2.1 Patient selection

Following institutional review board approval (IRB No. ERB-110968; approval date: 2 April 2025), we retrospectively reviewed consecutive patients who underwent primary total knee arthroplasty (TKA) using a navigation system (PRECISION®, Stryker, Michigan, USA) between May 2019 and January 2023 for end-stage primary knee osteoarthritis. The requirement for informed consent was waived due to the retrospective nature of the study.

Beginning in January 2021, functional alignment (FA) was introduced into our clinical practice, and 21 patients subsequently underwent FA-TKA. For comparison, 26 patients who had undergone mechanical alignment (MA)-TKA prior to the introduction of FA (from May 2019 to December 2020) were identified. To reduce potential bias associated with the temporal transition from MA to FA practice, only cases performed using the same implant system (Triathlon CS® Total Knee System, Stryker), a cruciate-substituting design, were included in both groups.

Patients with rheumatoid arthritis, prior fracture around the knee, or previous osteotomy involving the affected knee were excluded prior to matching.

Propensity score matching was performed in a 1:1 ratio based on age, sex, preoperative range of motion (ROM), and preoperative WOMAC pain score. This process yielded 21 matched pairs (42 patients in total). Five MA cases were excluded due to unmatched baseline characteristics (Table 1).

2.2 Surgical techniques

All procedures were performed by four experienced surgeons using an identical medial parapatellar approach and implant design. Perioperative protocols were standardized between groups. The key difference lay in the alignment philosophy.1Navigation-assisted mechanical alignment

MA-TKA was performed according to standard protocols using the navigation system. Tibial and femoral resections were perpendicular to their respective mechanical axes in the coronal plane. Femoral component rotation was set parallel to the surgical epicondylar axis. The posterior tibial slope was determined based on the patient's native slope, limited to a maximum of 8°. Ligament balancing was achieved by releasing tight structures to create rectangular extension and flexion gaps. Following meniscal resection, trial implants were inserted, and intraoperative knee kinematics were recorded using the navigation system.2Navigation-assisted functional alignment

Although FA was initially developed for robotic systems, it can also be performed using the “modified implant positioning” function of an image-free navigation platform.¹⁵ Preoperative 3D-CT planning (ZedKnee®, LEXI Co., Tokyo, Japan) was used to predict implant size and approximate alignment; however, final positioning was determined intraoperatively based on real-time assessment of extension and flexion gaps.

After registration of anatomical landmarks, osteophytes were thoroughly removed. Initial simulation under manual traction aimed for a medial gap of 17 mm in both extension and flexion, with a lateral gap of 18–19 mm, verifying overall limb alignment and bone cut thickness. The target medial gap of 17 mm was based on the implant thickness of the Triathlon® Total Knee System, consisting of an 8-mm femoral component and a tibial component of at least 9 mm. FDA-approved boundaries were applied (6° varus to 3° valgus for HKA and tibial resections; 6° valgus to 3° varus for distal femur; ±6° to the surgical transepicondylar axis for rotation),¹⁶ which encompass approximately 85% of normal anatomical variation.⁴ To mitigate implant failure risk, tibial coronal resection was further restricted within ±3° of the mechanical axis.¹⁷

After tibial resection and removal of posterior osteophytes, the second simulation was performed using a tensor with 30 lbs of traction, and final femoral component positioning was decided. If the simulated alignment exceeded boundaries, selective soft-tissue release was performed and reassessed. Menisci were excised, trial components inserted, and intraoperative kinematic data were collected at 0°, 30°, 45°, 60°, 90°, and maximum flexion, including flexion-extension angle, coronal limb alignment (varus = negative, valgus = positive values), tibial rotation, and femoral anteroposterior translation.

2.3 Data collection

Patient demographics, intraoperative records, and patient-reported outcomes (pain visual analog scale [VAS] at rest and during walking, and Western Ontario and McMaster Universities Osteoarthritis Index [WOMAC]) were retrieved from electronic medical records preoperatively and at 6 months postoperatively.

For both groups, final intraoperative kinematic parameters—including range of motion (extension and flexion), coronal limb alignment throughout flexion, tibial internal rotation, and posterior femoral translation—were extracted from the navigation database. Kinematic assessment was performed according to a previously described navigation-based protocol.¹⁸ During intraoperative kinematic evaluation, the dissected fascia was temporarily approximated using forceps to simulate soft-tissue tension. Kinematic assessment was performed once per knee by the operating surgeon using the navigation system. The knee was passively flexed by supporting the patient's heel in the surgeon's open palm to allow free tibial rotation, while the other hand stabilized the thigh. Care was taken to avoid intentional rotational stress during flexion. Intraoperative extension and flexion angles were measured under gravity without additional manual force.¹⁹ The coordinate system was defined according to the navigation registration protocol, in which femoral and tibial mechanical axes were established using digitized anatomical landmarks. Axial rotation and coronal alignment were calculated relative to the tibial mechanical axis, with all parameters expressed with reference to the femoral center, as previously described by Wada et al.¹⁸ The navigation system automatically recorded tibial rotation (internal rotation defined as positive) and coronal limb alignment (valgus defined as positive) at maximum extension, 0°, 30°, 45°, 60°, and 90° of flexion, as well as at maximum flexion during passive motion. Measurement accuracy was 0.5°, and results are reported to one decimal place. Previous studies have demonstrated acceptable repeatability and reproducibility of this methodology.

Radiographic assessment included the medial proximal tibial angle (MPTA), lateral distal femoral angle (LDFA), and hip–knee–ankle (HKA) angle obtained from long-leg standing radiographs. Patellar tilt angle (PTA) was measured on skyline radiographs at 60° flexion one week postoperatively. Coronal phenotypes were classified according to the Coronal Plane Alignment of the Knee (CPAK) system,²⁰ based on arithmetic HKA (aHKA = MPTA − LDFA) and joint line obliquity (JLO = MPTA + LDFA).

2.4 Statistical analysis

Propensity score matching was performed to reduce baseline imbalance between groups given the modest sample size. Propensity scores were estimated using a logistic regression model with group assignment (FA vs MA) as the dependent variable and the following covariates: age, sex, baseline extension and flexion range of motion, and preoperative WOMAC pain score. One-to-one nearest-neighbor matching without replacement was conducted using the MatchIt package with logit distance and no caliper. Patients with missing values in matching covariates were excluded (complete-case matching). Covariate balance after matching was assessed using standardized mean differences (SMDs), with absolute values < 0.2 considered indicative of acceptable balance. After matching, all covariates demonstrated adequate balance (all absolute SMDs <0.2). All subsequent analyses were performed using the matched cohort only.

Normality was assessed by the Shapiro–Wilk test. Continuous variables were summarized as mean (SD) when normally distributed and as median (IQR) otherwise. Group comparisons used independent-samples t-tests, with Wilcoxon–Mann–Whitney tests as robustness checks. Categorical variables were compared using χ² or Fisher's exact tests.

Kinematic data were analyzed by two-way mixed ANOVA (Group × Angle), testing: (1) internal–external rotation, (2) varus–valgus angle, and (3) femoral posterior translation. Statistical significance was set at p < 0.05.

As a supplementary assessment of statistical precision, we calculated the minimum detectable difference (MDD) for postoperative maximum flexion given the matched sample size (21 per group), assuming a two-sided α of 0.05 and 80% power, using the pooled standard deviation observed in the matched cohort.

4.1 Kinematic findings

The principal finding of this study is that functional alignment was associated with differences in axial rotational behavior. Compared with mechanical alignment, FA demonstrated greater mean tibial internal rotation during flexion. Although a significant main effect of group was observed, post hoc comparisons confirmed a statistically significant between-group difference primarily at maximum flexion, suggesting that FA influences overall rotational behavior.

This rotational pattern is consistent with the gap-based philosophy of FA. By prioritizing medial stability while permitting relatively greater lateral laxity in flexion, FA may facilitate posterior movement of the lateral femoral condyle, thereby increasing tibial internal rotation. In native knee kinematics, internal rotation during flexion reflects posterior translation of the lateral femoral condyle around a relatively stable medial compartment, creating a medial pivot pattern.¹⁴ Recent investigations of FA-TKA have similarly reported alterations in axial rotational behavior and contact patterns consistent with this medial pivot concept, further supporting the linkage between gap balancing strategy and rotational mechanics.^12,13 In this context, the increased internal rotation observed in the FA group may reflect a rotational pattern consistent with aspects of native knee kinematics, although restoration of full physiological knee mechanics cannot be assumed.

Importantly, this effect was predominantly rotational rather than translational. Although posterior femoral translation did not differ significantly between groups, FA demonstrated a modest mean posterior shift of approximately 1-2 mm. The absence of statistical significance may relate to sample size limitations or to the passive intraoperative assessment environment. More fundamentally, increased axial rotation does not necessarily require proportional sagittal translation, particularly in a cruciate-substituting implant where rollback mechanics differ from cruciate-retaining designs.

The association between internal tibial rotation and flexion capacity has been reported previously. Matsuzaki and Ishida demonstrated that greater intraoperative internal rotation correlates with improved postoperative flexion outcomes.^21,22 Our findings extend this concept by showing that alignment strategy itself may influence rotational behavior, thereby providing a potential mechanistic explanation for flexion differences observed between alignment philosophies.

However, caution is warranted. Deep flexion in this study was assessed intraoperatively under passive, gravity-assisted, non–weight-bearing conditions without external manual force,¹⁹ and therefore may not reflect active, weight-bearing knee mechanics during functional activities. While FA promoted a rotational pattern that resembles aspects of native kinematics, this should not be interpreted as complete biomechanical normalization.

4.2 Alignment findings

Whereas FA predominantly influenced axial rotational behavior during flexion, its effect on static coronal limb alignment was comparatively modest. FA demonstrated greater varus alignment during intraoperative kinematic assessment; however, this difference was not statistically significant on postoperative radiographs. This apparent discrepancy likely reflects the distinction between dynamic intraoperative measurements obtained under navigation guidance and static standing radiographic evaluation. Notably, the mean postoperative HKA angle in the FA group was approximately −1°, indicating a mild tendency toward varus alignment compared with MA, though within the predefined restricted alignment boundaries.

Beyond global limb alignment, postoperative radiographs revealed a more medially inclined tibial component in the FA group, consistent with a tendency toward constitutional varus and joint line obliquity as described by Bellemans et al..⁴ While such alignment characteristics may approximate native joint orientation and may contribute to the rotational behavior observed intraoperatively, they also raise theoretical concerns regarding long-term tibial component loading. The restricted alignment philosophy of FA is intended to prevent excessive deviation from neutral mechanical alignment, thereby balancing kinematic restoration with implant durability. Importantly, long-term data from unrestricted kinematic alignment techniques have demonstrated favorable 10-year implant survival,⁶ suggesting that controlled alignment modification may not necessarily compromise implant longevity, although continued longitudinal follow-up remains essential.

The greater PTA observed in the FA group may reflect internal femoral rotation applied during flexion-gap balancing combined with a restricted tibial varus setting. This differs from previous reports in which no significant PTA differences were observed when greater tibial varus inclination was permitted.²³ Because most contemporary implant designs were originally optimized for neutral mechanical alignment principles, alignment-strategy–dependent variations in component orientation may influence patellofemoral tracking. Future refinement of implant geometry tailored to diverse alignment philosophies may therefore help mitigate potential patellofemoral imbalance.

4.3 Patient-reported outcomes

Despite the observed differences in intraoperative kinematics and alignment, short-term PROMs at six months were comparable between FA and MA in the present cohort. This finding is consistent with prior comparative studies demonstrating minimal or no early differences in patient-reported outcomes between alignment strategies. Parratte et al. reported slightly faster early recovery at six months with anatomo-functional alignment; however, clinical scores converged by one year.²³ Similarly, Young et al. found no significant overall PROM differences at two years in a large randomized trial, despite FA requiring less soft-tissue release and yielding marginally higher satisfaction and recommendation rates.²⁴

Emerging evidence suggests that potential benefits of FA may be phenotype-dependent rather than universal. In particular, recent analyses have indicated that patients with constitutional varus morphology—such as those classified as CPAK Type I—may experience higher Forgotten Joint Scores and KOOS-QOL following FA compared with MA.^24,25 These observations raise the possibility that alignment strategies restoring individualized joint line orientation may confer perceptual advantages in selected subgroups.

In the present study, the absence of PROM differences at six months may reflect the relatively short follow-up period and the inclusion of heterogeneous coronal phenotypes. Early postoperative recovery is influenced by multiple factors beyond alignment and kinematic patterns, including pain control, rehabilitation, and patient expectations. Longer-term follow-up and phenotype-stratified analyses are therefore necessary to determine whether alignment-strategy–dependent kinematic differences translate into sustained patient-perceived functional improvements.

4.4 Limitations

This study is limited by its retrospective design, modest sample size, and short follow-up duration. Although propensity score matching was employed to reduce baseline imbalance, the matched cohort represents a selected subset of the original population, and estimates derived from small-sample matched analyses should be interpreted cautiously as mechanistic observations rather than confirmatory evidence. Based on the observed variance in the matched cohort, the present study had 80% power to detect between-group differences of approximately 8° in maximum flexion. Accordingly, the study may have been underpowered to detect smaller differences.

Second, intraoperative kinematic assessment was performed under passive, non–weight-bearing conditions during anesthesia. The absence of muscle activation and physiologic loading may influence tibiofemoral mechanics, particularly in deep flexion, and therefore may not fully reflect in vivo functional kinematics during daily activities.

Third, kinematic measurements were obtained once per knee due to practical surgical time constraints. Although prior studies have demonstrated acceptable repeatability of navigation-based assessment, the lack of repeated measurements may limit assessment of intraobserver variability.

Finally, posterior femoral translation was calculated based on navigation-derived reference points rather than compartment-specific contact point analysis. As such, direct comparison with fluoroscopic or sensor-based studies evaluating medial and lateral condylar rollback should be interpreted with caution.

Longer-term prospective studies incorporating weight-bearing dynamic assessments and phenotype-stratified analyses are warranted to determine whether alignment-strategy–dependent kinematic differences translate into sustained improvements in implant survival and patient-reported outcomes.