3.1 Demographic data

A total of 3208 knees were included, of which 1901 received MA and 1307 received rKA. The average age of patients receiving MA was 69.2 years (range, 63 - 75.5) whilst the average age of patients receiving rKA was 70.4 years (range, 67 - 76.4). The MA population had an average BMI of 29.8 kg/m² (range, 25.2 - 35) with the average BMI being 29.2 kg/m² (range, 26.4 - 32) in the rKA population. A male: female ratio of 999:835 and 594:644 for the MA and rKA groups respectively was observed across studies which reported on these domains. The average follow-up was 20.3 months (range, 6-53).

Few studies report or analyse postoperative outcomes in the context of demographic data. Winnock De Grave et al. reported decreasing postoperative Oxford Knee Score (OKS) with univariable analysis when adjusted for age (p = 0.047) and worse OKS for women on multivariable analysis (p = 0.045).²² Ma et al. also carried out demographic subgroup analysis and found no significant difference for gender and BMI on Knee Injury and Osteoarthritis Outcome Score (KOOS), EQ (5D) and Forgotten Join Score (FJS).¹⁹ However, their analysis did reveal worse outcomes on these aforementioned PROMs when adjusted for age (p < 0.05).¹⁹ This is, however, insufficient evidence to ascertain whether age, gender or BMI have an effect on PROMs following TKA. Furthermore, there was no analysis within the included studies in relation to radiological outcomes. Other studies have supported the role of age, gender and BMI in increasing the rate of adverse outcomes following TKA, but this needs further consideration.²⁸

3.2 Bias assessment of included studies

Of the three included RCTs, one had a low risk of bias as assessed with RoB-2 and two had some concerns as seen in Fig. 2. Due to the nature of the intervention, surgeons were not blinded at the point of delivery. This was mitigated by Matsumoto et al.¹⁶ who reported clear blinding of intraoperative outcome assessors. Ettinger et al.²⁰ had a relatively small sample size with a loss to follow-up of more than 5% (see Fig. 3).

Close

Fig. 2 Risk of Bias 2 summary of RCTs.

Export to PPT

Close

Fig. 3 Pooled mean difference of PROMs.

Export to PPT

The NOS tool includes assessment of three aspects: selection of study subjects, comparability between groups and outcome measurement. The total score is 9 with 5 – 9 accepted as higher quality observational research. According to the NOS criteria, all 17 included cohort studies had a score ≥5 which is considered to be the acceptable standard to minimise risk of bias.

3.3 Clinical scores

Five studies reported KSS outcomes, of which three demonstrated results favouring rKA. Abhari et al.¹⁷ observed significant improvements in both clinical and functional KSS subscales in the rKA group. Specifically, the clinical KSS improved from 41 ± 9.4 to 93 ± 9.4, and the functional KSS improved from 51 ± 10 to 85 ± 16, both with p-values <0.001. Matsumoto et al.¹⁶ also reported significantly higher functional scores in rKA patients (73.6 ± 13.7) compared to those who received MA (63.8 ± 18.8, p = 0.04), alongside significantly better outcomes in objective indicators (p < 0.001), patient satisfaction (p = 0.036), and overall functional activity (p = 0.008). Ettinger et al.²⁰ demonstrated statistically significant improvements in the satisfaction and expectation subscales at both 12 and 24 months in the rKA group, with KSS function showing significant improvement at 24 months (84.3 ± 12.5 for rKA vs. 77.9 ± 13.4 for MA, p = 0.028). The same study reported satisfaction scores of 34.0 ± 6.7 (rKA) vs. 30.7 ± 7.1 (MA) and expectation scores of 10.4 ± 2.7 (rKA) vs. 9.1 ± 2.5 (MA), both favouring rKA with p-values <0.05.

Two studies reported no significant differences in KSS between the alignment strategies. Sappey-Marinier et al.²¹ observed improvements in both groups (52.0 ± 30.4 vs. 57.5 ± 29.8), but without statistically significant differences across any of the subscales. Similarly, Parente et al.²⁴ reported no significant difference in subjective KSS (89.9 ± 2.6 for rKA vs. 88.5 ± 4.6 for MA) or functional KSS (88 ± 4.1 vs. 86.6 ± 8.5).

Pooled analysis from the RCTs yielded a significant mean difference in KSS favouring rKA, with a pooled mean difference of 7.3 points (95% CI: 2.97 to 11.70).

Two studies assessed outcomes using the WOMAC score. Abhari et al.¹⁷ reported a substantial improvement in WOMAC scores for the rKA group, increasing from 43 ± 15 to 90 ± 11 postoperatively, compared with a postoperative score of 85 ± 1.8 in the MA group (p = 0.039). In contrast, Ettinger et al.²⁰ found no significant differences at either 12 or 24 months postoperatively. The pooled analysis demonstrated a non-significant trend favouring rKA, with a mean difference of 2.4 points (95% CI: −0.24 to 5.10).

KOOS outcomes were reported in five studies. Abhari et al.¹⁷ found a significant improvement in KOOS for the rKA group compared to the MA group (87 ± 15 vs. 78 ± 1.9, p = 0.001). Ma et al..¹⁹ reported a statistically significant improvement at both early (T1, 10 days post-op) and later (T6, 6 months post-op) follow-up points. At T1, KOOS improved by 26.053 in the rKA group and 18.607 in the MA group (p < 0.001), and at T6, the improvements were 51.017 and 46.896 respectively (p = 0.023). However, three studies—Kobayashi et al.,¹⁵ Parente et al.,²⁴ and MacDessi et al.¹⁸ - found no significant differences in KOOS scores between the two alignment groups.

Four studies assessed the Forgotten Joint Score (FJS). Ettinger et al.²⁰ reported a significant difference favouring rKA at 12 months post-op (62.2 ± 22.9 vs. 52.4 ± 23.8, p = 0.044), though this difference was not sustained at 24 months (65.1 ± 23.2 vs. 56.9 ± 26.9, p = 0.127). The remaining studies—Kobayashi et al.,¹⁵ Ma et al.,¹⁹ and MacDessi et al.¹⁸ - did not demonstrate any statistically significant differences in FJS scores between rKA and MA.

Oxford Knee Score (OKS) was assessed in three studies,^15,20,22 none of which found a statistically significant difference between the two alignment methods. EQ-5D scores were reported in two studies.^18,19 Ma et al.¹⁹ found significantly greater improvements in the rKA group at both 10 days (0.457 vs. 0.367, p < 0.001) and 6 months (0.606 vs. 0.565, p = 0.01) postoperatively, whereas MacDessi et al.¹⁸ observed no difference. Visual Analogue Scores (VAS) were reported in two studies. Winnock De Grave et al.²² reported significantly better pain scores in the rKA group (9.2 ± 0.8 vs. 8.5 ± 1.3, p = 0.012) using a tibia-first technique (inverse restricted kinematic alignment). Parente et al.²⁴ found no statistically significant difference (see Table 2).

3.4 Lower limb alignment

Several radiographic measurements were recorded for analysis of lower limb alignment, as seen in Table 3. These included HKA, lateral distal femoral angle (LDFA) and medial proximal tibial angle (MPTA). Amongst rKA participants, the average preoperative and postoperative HKA was 174.0 and 178.5, respectively. The average preoperative and postoperative LDFA was 88.5 and 89.4, respectively. The average preoperative and postoperative MPTA was 86.7 and 88.1, respectively. In the MA group, the average preoperative and postoperative HKA was 174.6 and 179.0, respectively. The average preoperative and postoperative LDFA was 88.7 and 89.8, respectively. The average preoperative and postoperative MPTA was 86.8 and 89.7, respectively.

Various other alignment parameters were recorded across studies. Kobayashi et al.¹⁵ reported on Joint Line Orientation Angle (JLOA) and found a significant difference in postoperative values between the two comparators (in favour of rKA). In the MA group, the mean preoperative JLOA was −0.8° (range: −8.5 to 4.0, SD: 2.8), as compared to −0.3° (range: −6.5 to 4.0, SD: 2.7) in the rKA group (p = 0.326). In the MA group, the mean postoperative JLOA was −4.1° (range: −7.0 to −0.5, SD: 1.6) versus a mean postoperative JLOA of −2.0° (range: −6.0 to 3.0, SD: 2.0) in the rKA group (p < 0.001).

Matsumoto et al.¹⁶ found a significant improvement in tibial internal rotation between 60 and 120° of flexion in the kinematic group, with a mean rKA value of 21.1° ± 7.5 and a mean MA value of 15.8° ± 10.1 (p = 0.030). Sappey-Marinier et al.²¹ observed a statistical difference in tibial slope between rKA and MA groups (88.6 ± 1.1° in MA versus 86.6 ± 1.4° in rKA, p < 0.001), however found no significant difference in patellar tilt between the two groups (0.7 ± 3.5° in MA versus 0.8 ± 3.4° in rKA).Winnock De Grave et al.²² found no significant difference in tibial slope, with a mean MA value of 4.1 ± 1.6° and a mean irKA value of 4.2 ± 1.2°.

3.5 Complications

Complication rates were not invariably reported. Sappey-Marinier et al.²¹ reported a 16% revision rate for aseptic tibial loosening amongst rKA participants. All revisions occurred between 11.9 and 25.5 months postoperatively. In the MA group, 2 patients were revised due to aseptic tibial loosening at 38.4 and 45.1 months postoperatively. Additionally, 1 patient developed an early infection 2 months postoperatively, leading to a complete revision with component exchange. 1 patient underwent arthroscopic arthrolysis at 4 months postoperatively due to stiffness. At 50 months follow-up, the overall implant survivorship was significantly higher for the MA group compared to the rKA group (97% vs, 84% respectively, P = 0.002). Kobayashi et al.,¹⁵ Ma et al.,¹⁹ Ettinger et al.,²⁰ Winnock De Grave et al.²² and Parente et al.²⁴ reported no significant tendency for complications and/or implant survival between groups.

5.1 PROMs and clinical outcomes

Multiple studies demonstrated improved functionality, early proprioception and joint awareness in favour of rKA.^16,17,20 PROMs highlighted this consistently, with rKA either equivocal or superior to MA across various domains.^{16,17,19,20,22} For example, rKA has shown statistically superior KSS function scores at 24 months (p = 0.028),²⁰ higher FJS at 12 months (indicating improved neurophysiological adaptation to the prosthesis)²⁰ and non-inferiority across WOMAC, KOOS and UCLA activity scores, suggesting that rKA does not compromise functional recovery. However, a key surgical consideration is whether these apparent improvements translate into long-term biomechanical advantages or whether they are transient benefits attributable to early adaptation and soft tissue accommodation.

5.2 Radiographic outcomes

One of the fundamental concerns associated with rKA is the potential for excessive coronal deviation, particularly in the varus knee phenotype. Results demonstrated, however, that postoperative HKA remains within ±3° of neutral across most studies, supporting rKA's claim of maintaining a controlled alignment philosophy ^{15,17,19–21}. MPTA and LDFA adjustments ensured that resections remained within implant tolerances, mitigating concerns over component malpositioning. Nonetheless, an important surgical nuance emerged in the strategy employed for tibial resection. Unlike MA, which relies on perpendicular tibial cuts, rKA retains a slight native varus in constitutional varus knees, potentially altering load distribution across the polyethylene insert. This then raises concerns over asymmetrical wear patterns, especially in cases where posterior tibial slope (PTS) and component rotation are not precisely optimised.^29,30 Whilst short-term radiographic and alignment assessments suggested this approach remained within safe tolerances, long-term polyethylene wear studies will be imperative in determining whether this strategy has an impact on implant survivorship.

5.3 Surgical implications

Whilst rKA seemingly provided tangible functional benefits, its intraoperative execution was highly variable. A critical distinction across studies was the lack of a universal rKA technique. Some surgeons utilised robotic-assisted planning whereas others employed calipered kinematic adjustments or navigation-assisted resections ^{16–19,22,23}. This raised several key surgical considerations; whether rKA should follow a systematic algorithm for tibial slope and coronal cut restrictions or be case-dependent, whether rKA should strictly adhere to native posterior condylar reference points or if adjustments should be made based on intraoperative soft tissue tensioning, and whether rKA should be used preferentially in cases with preoperative ligament imbalance (i.e. medial laxity in varus knees), given that rKA theoretically reduces the requirement for soft tissue releases.¹⁰ Addressing these questions is vital, as inconsistent surgical execution may introduce variability in clinical outcomes which will increase the difficulty of establishing rKA as a new standard.

5.4 Complications and implant longevity

One of the most controversial aspects of rKA identified, was its implications for implant longevity. Results suggested no statistically significant difference in complication rates between rKA and MA however isolated studies raised concerns about aseptic tibial loosening ^{15,19–22,24}. The biomechanical rationale behind this concern is that retaining slight native varus alignment may increase compressive forces on the medial tibial plateau, potentially exceeding polyethylene or cement tolerances over time. Despite this, cases of implant loosening have not been consistently observed across studies, suggesting that patient selection, component, design and surgical technique may have been confounding variables. Notably, no included study reported an exceedingly high failure rate with rKA, nor was there a clear signal for increased revision risk. Morcos et al.³¹ conducted a prospective, longitudinal cohort investigation of 104 consecutive primary rKA TKAs. Computer-assisted navigation (Orthomap ASM, Stryker) and cemented cruciate-retaining implants (Triathlon CR, Stryker) were used. The cohort included 89 patients (104 knees), with 21 men and 83 women. The mean age at surgery was 67 years (range 33–85), and the mean follow-up was 11.3 years (range 10–13). The study reported implant survivorship of 99.0% at a mean follow-up of 11.3 years, with only one early revision (1.0%) due to femoral component malrotation causing instability and recurrent hemarthroses, which was successfully treated by femoral component revision. One patient sustained a femoral periprosthetic fracture at 5 years, treated with open reduction and internal fixation without implant revision. Furthermore, Vermue et al.³² employed a retrospective cohort study including 143 eligible patients who underwent primary rKA-TKA with a rotating-platform deep-dish design. Patient demographics included a mean age of 71.3 ± 8.6 years, mean BMI of 30.0 ± 6.0 kg/m² with 69% women and 31% men. Charnley classification (A- 40%, B1- 51%, C1- 1%, C3- 7%) and Devane classification (3- 36%, 4- 62%, 5- 2%) were reported. Follow-up lasted 5 ± 1 years, with 123 patients completing this. The study reported a revision-free survival rate at five years of 98%. Complications observed during the five-year follow-up included periprosthetic joint infection (1.6%), arthrofibrosis (1.6%), aseptic loosening of a cementless femoral component (0.8%) and patellar dislocation (0.8%). These results indicate excellent mid-term and long-term implant survivorship with a low rate of complications.

5.5 Robotic-assisted versus manual techniques

MA and rKA were performed using both traditional manual instrumentation (jig-based) and robotic-assisted techniques, each with distinct biomechanical implications, precision advantages and clinical trade-offs. Analysis of the included studies revealed a trend towards improved alignment accuracy and soft tissue balance with robotic-assisted techniques for both MA and rKA.^17,22,23 However, the clinical significance of these refinements remains debated. For MA, robotic-assisted surgery consistently improved HKA accuracy, reducing outliers beyond ±3° of neutral alignment. However, there was no clear functional advantage over manual MA in PROMs. For rKA, robotic guidance allowed for more precise restoration of native knee morphology, particularly in tibial slope and femoral component rotation.²² While robotic-assisted rKA appeared to offer greater intraoperative control over bone resection and GB, its functional superiority over manual rKA remains unproved. Given the inherently patient-specific nature of rKA, the argument can be made that precision execution is more critical in rKA than MA thereby justifying the added cost and complexity of robotic systems. Deckey et al.³³ deduced that robotic-assisted TKA achieved high precision in implant placement, with minimal deviations in coronal and rotational alignment. Furthermore, a meta-analysis conducted by Alrajeb et al.³⁴ concluded that robotic TKA resulted in significantly better post-operative anatomical (OR − 0.82; 95% CI, −1.027 to − 0.58, p < 0.00001) and mechanical restoration (OR − 0.95; 95% CI, −1.49 to − 0.41, p < 0.0006) compared to conventional jig-based TKA. Additionally, knee range of motion (OR − 2.23; 95% CI − 4.89–0.43, p = 0.1) and femoral prosthesis position (OR − 0.98; 95% CI, −2.03–0.08, p = 0.07) also favoured robotic knees, though these differences did not reach statistical significance.

5.6 Future directions

To establish rKA as a widely accepted alignment philosophy, future research should prioritise RCTs with standardised surgical protocols to eliminate variability in execution, longitudinal studies (>10 years) evaluating polyethylene wear patterns, aseptic loosening rates and implant failure mechanisms, and standardised reporting of adverse events and effect sizes to facilitate comparisons across future studies.^35,36 Additionally, patient selection criteria must be refined. Whilst rKA appeared to be beneficial for most primary TKAs, its role in high-demand patients, post-traumatic knees, and severe varus/valgus deformities is unclear.