Revision of partial to total knee arthroplasty using robotic assistance

Abstract

Aims

To evaluate the feasibility, radiological accuracy, and early clinical outcomes of robotic-assisted conversion of partial knee arthroplasty (PKA) to total knee arthroplasty (TKA) using the ROSA system. Feasibility was defined as successful robotic registration and completion of the procedure using standard primary components without femoral components or augments.

Methods

A retrospective cohort of 23 consecutive conversions (medial or lateral unicompartmental and patellofemoral arthroplasties) was analyzed. The minimum follow-up was 12 months. Clinical outcomes were assessed with the Oxford Knee Score (OKS), visual analogue scale (VAS) for pain, and patient satisfaction. Radiological accuracy was evaluated by comparing planned with achieved component alignment (medial proximal tibial angle (MPTA), lateral distal femoral angle (LDFA), and hip-knee-ankle (HKA)) using paired t-tests, two one-sided tests for equivalence (SD 1.5°), and Bland-Altman analysis.

Results

All cases were successfully registered with the robot and completed without femoral components or augments; 22 of 23 used standard primary components. The mean OKS improved from 17.2 (SD 5.5) to 40.0 (SD 5.0) and VAS pain decreased from 6.7 (SD 1.5) to 1.6 (SD 1.6) (p < 0.001). The mean absolute error between planned and achieved alignment was 0.8° for MPTA and 0.3° for LDFA, confirming accurate restoration, while HKA showed greater variability (mean bias 1.37°; limits -4.1° to 6.9°). One secondary patellar resurfacing was required; no other complications occurred.

Conclusion

Robotic-assisted conversion of PKA to TKA was feasible, bone-preserving, and resulted in significant clinical improvement at 12 months, with precise component alignment but slightly more variation in overall mechanical axis restoration.

Cite this article: Bone Jt Open 2026;7(3):283–290.

Introduction

Knee osteoarthritis is a leading cause of pain and disability worldwide. Surgical interventions such as partial knee arthroplasty (PKA), including medial unicompartmental knee arthroplasty (UKA), lateral UKA, and patellofemoral arthroplasty, and total knee arthroplasty (TKA) are well-established treatments to alleviate symptoms and restore function.¹ PKA is frequently preferred in patients with isolated compartment degeneration due to preservation of native joint anatomy, reduced blood loss, faster recovery, and improved restoration of knee kinematics compared with TKA.^2-5

Despite these advantages, PKA has a higher revision rate than TKA. According to the latest national joint registries, ten-year survival rates are approximately 90% for PKA and 94% to 95% for TKA, consistent across the 2023 Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR) and 2024 National Joint Registry (NJR) reports.^6-9 One key reason is the lower threshold for revision: a PKA with poor functional scores is more likely to be revised than a TKA with similar symptoms. For instance, a PKA with a low Oxford Knee Score (OKS) has a 60% likelihood of being revised, compared with only 10% for TKA.^10,11 This may partly reflect differences in surgeon perception, patient expectations, and the technical feasibility of revising a PKA.

Failed PKA is typically converted to a TKA, which remains an effective salvage strategy. However, revision TKA presents significant technical challenges due to potential bone loss, altered joint lines, and soft-tissue compromise, requiring advanced reconstructive techniques such as modular augments, bone grafting, or cement with screws.^12,13

Robotic-assisted knee arthroplasty has emerged as a promising technology to improve implant positioning and alignment accuracy, reducing outliers in both UKA and TKA.^14-16 While evidence supports its use in primary procedures, little is known about its application in revision settings, particularly for the conversion from PKA to TKA.

The aim of this study was to evaluate the feasibility, defined as successful robotic registration and completion with standard primary components, together with the radiological accuracy and 12-month clinical outcomes of robotic-assisted conversion of partial to total knee arthroplasty.

Methods

Study design and patient selection

This retrospective cohort study included prospectively collected cases of robotic-assisted conversion of unicompartmental (medial or lateral) or patellofemoral knee arthroplasty (PKA) to total knee arthroplasty (TKA) performed at a single tertiary orthopaedic centre between May 2022 and September 2023. Only aseptic revisions were eligible; patients revised for periprosthetic joint infection or post-traumatic indications were excluded. A total of 23 consecutive patients met the inclusion criteria and were analyzed. The minimum follow-up available was 12 months.

Patient characteristics

A total of 23 patients were included in the study, with a mean age of 65.6 years (SD 10.5). The cohort consisted of 16 females and seven males. The index procedure was a medial unicompartmental knee arthroplasty (UKA in 17 patients (12 fixed-bearing and five mobile-bearing)), a patellofemoral joint (PFJ) arthroplasty in four patients, and a lateral UKA in two patients (both fixed-bearing). The mean time from index procedure to revision was 8.9 years (SD 5.98). Patient demographic details and baseline characteristics are summarized in Table I.

Table I.

Patient demographic details and index procedure characteristics.

Variable	Data
Mean age, yrs (SD)	65.6 (10.5)
Sex, n (%)
Female	16 (69.6)
Male	7 (30.4)
Index procedure type, n (%)
Medial UKA	17 (74.0)
Patellofemoral arthroplasty	4 (17.4)
Lateral UKA	2 (8.7)
Bearing type (UKA only), n (%)
Fixed bearing	14 (82.4)
Mobile bearing	3 (17.6)
Mean time from index to revision, yrs (SD)	8.9 (6.0)

UKA, unicompartmental knee arthroplasty.

Surgical technique

All procedures were performed by a single high-volume knee surgeon (FH) experienced in robotic-assisted arthroplasty. The ROSA Knee System (Zimmer Biomet, USA) was used in all cases with fully cemented implants. All cases were performed using the Persona Total Knee System (Zimmer Biomet), except for one patient who required a NexGen legacy constrained condylar knee (LCCK) tibial component due to extensive cystic bone loss. Planning and execution followed the principles of inverse kinematic alignment (iKA) to restore the joint line and native knee kinematics.¹⁷

A specific feature of our workflow was preresection robotic landmarking with the unicompartmental prosthesis in situ. Anatomical landmarks were registered directly on the femoral and tibial components prior to explantation, allowing intraoperative identification of joint line alteration, component malalignment, and soft-tissue imbalance, and guiding bone resections to maximize preservation of host bone and ligament structures.

Full-length standing radiographs were obtained preoperatively and at final follow-up for all patients. For illustrative purposes, a representative case using primary components is shown in Figure 1, and the only case requiring a LCCK tibial component is shown in Figure 2. These images also demonstrate coronal alignment restoration following robotic-assisted conversion.

Fig. 1.

Robotic-assisted conversion from medial unicompartmental knee arthroplasty (UKA) to total knee arthroplasty (TKA) using primary components in an 83-year-old female patient, left knee. A) Preoperative anteroposterior radiograph showing medial UKA. B) Postoperative anteroposterior radiograph at six-month follow-up demonstrating restored coronal alignment and primary TKA components.

Fig. 2.

Robotic-assisted conversion from medial unicompartmental knee arthroplasty (UKA) to total knee arthroplasty (TKA) requiring a revision-type tibial component in a 78-year-old female patient, left knee. A) Preoperative anteroposterior radiograph showing failed UKA with substantial cystic tibial bone loss. B) Postoperative anteroposterior radiograph at six-month follow-up demonstrating NexGen legacy constrained condylar knee tibial component and primary femoral component.

Clinical outcome measures

Clinical outcomes were assessed using the OKS,¹⁸¹⁹ visual analogue scale (VAS) for pain,²⁰ and a patient satisfaction score on a ten-point Likert scale. Data were collected preoperatively and at three, six, and 12 months postoperatively. The Patient Acceptable Symptom State (PASS) was defined as an OKS ≥ 37, and the minimum clinically important difference (MCID) as an improvement of ≥ 5 points compared with baseline.²¹

Radiological analysis

Radiological measurements included the medial proximal tibial angle (MPTA), lateral distal femoral angle (LDFA), and hip-knee-ankle (HKA) angle on standardized long-leg standing radiographs obtained preoperatively and at final follow-up.

HKA was measured under full weightbearing and compared with intraoperative robotic values obtained in the supine position, acknowledging that load-bearing and rotation can influence coronal alignment measurements.

These radiological values were compared with the planned alignment as defined intraoperatively by the ROSA system to assess implant positioning accuracy. Measurements were performed by two independent reviewers using digital templating software. An example is shown in Figure 1, with angular values indicated.²²

Statistical analysis

Descriptive statistics were used to summarize patient demographic details and baseline characteristics. Longitudinal changes in clinical outcome scores were analyzed using repeated measures analysis of variance (ANOVA). Radiological accuracy was assessed using the two one-sided test (TOST) procedure for equivalence, in addition to paired t-tests and Bland-Altman analysis comparing planned versus achieved alignment values. Statistical analyses were performed in SPSS Statistics v. 29 (IBM, USA). A p-value < 0.05 was considered statistically significant.

Ethical approval was obtained from the institutional review board, and written informed consent was obtained from all participants prior to inclusion.

Results

Indications and surgical details

The primary indications for revision were polyethylene wear (n = 7), progression of osteoarthritis (n = 6), aseptic loosening (n = 4), component malposition (n = 4), and instability (n = 2). These are presented in Table II.

Table II.

Indications for revision and implant type.

Index procedure	n	Main indication for revision	Tibial component	Femoral constraint design	Patellar management
Medial UKA	17	Polyethylene wear (6); OA progression (5); Aseptic loosening (3); Malposition (2); Instability (1)	16 Primary Persona; 1 LCCK	15 PS; 2 CR	No resurfacing: 16; Secondary: 1
Lateral UKA	2	OA progression (1); Loosening (1)	2 Primary Persona	2 PS	No resurfacing: 2
Patellofemoral arthroplasty	4	Component malposition (2); OA progression (2)	4 Primary Persona	2 CR; 2 PS	No resurfacing: 4
Total	23	—	22 Primary Persona; 1 LCCK	19 PS; 4 CR	22 No; 1 Secondary

CR, cruciate-retaining; LCCK, legacy constrained condylar knee; OA, osteoarthritis; PS, posterior-stabilized; UKA, unicompartmental knee arthroplasty.

A standard primary TKA system was used in all but one case, in which a NexGen LCCK tibial component was implanted due to substantial cystic bone loss; a primary femoral component was retained in this patient (Figure 1 and Figure 2). All PFJ revisions were performed using a cruciate-retaining (CR) design, while the remaining cases received a posterior-stabilized (PS) implant. A lateral retinacular release was performed in four patients; no additional ligament releases were required. The mean polyethylene insert thickness was 11.0 mm (SD 1.02; 10 to 13). This value is reported descriptively to illustrate restoration of joint-line height and was not treated as an outcome variable

Clinical outcomes

The Oxford Knee Score (OKS) improved significantly from a mean of 17.2 (SD 5.5; 7 to 28) preoperatively to 40.0 (SD 5.0; 26 to 47) at 12 months postoperatively (p < 0.001). Pain scores on the visual analogue scale (VAS) decreased from a mean of 6.7 (SD 1.5) to 1.6 (SD 1.6) over the same period (p < 0.001). Post-hoc Bonferroni-adjusted pairwise comparisons confirmed statistically significant improvement at all postoperative time points (p < 0.01 for both scores). At 12 months, 19 of 23 patients (83%) achieved the patient acceptable symptom state (PASS; OKS ≥ 37), and all patients (100%) exceeded the minimum clinically important difference (MCID; ΔOKS ≥ 5). These findings indicate that nearly all patients experienced a clinically meaningful improvement, and the vast majority reached a satisfactory symptom state. Clinical outcomes are detailed in Table III.

Table III.

Clinical outcomes.

Outcome measure	Preoperative, mean (SD)	Postoperative (12 mnths), mean (SD)	p-value
Oxford Knee Score	17.2 (5.5)	40.0 (5.0)	< 0.001
Visual analogue scale	6.7 (1.5)	1.6 (1.6)	< 0.001

OKS, Oxford Knee Score; VAS, visual analogue scale.

Radiological accuracy

Table IV summarizes the planned compared with achieved alignment values for the medial proximal tibial angle (MPTA), lateral distal femoral angle (LDFA), and hip-knee-ankle (HKA) angle. Paired t-tests revealed small but statistically significant differences for MPTA (mean difference 0.80°, p = 0.007) and HKA (–1.37°, p = 0.028), while LDFA showed no significant deviation from planned values (mean difference 0.30°, p = 0.466).

Table IV.

Radiological accuracy.

Angle	Mean planned, ° (SD)	Mean achieved, ° (SD)	Mean difference, °*	p-value	Equivalence ( ± 1.5°)
Medial proximal tibial angle	87.95 (0.93)	87.19 (1.44)	0.76	0.007	Yes
Lateral distal femoral angle	88.25 (1.04)	87.95 (2.23)	0.30	0.466	Yes
Hip-knee-ankle angle	-0.03 (0.79)	1.35 (3.01)	-1.38	0.028	No

^*Mean difference represents achieved minus planned.

HKA, hip-knee-ankle angle; LDFA, lateral distal femoral angle; MPTA, medial proximal tibial angle.

Equivalence testing using the two one-sided tests (TOST) method with a ± 1.5° margin demonstrated statistical equivalence for MPTA and LDFA. For MPTA, the 90% CI ranged from –1.28° to –0.19° (p = 0.011); for LDFA, the interval was –0.21° to 0.78° (p < 0.001). The HKA angle exceeded the equivalence margin, with a 90% CI from –2.87° to –1.66° (p = 0.416), indicating that postoperative mechanical alignment was not equivalent to the plan. Across all alignment parameters, 18 of 69 individual measurements (26%) fell outside the ± 1.5° equivalence margin, predominantly in the HKA. These outliers represented small deviations that did not correlate with inferior clinical outcomes, suggesting limited clinical significance despite statistical variability.

Bland–Altman analysis further illustrated agreement between planned and achieved values (Figure 3). For MPTA, the mean bias was –0.77° with 95% limits of agreement from –3.24° to 1.70°. For LDFA, the mean bias was –0.30° (–4.04° to 3.44°), with a slight proportional trend indicating undercorrection at higher angles. The HKA showed greater variability, with a mean bias of 1.37° and limits of agreement from –4.13° to 6.87°, reflecting inconsistent restoration of the mechanical axis.

Fig. 3.

Bland-Altman plots comparing planned and achieved alignment angles for A) medial proximal tibial angle (MPTA); B) lateral distal femoral angle (LDFA); and C) hip-knee-ankle (HKA) angle. The solid horizontal line represents the mean bias (difference between planned and achieved values), and the upper and lower lines indicate the 95% limits of agreement. Robotic-assisted alignment demonstrated minimal bias and narrow limits of agreement for MPTA and LDFA, while greater variability was observed for HKA, reflecting differences between supine intraoperative and standing radiological measurements.

Complications

One patient who had not received patellar resurfacing underwent secondary resurfacing of the patellar component at six months postoperatively due to persistent anterior knee pain. No infections, thromboembolic events, or reoperations for aseptic loosening were observed.

Discussion

Robotic-assisted techniques are well established in primary TKA, where they consistently reduce alignment outliers and improve coronal plane accuracy compared with manual instrumentation.^15,16,23 Their application in revision surgery, however, has been far less extensively studied.²⁴ This study demonstrates that robotic-assisted conversion of PKA to TKA using the ROSA system is both technically feasible and clinically effective. All procedures were successfully registered and completed with standard primary components in nearly every case, confirming the practical feasibility of the technique. At 12 months, patients demonstrated substantial improvement in function and pain, with mean OKS increasing from 17.2 to 40.0 and VAS decreasing from 6.7 to 1.6; 83% achieved PASS and 100% exceeded the OKS MCID, confirming meaningful clinical benefit.

Several aspects of the robotic workflow appear advantageous during conversion from PKA to TKA. Preresection landmarking with the prosthesis in situ allowed precise assessment of joint-line distortion and component malalignment despite metallic artefact, guiding bone-preserving resections (Figure 4). Following inverse kinematic alignment (iKA) principles enabled restoration of the native joint line while minimizing collateral releases. Consistent with this, nearly all cases were completed using standard primary components and modest polyethylene thickness (mean 11 mm; 10 to 13), confirming bone and ligament preservation without excessive resection.^17,25,26

Fig. 4.

Intraoperative photograph showing robotic pointer registration on the femoral component of the unicompartmental prosthesis prior to explantation, enabling accurate preresection landmarking.

Radiological accuracy analysis confirmed equivalence between planned and achieved component angles for MPTA and LDFA, indicating precise control at the component level. In contrast, variability was greater for global mechanical alignment (HKA), with approximately one-quarter of measurements (≈ 26%) exceeding the ± 1.5° equivalence margin. These small deviations did not correlate with inferior OKS or VAS outcomes, suggesting limited clinical relevance. This observation aligns with contemporary understanding that coronal alignment within ± 3° of neutral remains functionally acceptable, especially under kinematic or inverse kinematic alignment philosophies.^27-29 The combination of component-level precision and bone-sparing execution highlights the potential of robotics to optimize complex revisions

The high rate of primary component use could be interpreted as a bias toward technically straightforward cases. However, we believe this reflects the advantages of robotic assistance rather than case selection: the system’s precise mapping, controlled resections, and reproducible joint-line restoration expanded the feasibility of using primary implants, even in challenging scenarios. The single LCCK case reflected pre-existing cystic bone loss rather than a failure of robotic planning.

Previous robotic-assisted UKA-to-TKA studies have mainly reported clinical outcomes comparable with primary robotic TKA^30,31 Our study adds quantitative radiological evidence through the combined use of TOST equivalence testing and Bland-Altman analysis, providing new insights into precision and variability in revision settings.

One patient underwent secondary patellar resurfacing for persistent anterior knee pain, consistent with meta-analytic data indicating symptomatic benefit in such cases.³² Our practice has evolved towards selective resurfacing, typically resurfacing in PS designs while retaining the patella in CR and MC designs.³³ No infections, thromboembolic events, or aseptic rerevisions occurred.

Limitations include the retrospective design, small cohort size, and short follow-up. All procedures were performed by a single high-volume robotic surgeon, ensuring consistency but limiting generalizability. The absence of a manual comparator prevents direct evaluation of robotic compared with conventional revision. The study was not powered to detect small differences between implant subtypes or alignment philosophies.

Within these limitations, this study provides early evidence that robotic-assisted conversion from PKA to TKA is technically feasible, radiologically accurate, and clinically effective. Small residual axis variability did not affect outcomes and should be interpreted in the context of standing compared with supine measurements. Future multicentre studies with longer follow-up are warranted to confirm whether these technical advantages translate into sustained functional benefit and cost-effectiveness compared with manual techniques.

Take home message

- This study demonstrates that robotic-assisted conversion of partial to total knee arthroplasty (TKA) is feasible and achieves high radiographic accuracy in component positioning.

- The use of robotic guidance may help surgeons restore alignment more precisely while maintaining an inverse kinematic alignment strategy in a technically demanding revision setting.

- These findings suggest that robotic assistance could improve surgical reproducibility and support satisfactory early clinical outcomes following unicompartmental knee arthroplasty-to-TKA revision.

Data Availability

The datasets generated and analyzed in the current study are not publicly available due to data protection regulations. Access to data is limited to the researchers who have obtained permission for data processing. Further inquiries can be made to the corresponding author.