, and this was performed using Minitab vs12 (Minitab Inc., State College, Pennsylvania).

Study oversight

The study complied with the principles of the Declaration of Helsinki and was approved by the local ethics committee of the West of Scotland Research Ethics Service (10/S0704/12) and registered with the ISRCTN (International Standard Randomised Controlled Trial Number) Registry (ISRCTN77119437).

Follow-up

A total of 139 patients underwent randomisation (Fig. 1); 64 of 70 (91%) in the robotic arm-assisted group and 62 of 69 (90%) in the manual surgery group completed the 12-month follow-up. The mean three month and one year follow-up times for the robotic arm-assisted group was 3.2 months and 13.0 months, and for the manual surgery group it was 3.2 months and 13.0 months.

Pre-operative demographics

Pre-operative demographics (supplementary material) were well balanced for all variables other than pre-operative anxiety. Although the difference in pre-operative anxiety between the groups reached statistical significance (p < 0.05), the higher anxiety in the manual group is unlikely to be a clinically relevant difference (0.5).

Early post-operative pain

Pre-operative pain levels were not significantly different between the two groups (supplementary material). However, from the first post-operative day through to week 8 post-operatively, the median pain scores for the robotic arm-assisted group were 55.4% lower than those observed in the manual surgery group (Fig. 2) (p = 0.040). However, by three months and one year post-operatively, there was no difference in pain scores between the groups (supplementary material). All patients were offered the same analgesic pathways (supplementary material), demonstrating no difference in the overall analgesic use for either patient group.

Fig. 2

Graph showing early post-operative pain. Visual analogue scale (VAS) pain scores were recorded by the patient daily for the first seven days, and then subsequently weekly.

Clinical outcomes

At three months post-operatively, the robotic arm-assisted group had better AKSS (Fig. 3, and supplementary material), although no difference was noted in OKS. The FJS is a measure of a patient’s awareness of their joint. Although there was no overall statistical difference between the two groups, the proportion of patients achieving a forgotten joint (FJS > 80%) was almost double in the robotic arm-assisted group (15% versus 8%, p = 0.265).

Fig. 3

Graph showing three-month post-operative American Knee Society Score (AKSS) (p = 0.0405).

At one year post-operatively, the reported differences with the AKSS had narrowed from a mean of 21 points to a mean of 7 points (p = 0.106) (Fig. 4, and supplementary material), with 44% of the robotic arm-assisted group and 26% of the manual surgery group reaching the ceiling limit of the score (AKSS > 180/200). The proportion of patients with a forgotten joint had increased proportionately in both groups to 26% in the robotic arm-assisted group and 13% in the manual surgery group (p = 0.067). No difference was observed with the OKS, and almost half of each group reached the ceiling limit of the score (OKS > 43) (supplementary material).

Fig. 4

Graph showing one-year post-operative American Knee Society Score (AKSS) (p = 0.1056).

A greater proportion of patients receiving robotic arm-assisted surgery improved their UCLA Activity Score from pre-operatively to one year post-operatively by more than one level, 69% versus 52% (p = 0.06).

There were no significant differences between the two groups using the general health outcome measure SF-12, nor were there any significant differences in complications noted. No revision surgery was performed on any patient within the first 12 months after surgery. There were a number of minor wound complications which were more common in the manual surgery group (supplementary material), but there were no deep infections in either group.

Inpatient length of stay was shorter in the robotic-arm surgery group, with a difference of 0.54 days (p = 0.07) (supplementary material). Additionally, three months post-operatively, primary care utilisation calculated from the group proportions visiting their GPs, was 15% lower (p = 0.092) (supplementary material) in the robotic-arm group.

Factors predictive of excellent and poor clinical outcome

Using AKSS > 180, OKS > 43 and FJS > 80% as markers of excellent clinical outcome, the key factors associated with achieving excellent outcome were a pre-operative activity level > 5 on the UCLA activity score (all three outcome measures; AKSS, OKS, FJS), use of robotic-arm surgery (two outcome measures; AKSS, FJS), and not having pre-operative depression (one outcome measure; AKSS) (supplementary material).

Subanalysis of patients with pre-operative UCLA activity scores > 5 revealed differences in the outcome between robotic arm-assisted and manual surgery for the AKSS (p = 0.0064), the OKS (p = 0.0106) and the FJS (p = 0.0346) (Figs 5 and 6, and supplementary material).

Fig. 5

Graph showing one-year post-operative American Knee Society Score (AKSS) for patients with pre-operative University of California, Los Angeles Activity Scale score > 5. The lack of depressed patients in this higher functioning group further represents association of depression and poor clinical outcomes. Two depressed patients in the robotic arm group achieved higher levels of activity (p = 0.0064).

Fig. 6

Graph showing one-year post-operative Oxford Knee Score for patients with pre-operative University of California, Los Angeles Activity Scale score > 5. The lack of depressed patients in this higher functioning group further represents association of depression and poor clinical outcomes. Three depressed patients in the robotic arm group achieved higher levels of activity (p = 0.0106).

Factors associated with poor outcome were pre-operative depression (three outcome measures; AKSS, OKS, FJS) and pre-operative anxiety (one outcome measure; OKS) (supplementary material).