A retrospective analysis of the clinical records of 2 cohorts for a total of 95 lateral UKA implants was performed. All patients underwent surgery between 2011 and 2017. The first cohort consisted of 43 patients with cemented lateral UKA who underwent the standard nonrobotic procedure (conventional group). In comparison, the second cohort consisted of 52 patients who received robot-assisted cemented lateral UKA (Robotic group).

To eliminate selection bias, the inclusion and exclusion criteria for the study were restricted as follows for a more detailed comparison and to obtain similar preoperative values (Table 1): patients undergoing lateral UKAs procedure; minimum 60-month follow-up; age between 40 and 80 years; nonprevious surgery of the affected knee (except meniscectomy or anterior cruciate ligament [ACL] reconstruction); preoperative absence of systemic disease (e.g. diabetes, rheumatoid arthritis); preoperative BMI < 40.

Table 1 Demographics of the patient population

All the procedures were performed in two different high-volume centres specialised in knee surgery:

1.

IRCCS Istituto Ortopedico Galeazzi, Milan, Italy. All UKAs in the Conventional Group were performed by a single surgeon (43 patients).

2.

Policlinico di Abano Terme, Abano Terme, Italy. All robotic group procedures were performed by four different surgeons (52 patients).

The indications for surgery were as follows: Kellgren–Lawrence of the lateral compartment grade III or IV OA; avascular necrosis or osteonecrosis isolated of the lateral femoral condyle; idiopathic or secondary osteoarthritis of the lateral femoral compartment of the knee; knee flexion > 100°; flexion contracture < 15°; valgus deformity (measured on hip–knee–ankle angle) < 10°; integrity of cruciate and collateral ligaments; osteoarthritis of the medial compartment and patellofemoral grade I or II according to Kellgren–Lawrence classification [5].

Magnetic resonance imaging (MRI) was assessed preoperatively to confirm the anatomical integrity of the cruciate and collateral ligaments in all patients and the absence of OA of the medial and patellofemoral knee compartment.

Patient recruitment

A total of 117 patients were initially screened. Of them, 22 were deemed ineligible for the following reasons: severe OA of the patella requiring resurfacing (N = 9), previous high tibial osteotomy (N = 8) and previous femoral/tibial fracture (N = 5). A total of 95 patients were included in the present study: 52 were allocated to the Conventional group and 43 to the Robotic group (Fig. 1) [6]. No patients were lost to follow-up.

Fig. 1Surgical technique: conventional group

All patients were placed in a supine position on a standard operating table after spinal anaesthesia had been induced, as the knee and ipsilateral hip should be freely mobile without the use of a tourniquet. Good exposure was obtained with a skin incision, starting from the lateral margin of the patella to a point approximately 4–5 cm distal to the joint line. Deep exposure was achieved with a lateral parapatellar approach through the subcutaneous tissues to the joint capsule, and the patella was medially dislocated. Inspection of the patellofemoral and medial compartments was routinely performed. The ACL status was checked. Lateral and intercondylar osteophytes were removed, and an anterior tibial precut was performed to gain adequate posterior and articular view and access. For balancing, a spacer block was inserted into the joint space until the anterior stop contacted the anterior tibia to assess the gap. The spacer block was placed to sit flat on the resected tibial surface to ensure that the proper amount of distal femoral bone was resected.

In all cases, the fixed-bearing ZUK Unicompartmental Knee (LimaCorporate, Villanova di San Daniele del Friuli, Udine, Italy) was implanted using the corresponding instrumentation, extramedullary tibial guide and femoral and tibial cutting guides. All components were cemented using Refobacin® Bone Cement R (Zimmer Biomet, Warsaw, Indiana, USA). Tibial coverage was maximised without any overhang while targeting the natural tibial slope [7].

Surgical technique: robotic group

All the prostheses were fixed-bearing metal-backed UKAs (Restoris MCK, Stryker, Mahwah, New Jersey) implanted with a semiautonomous robot, the MAKO robotic arm system (Stryker). A lateral parapatellar approach was performed [8].

Before the surgery, a 3D model of the patient’s knee was created from a preoperative CT scan. Then, surgical planning was performed, defining the size and position of the tibial and femoral components, the amount of bone resection and the implant alignment. The planning was changed after bone registration and pose capture, applying varus stress to correct the valgus deformity and tension in the lateral collateral ligament. Once gap balancing and implant tracking were checked, bone resection was performed using a 6 mm burr. The final implants were cemented using Biomet® Bone Cement R (Zimmer Biomet, Warsaw, Indiana, USA).

Rehabilitation protocol

Both patient groups followed the same rehabilitation protocol, which involved passive mobilisation on the same day of surgery; from day one, patients were started on active progressive joint mobilisation and assisted walking with two crutches. Gradually and according to each patient, recommendations were made to increase the load during walking and continue with isometric muscle toning exercises until patients could walk independently without the use of walking aids [9].

Survivorship

A revision was defined as the failure of the implant (periprosthetic joint infection [PJI], periprosthetic fracture, aseptic loosening or revision with TKA for medial or patellar cartilage deterioration), and survival was based on implant revision. PJI was diagnosed according to the New Definition for Periprosthetic Joint Infection: From the Workgroup of the Musculoskeletal Infection Society [10]. Periprosthetic fractures were defined as fractures of the femur or tibia occurring within 15 cm of the joint line or 5 cm of the endomedullary stem if present [11]. Patients were classified as having aseptic loosening if they had symptoms including pain, instability or swelling and radiographic evidence of loosening and did not meet the Musculoskeletal Infection Society criteria for infection [12].

Clinical evaluation

All clinical assessments were performed by two independent clinicians who were not involved in the index surgery. The clinical evaluation entailed evaluating the Knee Injury and Osteoarthritis Outcome Score (KOOS), which is divided into subscales (symptoms and stiffness, pain, function in daily living, function in sport and recreation and quality of life) for each patient. Each patient was clinically evaluated on the day before surgery (T0), at a minimum 1-year follow-up (T1) and at a minimum 5-year follow-up (T2) [13].

All the procedures involving human participants in this study followed the institutional or national research committee ethical standards, the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Informed consent was obtained from all the participants included in the study [14]. Appropriate ethical approval was also obtained from the local ethics committee (Ethical Committee of San Raffaele Hospital—CE 236/2017).

Statistical analysis

Summary statistics are presented as the mean and standard deviation (SD) values or absolute frequencies and percentages. After testing the distribution of continuous variables, a t test or a Wilcoxon–Mann–Whitney test was performed to assess preoperative differences between the Conventional and Robotic groups, and a chi-square test was used to evaluate the categorical variables.

The groups’ KOOS total and subscores were compared at each time point with a t test or a Wilcoxon–Mann–Whitney test as appropriate. Then, for each group, pre- and postoperative scores were compared with a paired t test or a Wilcoxon signed-rank test.

Failures were recorded for each group, and Fisher’s exact test was performed to assess any differences in the total number of failures between the conventional and robotic groups. In addition, survival curves were estimated for each group to account for the time of failure onset from surgery. A Cox regression model was created using ‘failure’ as an independent variable and the specific group as a covariate. Age, BMI, sex and surgical side were also added as covariates to the Cox model, but no statistical significance was found. All tests were two-sided, and a p value of less than 0.05 was considered statistically significant. Statistical analyses were conducted in R (version 4.1.1).

Sample size

An estimated sample of 80 subjects, 40 for each group, was required to compare KOOS scores between groups with a two-sided Wilcoxon–Mann–Whitney test, assuming a mean difference of 20, an SD of 20 for both groups, a 5% alpha and 99% power. Given the same parameters, this sample also had 99% power to detect a prepost difference using a Wilcoxon signed-rank test [15].