This single-centre retrospective cohort study was performed at a tertiary hospital. The protocol was approved by the Ethics Committee of the Affiliated Mindong Hospital, Fujian Medical University (Fujian, China; approval No. H2025010301). All patients provided written consent at the time of surgery for the future use of anonymised clinical and radiographic data. This analysis used anonymised data only and required no additional study-specific consent. The study was conducted in accordance with the Declaration of Helsinki and reported in line with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidance for cohort studies. No animal experiments were involved.

This retrospective study included consecutive patients with isolated LCOA and coexisting AKP who underwent primary knee arthroplasty at our institution between January 2018 and November 2023. Patients were identified from the institutional arthroplasty registry and verified using outpatient records and preoperative imaging. Patients with follow-up shorter than 24 months or incomplete key outcome data were excluded.

Inclusion criteria were: age ≥ 40 years; correctable valgus deformity with valgus alignment ≤ 10°; knee flexion ≥ 100° and flexion contracture ≤ 10°; radiographic LCOA with Kellgren–Lawrence (KL) grade III–IV in the lateral compartment and preserved medial compartment joint space (KL grade 0–II); and mild-to-moderate patellofemoral compartment degeneration (KL grade I–II). AKP was defined as patient-reported pain around or behind the patella during activities such as stair climbing, rising from a chair, or prolonged sitting, supported by clinical assessment.

Exclusion criteria were advanced patellofemoral osteoarthritis (KL grade III–IV), inflammatory arthritis, major coronal plane instability (medial or lateral collateral ligament insufficiency), severe bone loss of the lateral tibial plateau, and loss to follow-up before 24 months. In addition, implant-specific eligibility considerations informed selection of LUKA versus TKA, including ligament integrity, deformity correctability, and the extent of lateral tibial bone loss.

Treatment allocation was non-randomised and reflected routine clinical decision-making documented in the medical records. Procedure choice was based on preoperative clinical and radiographic assessment of compartment involvement, deformity correctability, ligament integrity, and lateral tibial bone loss. LUKA was generally selected when isolated lateral disease with a preserved medial compartment and adequate ligament stability was present and the deformity was clinically correctable. TKA was selected when these implant-specific conditions were not met or when both options were considered feasible but the patient preferred TKA after counselling.

All procedures were performed by a single senior orthopaedic surgeon using a standardised lateral parapatellar approach with a layered, Z-shaped capsulotomy. In our institution, this constituted a unified lateral parapatellar pathway applied to both LUKA and TKA to standardise exposure, patellofemoral assessment, and perioperative management across implant types. Compared with conventional medial parapatellar workflows, the lateral arthrotomy provides direct lateral access for balancing and facilitates intraoperative patellar tracking assessment, thereby reducing variability in patellofemoral handling between groups. Patellar denervation was achieved by circumferential electrocautery along the patellar rim (continuous rim cauterisation along the patellar margin), followed by lateral retinacular release in a graded manner when indicated by persistent lateral tilt or subluxation on intraoperative tracking assessment to facilitate patellofemoral tracking. Patellar tracking was assessed after trial implantation through passive knee flexion–extension and recorded as satisfactory/unsatisfactory in the operative note. The patella was not resurfaced in any case. In LUKA, bone resections were performed using conventional instrumentation, and a cemented fixed-bearing unicompartmental implant was inserted. In TKA, femoral and tibial resections were performed using intramedullary and extramedullary alignment guides, respectively, followed by implantation of a cemented fixed-bearing total knee prosthesis; a cruciate-retaining or posterior-stabilised design was selected according to posterior cruciate ligament integrity.

All patients received perioperative antibiotic prophylaxis, standardised thromboprophylaxis, and postoperative multimodal analgesia according to institutional protocols. Rehabilitation began on postoperative day 1 and followed the same programme in both groups.

Collected variables included demographics (age, sex, body mass index [BMI]), perioperative parameters (operative time, incision length, and intraoperative blood loss), length of hospital stay, complications and reoperations, and clinical outcomes assessed preoperatively and at 3 and 24 months. Intraoperative blood loss was extracted from the operative record as estimated blood loss (EBL); no formula-based derivation was applied. Functional outcomes were evaluated using the Knee Society Score (KSS) (20) and the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (21). The prespecified primary outcome was the KSS at 24 months. Key secondary outcomes were the WOMAC at 24 months and compartment-specific pain ratings (LC-VAS and PFC-VAS) at 24 months. Exploratory outcomes included KSS, WOMAC, LC-VAS, and PFC-VAS at 3 months, perioperative parameters (operative time, incision length, intraoperative blood loss, and length of hospital stay), complications and reoperation-free survival (any reoperation on the index knee as an event).

Pain was measured using a 0–10 VAS (22), where 0 indicated no pain and 10 indicated the worst imaginable pain. At each assessment, patients were asked to rate their average pain during usual daily activities over the preceding week (i.e., activity-related pain rather than pain at rest, typically during weight-bearing ambulation and stair negotiation). All LC-VAS and PFC-VAS ratings were obtained face-to-face during routine outpatient follow-up visits and were recorded by a trained assessor using scripted, standardised instructions. Patients rated pain predominantly perceived in the lateral tibiofemoral compartment (LC-VAS) and around or behind the patella (PFC-VAS) separately, using standardised anatomical diagrams and uniform verbal instructions. During the physical examination, the assessor indicated the corresponding anatomical areas directly on the patient’s knee (lateral joint line/lateral tibiofemoral region and the peripatellar region) and asked the patient to describe and rate pain attributable to each indicated location separately (0–10 VAS). This anatomy-guided localisation approach was adapted from knee pain mapping and digital pain manikin methodologies that have been used to characterise pain location patterns in KOA and after knee arthroplasty (16–18). Because LC-VAS and PFC-VAS have not undergone formal psychometric validation for compartment-specific discrimination, they were prespecified as exploratory endpoints to describe patient-reported pain distribution and were interpreted accordingly. Between-group comparisons for these outcomes were based on the unadjusted analyses described below.

Postoperative radiographs were obtained as part of routine care; however, quantitative radiographic assessments (e.g., limb alignment, component coronal alignment/varus–valgus, tibial slope, and implant positioning) were not systematically measured and were therefore not included in the predefined outcomes. Reoperation-free survival was summarised using Kaplan–Meier curves. Any reoperation on the index knee was treated as an event. Given the low event rate and the fixed follow-up window, survivorship analyses were interpreted descriptively (through 24 months).

All analyses were performed using IBM SPSS Statistics (version 26.0; IBM Corp., Armonk, NY, United States) and R (version 4.3.2; R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were compared using independent-samples t tests (Welch’s t test when appropriate) or the Mann–Whitney U test, and categorical variables using the chi-square or Fisher’s exact test. For the prespecified primary outcome (24-month KSS) and key secondary outcome (24-month WOMAC), exploratory multivariable linear regression was performed with treatment group as the main predictor, adjusting for age, sex, BMI, ASA class, operated side, and the corresponding preoperative score.

To assess robustness to measured confounding, propensity score–weighted sensitivity analyses (stabilised inverse probability of treatment weighting [IPTW] and overlap weighting) were conducted as robustness checks alongside the primary analyses, recognising that these methods address measured covariates only. Propensity scores were estimated using logistic regression with the same prespecified covariates, and covariate balance was evaluated using absolute standardised mean differences (<0.10). Treatment effects were reported as mean differences with 95% confidence intervals from weighted linear regression with robust (sandwich) standard errors. Reoperation-free survival was summarised descriptively using Kaplan–Meier curves given the low event rate. All tests were two-sided, and p < 0.05 was considered statistically significant.

As this was a retrospective cohort with a fixed study period, no a priori sample size calculation was performed. Accordingly, inference emphasised effect estimation with 95% confidence intervals, and no equivalence or non-inferiority framework was prespecified.

After exclusions (two patients with incomplete records, one lost to follow-up, and one who declined research use of their data), 115 patients were included in the final analysis: 52 underwent LUKA and 63 underwent TKA. Baseline characteristics were broadly comparable between groups, including age, sex, body mass index (BMI), ASA class, operated side, and Kellgren–Lawrence grades in the lateral compartment and patellofemoral compartment. Baseline pain and function scores (LC-VAS, PFC-VAS, KSS, and WOMAC) were also similar between groups. Baseline variables are summarised in Table 1.

Patients undergoing LUKA had a significantly shorter operative time than those undergoing TKA (54.38 ± 4.58 vs. 71.22 ± 5.14 min, p < 0.001). In addition, LUKA was associated with lower intraoperative blood loss (58.50 [54.00–70.00] vs. 208.00 [180.00–239.00] mL, p < 0.001), a shorter incision length (8.55 [8.10–9.80] vs. 14.20 [13.45–15.10] cm, p < 0.001), and a shorter hospital stay (5.00 [4.00–5.00] vs. 7.00 [6.00–8.00] days, p < 0.001) (Table 2).

At 3 months, LUKA was associated with better function than TKA, with higher KSS and lower WOMAC, and lower lateral compartment pain (LC-VAS), while patellofemoral pain (PFC-VAS) did not differ significantly (Table 3). By 24 months, KSS, WOMAC, LC-VAS, and PFC-VAS did not differ significantly between groups (Table 3). In exploratory multivariable linear regression adjusting for age, sex, body mass index, ASA class, operated side, and the corresponding preoperative score, the adjusted mean difference in 24-month KSS was 0.51 points (95% confidence interval [CI] −0.58 to 1.60; p = 0.357) and the adjusted mean difference in total WOMAC was −0.62 points (95% CI −1.85 to 0.62; p = 0.326). Propensity score–weighted sensitivity analyses were consistent with the main models. Stabilised IPTW achieved good covariate balance (maximum absolute standardised mean difference = 0.058; maximum stabilised weight = 2.63), yielding a 24-month KSS mean difference of 0.46 (95% CI −0.55 to 1.47) and a WOMAC mean difference of −0.72 (95% CI −1.92 to 0.48). Overlap weighting produced similar estimates (KSS 0.36, 95% CI −0.66 to 1.37; WOMAC −0.63, 95% CI −1.84 to 0.57). At 24 months, LC-VAS and PFC-VAS values were close to the lower bound and KSS values were high, suggesting potential floor and ceiling effects that may reduce the ability to detect small residual between-group differences.

No revisions occurred in the LUKA group. In the TKA group, one patient (1.6%) underwent debridement, antibiotics, and implant retention (DAIR) for periprosthetic joint infection (PJI). When any reoperation on the index knee was treated as an event, Kaplan–Meier curves were used to summarise reoperation-free survival through 24 months. With only one event recorded, the curves were interpreted descriptively (Figure 1).

To provide radiographic context, one representative case from each group is shown. In the LUKA case, the KSS improved from 60 preoperatively to 92 at 24 months, and the WOMAC decreased from 65 to 16 (Figure 2). A patient treated with TKA showed a comparable change in KSS and WOMAC over the same follow-up period (Figure 3). These images illustrate typical preoperative findings and implant positioning at follow-up.