1 Introduction

Among the various parameters influencing alignment, the coronal orientation of the knee joint line —commonly referred to as joint line obliquity (JLO), joint line orientation, or joint line orientation angle (JLOA)—has received increasing attention in both high tibial osteotomy (HTO) and total knee arthroplasty (TKA).

In the setting of HTO, several studies have demonstrated that altering the native joint line through inappropriate patient selection or excessive correction can result in a laterally tilted JLO or JLOA ^1–5. This alteration may lead to elevated shear stress across the lateral compartment, rather than simple lateral load redistribution, thereby potentially accelerating cartilage degeneration and joint deterioration ^1–4. Babis et al.⁵ emphasized the importance of preserving JLO by employing a double-level osteotomy technique, arguing that excessive lateral tilt following HTO can disrupt normal knee kinematics and compromise long-term outcomes.

In the context of TKA, the coronal orientation of the joint line has become a key consideration with the advent of alignment strategies such as kinematic alignment.^6,7 Moreover, this parameter is a central component of the recently proposed Coronal Plane Alignment of the Knee (CPAK) classification system, which aims to capture individual knee phenotypes and assist in personalized surgical planning.⁸ Many recent studies have highlighted the clinical relevance of JLO in TKA. While earlier literature emphasized that kinematically aligned TKA (KA-TKA) leads to a more horizontal joint line and superior clinical outcomes compared to mechanically aligned TKA (MA-TKA) ^9–12, some findings have shown that patients with a laterally tilted JLOA—typically associated with less horizontal orientation—demonstrated more favorable outcomes ^13–15. These findings suggest that the relationship between JLOA and functional outcomes may be more nuanced than previously understood.

Despite increasing interest in JLO and JLOA, no consensus has been established regarding clinically meaningful thresholds. Maintaining parallelism of the knee joint surface relative to the floor may be important; however, this can be compensated for by adjacent joints or may vary dynamically during activity. Therefore, reference values for JLOA and related radiographic parameters in healthy adults are needed, along with investigations into factors influencing JLOA. Furthermore, if restoration of CPAK types is to be set as a surgical target in procedures such as KA-TKA, it is essential to analyze differences in JLOA across CPAK types and the bilateral symmetry of these parameters between right and left limbs, thereby clarifying their potential clinical implications for achieving personalized alignment.

The present study aims to evaluate knee-JLOA (K-JLOA) and related radiographic parameters, including the mechanical hip–knee–ankle angle (mHKA), medial proximal tibial angle (MPTA), lateral distal femoral angle (LDFA), arithmetic hip–knee–ankle angle (aHKA), joint line convergence angles of the knee and ankle (K-JLCA, A-JLCA), talar inclination, and knee-intercondylar and ankle-intermalleolar distances (K-ICD, A-IMD) in a healthy population. Specifically, we sought to identify factors affecting JLOA, compare K-JLOA across CPAK types, and analyze bilateral symmetry of these parameters and CPAK types between right and left knees.

We hypothesized that K-JLOA and related radiographic parameters in healthy adults would demonstrate consistent normative ranges, providing reference values for future clinical application. We further hypothesized that K-JLOA would be associated with tibial and femoral geometry, as well as with parameters reflecting joint space and foot position. In addition, we anticipated that K-JLOA would vary across CPAK types. Finally, we hypothesized that while most radiographic parameters would show strong bilateral concordance between right and left knees, CPAK phenotypes would demonstrate lower agreement due to the variability of aHKA and JLO and the impact of classification boundary effects.

2 Materials and methods

This study was designed as a retrospective observational analysis of standing, full-length anteroposterior long-leg radiographs (LLRs) obtained from healthy adults aged 20–50 years. This study was approved by the Institutional Review Board (IRB No. 2025-11-020). The requirement for informed consent was waived due to the retrospective nature of the study and the use of anonymized imaging data.

Radiographic measurements of K-JLOA and related parameters were performed on the LLRs. Radiographs were screened from institutional imaging archives between 2020 and 2024, and subjects were included only if they met the predefined inclusion and exclusion criteria. After this selection process, a total of 100 individuals (200 knees) were included for final analysis. Inclusion criteria were the absence of radiographic osteoarthritis (Kellgren–Lawrence grade ≤1), no prior history of lower limb surgery or trauma, and symmetric lower limb alignment without gross deformity. Exclusion criteria included radiographs with technical errors, inadequate visualization, or malpositioning, such as excessive internal or external rotation or flexion contracture. Although this was a retrospective study, all LLRs were obtained using a standardized institutional protocol. Radiographs were acquired under full weight-bearing conditions with the patellae oriented forward and the feet placed shoulder-width apart on a pre-marked plate to ensure consistent foot positioning. Radiographs that did not meet these criteria were excluded from analysis.

Imaging data were analyzed using the digital goniometer and caliper tools available in the Picture Archiving and Communication System (PACS) software (INFINITT, Seoul, Korea). The following radiographic parameters were measured or calculated: mHKA, MPTA, LDFA, aHKA, JLO, knee and ankle JLOA (K-JLOA and A-JLOA), K-JLCA and A-JLCA, talar inclination, K-ICD, and A-IMD (Fig. 1). To identify factors influencing K-JLOA, multiple regression analysis was performed using these radiographic parameters entered as independent variables.

Close

Fig. 1 Measurement methods of radiographic parameters related to joint line orientation of knee and ankle mHKA, mechanical hip-knee-ankle angle; MPTA, medial proximal tibial angle; LDFA, lateral distal femoral angle; fMA, mechanical axis of the femur; tMA, mechanical axis of the tibia; a, femoral joint line tangent to both femoral condyles; b, tangential line at the distal tibia plafond; c, tangential line at the talar dome; d, horizontal reference line parallel to the floor; K-JLOA, joint line orientation angle of the knee (a–d): as shown in this figure, medial inclination is expressed as a negative value; A-JLOA, joint line orientation angle of the ankle (b–d): as shown in this figure, lateral inclination is expressed as a positive value; K-JLCA, joint line convergence angle of the knee (A): as shown in this figure, lateral opening is expressed as a positive value; A-JLCA, joint line convergence angle of the ankle (B); K-ICD, intercondylar distance between both knees; A-IMD, intermalleolar distance between both ankles.

Export to PPT

To compare K-JLOA across CPAK types, group differences were analyzed using one-way ANOVA; post-hoc pairwise comparisons were performed with Tukey's HSD and Duncan's multiple range test.

For bilateral symmetry analyses, right–left associations for each parameter (mHKA, MPTA, LDFA, aHKA, JLO, K-/A-JLOA, K-/A-JLCA, talar inclination) were quantified using Pearson correlation coefficients (r), while concordance of CPAK phenotypes between contralateral knees was expressed as percentage agreement and summarized in a cross table.

The minimum difference that the PACS software could detect was 0.1° and 0.1 mm. Two independent investigators measured all of the radiographic parameters to reduce observation bias. Inter- and intraobserver reliabilities of the radiographic variables were assessed using the intraclass correlation coefficient (ICC), based on a two-way random-effects model with absolute agreement for averaged measurements (ICC[2,2]). All measurements demonstrated excellent reliability (ICC >0.80). The ICC for K-JLOA was 0.91 (95% CI, 0.86–0.95). Therefore, the mean values of the two observers were used for analysis.

3 Results

Analysis of 100 LLRs from healthy adults demonstrated the typical distribution of key alignment parameters in non-arthritic knees. The mean mHKA was −1.0° ± 2.8°, with a distribution skewed slightly toward varus alignment. The MPTA averaged 86.5°, and the LDFA averaged 86.8°. The K-JLOA averaged −3.2°, ranging from −8.9° (medial tilt) to 7.2° (lateral tilt). The A-JLOA averaged 0.6°, ranging from −16.1° (medial tilt) to 12.9° (lateral tilt). The K-JLCA and A-JLCA were 0.9° (range, −6.0° to 8.9°) and 0.7° (range, −5.8° to 9.9°), respectively. Mean K-ICD was 95.8 mm, and mean A-IMD was 109.6 mm. A summary of all parameters is presented in Table 1.

The study population consisted of 57 males and 43 females. Sex-based comparisons of radiographic parameters according to side are summarized in Supplementary Table 1. For both knees, most radiographic parameters did not differ significantly between sexes, with the exception of mHKA, aHKA, and A-JLCA.

In multiple regression analysis to identify independent radiographic predictors of K-JLOA, stepwise selection identified five significant parameters: LDFA, MPTA, K-JLCA, A-JLCA, and A-IMD. The final model was statistically significant (F = 43.17, p < 0.001) and explained 52.6% of the variance in K-JLOA (adjusted R² = 0.526) (Table 2). K-JLOA was positively associated with MPTA and LDFA, but negatively associated with K-JLCA, A-JLCA, and A-IMD (Table 2).

The most frequently observed types were I, II, and III, with mean K-JLOA values ranging from −2.8° to −4.3°, demonstrating significantly greater medial inclination compared with rare types (Table 3). One-way ANOVA revealed significant differences in K-JLOA across CPAK types (F[6,193] = 7.63, p < 0.001). Post-hoc Tukey tests showed that types II and III demonstrated significantly greater medial inclination compared with types IV and V (both p < 0.05), while type IX exhibited significantly more lateral inclination than types I–III (all p < 0.05). Notably, type I showed comparable K-JLOA values to types V and VI (p > 0.05), and the “neutral axis” group (types V–VI) still exhibited medially inclined JLOA.

Most radiographic parameters demonstrated significant bilateral concordance (all p < 0.001). Strong correlations were observed for mHKA (r = 0.762), A-JLOA (r = 0.782), and talar inclination (r = 0.709) (Table 4). In contrast, JLO (r = 0.379), aHKA (r = 0.456), K-JLOA (r = 0.364), and K-JLCA (r = 0.315) showed weaker correlations, indicating substantial variability between right and left knees (Table 4). Bilateral concordance of CPAK phenotypes was poor. Only 31% of subjects had both knees classified into the same CPAK type, while the majority exhibited contralateral knee discordance, with paired knees frequently assigned to different CPAK quadrants (Table 5; Fig. 2).

Close

Fig. 2 Scattergram showing bilateral asymmetry between right and left knees in arithmetic hip–knee–ankle angle (aHKA), joint line obliquity (JLO), and Coronal Plane Alignment of the Knee (CPAK) phenotype Blue circles represent right knees. Red circles represent left knees. Each gray line connects the two knees of the same subject. Substantial right–left asymmetry was observed in many individuals, with paired knees frequently located in different CPAK quadrants.

Export to PPT

4 Discussion

The key finding of this study is that K-JLOA was only modestly influenced by CPAK type, and joint line–related parameters showed weaker bilateral concordance compared with mHKA (Fig. 3). In the present study, the JLO was 173.3° ± 3.5°, whereas the K-JLOA was −3.2° ± 2.7°. This indicates that the medial inclination of the joint line relative to the floor is relatively smaller than the medial inclination relative to the mechanical axis observed on LLRs. Sex-based analyses demonstrated no meaningful differences in K-JLOA on either side, suggesting that sex is unlikely to be a major confounder in the interpretation of knee joint line orientation in healthy adults.

Close

Fig. 3 Bilateral correlation of joint line obliquity (JLO), joint line orientation angle (K-JLOA), arithmetic hip–knee–ankle angle (aHKA), and mechanical hip–knee–ankle angle (mHKA) Scatter plots showing right–left correlations for JLO (A) and K-JLOA (B) in healthy adults. Both JLO and K-JLOA demonstrated weak bilateral correlations (JLO: R² = 0.144; K-JLOA: R² = 0.132). Scatter plots showing right–left correlations for aHKA (C) and mHKA (D). aHKA demonstrated a modest bilateral correlation (R² = 0.208), whereas mHKA showed a stronger correlation (R² = 0.581).

Export to PPT

In addition, differences in K-JLOA among CPAK types were minimal (Table 3). Although statistically significant differences in K-JLOA were observed across CPAK types, the absolute magnitude of these differences was small, and were therefore considered minimal in terms of clinical relevance. In a previous study based on three-dimensional analysis using upright CT, no significant differences in the 3D joint surface–floor angle were identified among the CPAK types,¹⁶ consistent with the results of the present study. Although the CPAK classification separates phenotypes by the relative sum and difference of MPTA and LDFA with respect to the mechanical axis, the actual K-JLOA relative to the floor showed only modest differences among types. As expected, CPAK types I–III, IV–VI, and VII–IX differed in K-JLOA, reflecting their definitional basis. Interestingly, type I demonstrated K-JLOA values comparable to types V and VI, and the neutral JLO group (types IV–VI) still exhibited medially inclined K-JLOA relative to the floor. These discrepancies suggest that CPAK classification derived from mechanical axis parameters does not necessarily correspond to functional joint line orientation relative to the floor.

Previous reports on both HTO and TKA have associated excessive K-JLOA with adverse biomechanical and clinical outcomes ^4,17–19, although several studies found no significant effect on clinical results ^{13–15,20,21}. Such inconsistency may be attributed not only to variations in measurement techniques but also to adjacent joint compensation and the inherently dynamic nature of joint surfaces during functional activity.^16,22,23 A study evaluating KA-TKA demonstrated that while the postoperative MPTA showed a wide and slightly varus-skewed distribution, the K-JLOA had a narrower, more normally distributed pattern centered near 1.0° varus.⁹ In a retrospective study following MA-TKA, the A-JLOA was defined as the angle between the tangent to the subchondral plate of the talus and the floor, with lateral inclination considered negative. The postoperative A-JLOA was −3.7° ± 4.8°.²⁴ In another study after HTO, the A-JLOA and talar inclination of the non-involved limb were 7.3° and 6.8°, respectively, indicating a laterally tilted orientation of the ankle joint line.²⁵ Unlike these reports, the present study found the A-JLOA averaged 0.6°, nearly parallel to the floor, but with nearly twice the standard deviation compared to the K-JLOA (Table 1). This suggests that the ankle joint exhibits greater variability in parallelism than the knee joint.

In the present study, several radiographic parameters were found to influence K-JLOA. As expected, both MPTA and LDFA affected not only JLO relative to the mechanical axis but also K-JLOA relative to the floor. Parameters such as K- and A-JLCA likely reflect the influence of periarticular soft tissue laxity, whereas A-IMD represents the effect of foot position. Therefore, K-JLOA is likely influenced by periarticular soft tissue condition, radiographic acquisition technique, changes in foot position after deformity correction, and dynamic alterations in foot positioning during activity. In two studies that defined lateral space opening as a positive value, the preoperative K-JLCA before HTO measured 8.5° ± 2.3°²⁶ and 3.4° ± 2.2°¹⁸ whereas in another study that defined it as negative, the preoperative K-JLCA was −3.6° ± 2.2°.²⁷ How this variable affects K-JLOA remains unclear.²⁷ In the present study of healthy adults, both K-JLCA and A-JLCA averaged less than 1° but showed a wide range of variation (Table 1). Even when LLRs were obtained with a standardized distance between the feet, differences in ankle circumference and coronal alignment produced variability between K-ICD and A-IMD (Table 1). In a study investigating radiographic parameters associated with changes in K-JLOA after HTO, the K-ICD decreased from 78.5 ± 22.7 mm preoperatively to 66.1 ± 20.8 mm postoperatively, whereas the A-IMD increased from 45.7 ± 39.1 mm to 60.9 ± 36.5 mm.²⁸ Postoperative changes in K-JLOA were more strongly correlated with K-ICD (r = 0.702) and A-IMD (r = 0.691) than with MPTA (r = −0.342),²⁸ suggesting that foot and knee positioning exert greater influence than bony correction alone. In particular, when standing LLRs are taken with the feet together, a reduced A-IMD can make the JLO appear visually lateralized. A retrospective study further demonstrated that K-JLOA after MA-TKA was strongly affected by foot distance during LLR acquisition, with the K-JLOA changing by 3.7° for every 10 cm increase in separation.²⁴ Therefore, when analyzing K-JLOA as an independent variable on LLRs, it is essential to account for associated radiographic and positional parameters.^24,28 Failure to adequately control for these factors may lead to misinterpretation of clinical outcomes.

In the final regression model of the present study, greater LDFA and MPTA were positively associated with increases in K-JLOA, indicating that a more varus-oriented distal femur and a more valgus-oriented proximal tibia tend to produce a laterally tilted joint line. Conversely, larger K-JLCA and A-JLCA values—both expressed as positive for lateral opening—were negatively associated with K-JLOA, suggesting that increased lateral joint space opening at the knee or ankle correlates with a more medially inclined joint line. A greater A-IMD was also negatively associated with K-JLOA, implying that a wider ankle distance is linked to medial tilt. Collectively, these findings highlight that osseous alignment parameters (LDFA, MPTA), soft tissue– or joint space–related parameters (K-JLCA, A-JLCA), and foot positioning parameters (A-IMD) independently contribute to the coronal orientation of the knee joint line. This underscores the importance of considering these confounding factors when evaluating and interpreting joint line orientation.

In terms of bilateral concordance, knee joint-related parameters such as MPTA, LDFA, aHKA, JLO, K-JLOA, and K-JLCA showed weaker correlations between sides compared with mHKA. In contrast, ankle joint-related parameters including A-JLOA, A-JLCA, and talar inclination demonstrated relatively stronger bilateral correlations. Taken together, the side-to-side differences observed in K-JLOA and other knee joint line–related parameters likely represent a combination of true constitutional asymmetry and methodological variability, including measurement error and compensatory alignment at adjacent joints.^29,30 This warrants caution when applying side-to-side comparisons in clinical practice and implies the challenges of implementing individualized alignment strategies in real-world settings.

Despite significant correlations among quantitative radiographic parameters, bilateral concordance of nominal CPAK phenotypes was poor. This discrepancy can be explained in part by classification boundary effects, whereby small inter-limb differences around cut-off thresholds result in discordant phenotype assignment. In addition, relatively weaker bilateral correlations of JLO and aHKA, which are key determinants of CPAK classification, further contribute to this poor concordance (Table 4). These findings indicate that while radiographic parameters provide reliable continuous measures, categorical classification into CPAK types may exaggerate side-to-side differences and should therefore be interpreted with caution.

This study has several limitations that should be acknowledged. This study characterized the variability of joint line–related parameters through radiographic analysis in a healthy adult population. However, it did not statistically evaluate the direct association between joint line orientation and clinical outcomes. This was intentional, as the primary aim was to establish a radiographic reference framework for future investigations rather than to draw direct clinical correlations. Second, the study population was derived from a single ethnic group, which may limit the generalizability of the findings to other populations.^31,32 Third, only static standing long-leg radiographs were analyzed. Although this method remains the clinical standard, it cannot capture dynamic joint line behavior during gait or loading. Additionally, variations in foot positioning, ankle morphology, and periarticular soft tissue laxity may not have been fully controlled despite standardized radiographic protocols, potentially introducing measurement variability. Fourth, no a priori sample size calculation or power analysis was performed because of the retrospective and exploratory design, which may limit the statistical power of the analyses. Furthermore, the relatively small sample size—particularly within certain CPAK subgroups—may have reduced the ability to detect subtle intergroup differences. Finally, biomechanical studies incorporating gait analysis, dynamic fluoroscopy, or three-dimensional upright imaging are warranted to clarify the functional relationship between true joint contact mechanics and radiographic joint line orientation under dynamic conditions.

5 Conclusion

In healthy adults, K-JLOA was slightly medially inclined and was influenced by tibial (MPTA) and femoral (LDFA) geometry, as well as joint space parameters (K-JLCA, A-JLCA) and ankle width (A-IMD). Although statistically significant differences in K-JLOA were observed across CPAK types, their magnitude was small. While mHKA demonstrated strong bilateral concordance, JLO and aHKA showed weaker agreement, which, together with classification boundary effects, contributed to poor CPAK phenotype concordance. These findings suggest that although K-JLOA and related parameters can be reliably measured, phenotype-based classification should be interpreted cautiously and supported by quantitative radiographic assessment when considering personalized alignment strategies.

Patient consent

Informed consent was waived due to the retrospective nature of the study and the use of anonymized data, as approved by the Institutional Review Board.

CRediT author statement

Sang Jun Song: Conceptualization, Writing – original draft, Supervision.

Jihoon Song: Data curation, Formal analysis.

Min Sung Kim: Data curation, Formal analysis.

Kang-il Kim: Investigation, Validation.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.