Longitudinal 3D Shape and Alignment Analysis of Post-traumatic Osteoarthritis After Tibial Plafond Fractures

Abstract

Background:

This study aimed to characterize longitudinal 3-dimensional morphologic, alignment, and radiographic changes of the ankle and hindfoot following surgical fixation of tibial plafond fractures and to identify structural features associated with post-traumatic osteoarthritis (PTOA) severity.

Methods:

Thirty-two patients with surgically treated pilon fractures from 3 institutions underwent weightbearing computed tomography (WBCT) at 6, 12, and 18 months post-operatively. At 18 months, patients were grouped into low-grade and high-grade PTOA groups. Statistical shape modeling was used to quantify morphologic and alignment variation across the tibia, fibula, talus, and calcaneus. Joint distances at the talocrural and subtalar joints were calculated using correspondence-based metrics. Radiographic parameters were automatically derived from 3-dimensional bone models using a MATLAB toolbox. Cross-sectional and longitudinal statistical differences were assessed.

Results:

Patients who developed high-grade PTOA demonstrated progressive flattening of the calcaneal pitch, posterior and inferior translation of the tibia and fibula, and narrowing of the central tibiotalar joint. In contrast, the low-grade group exhibited greater valgus hindfoot alignment, mild widening of the medial and posterior tibiotalar joint, and greater alignment variation over time. Radiographically, the high-grade group displayed a significant increase in tibiocalcaneal angle over time (P = .03) and a trend toward a more varus hindfoot configuration.

Conclusion:

Distinct longitudinal alignment and morphologic trajectories differentiate low-grade and high-grade PTOA development following pilon fracture fixation. Preservation of coronal alignment and subtalar adaptability may provide a protective mechanism against rapid PTOA progression. These findings highlight the clinical potential of WBCT and shape-based 3-dimensional analysis for early detection of malalignment patterns associated with post-traumatic degeneration.

Level of Evidence:

Level III, prognostic study.

Introduction

Post-traumatic osteoarthritis (PTOA) is a frequent and debilitating consequence of tibial plafond, or pilon, fractures, with reported rates approaching 40% within 2 years and exceeding 70% within 5 years after surgery.^1-4 These high-energy intra-articular injuries primarily affect younger, active individuals and often lead to early joint degeneration, pain, and loss of function. ⁵ Despite improvements in surgical fixation, effective strategies to prevent PTOA progression remain limited. ¹

Malalignment and abnormal joint loading are recognized contributors to PTOA following tibial plafond fractures. ⁶ Weightbearing computed tomography (WBCT) has enabled more precise assessment of joint alignment and joint space width than conventional radiographs, yet most studies remain cross-sectional and focused solely on the tibiotalar joint. ⁷ Longitudinal 3-dimensional (3D) analyses that capture the evolution of bone morphology and alignment, particularly across both the talocrural and subtalar joints, are lacking. Such investigations are needed to understand how mechanical adaptations differ between joints that rapidly develop PTOA and those that remain relatively preserved. ⁴

This study quantified cross-sectional and longitudinal morphology, alignment, and joint space changes in the ankle and hindfoot following surgical fixation of pilon fractures using WBCT-based statistical shape modeling (SSM), radiologic measurements, and quantitative joint metrics. We hypothesized that patients who develop severe (high-grade) PTOA within 18 months post-operative would demonstrate both morphologic and alignment differences, specifically a more cavovarus presentation, compared with patients with no or mild (low-grade) PTOA at the same time point. Identifying 3D structural markers may help inform early detection and guide post-operative management or surgical decision-making aimed at preserving ankle joint function.

Materials and Methods

Data Source and Study Population

This retrospective study was performed in accordance with the Declaration of Helsinki and under IRB approval (IRB 00139629). Thirty-two patients from 3 institutions (University of Utah, University of Iowa, and University of Washington) who sustained pilon fractures (Orthopaedic Trauma Association type 43B or 43C) ⁸ and underwent surgical fixation were included. Each of the 32 patients underwent a WBCT scan at 6 months (6M), 12 months (12M), and 18 months (18M) post-operatively, providing complete longitudinal data across all 3 time points. The 18M scans were then used to allocate the patients into high-grade tibiotalar PTOA (Kellgren-Lawrence scores of 3 or 4) and low-grade tibiotalar PTOA (Kellgren-Lawrence scores of 0, 1, or 2).^9,10 The primary grouping variable was 18M tibiotalar PTOA severity, whereas SSM, joint distance, and radiologic measurements were treated as complementary exploratory structural analyses used to characterize morphology, alignment, and joint distance patterns associated with PTOA severity. Exclusion criteria included hardware removal between time points or additional fractures beyond the tibia or fibula.

Image Acquisition and Processing

WBCT scans were acquired at each time point at one of 3 institutions: University of Utah (Curvebeam Pedcat; 0.37-mm voxels), University of Iowa (Curvebeam HiRise; 0.37-mm voxels), and University of Washington (Curvebeam Pedcat; 0.37-mm voxels). WBCT images were semi-automatically segmented (Bonelogic; DISIOR), manually verified, and consistently smoothed and remeshed (Mimics & 3matic; Materialise) to allow for consistent generation of 3D bone models of the distal tibia, distal fibula, talus, and calcaneus. ¹¹ Left feet were mirrored to right orientation, and the 4 bones were aligned as a combined multi-bone construct maintaining their native anatomical alignment using an iterative closest-point method. ¹² These aligned bone models were then the inputs for the SSM workflow, with group flags denoting time points and PTOA group.^13,14

Statistical Shape Modeling

A multi-bone SSM, a statistical model that captures average anatomy and anatomical variation across subjects, was generated in ShapeWorks v6.6.0 (Scientific Computing and Imaging Institute), an open-source software package. ¹⁵ The distal tibia, distal fibula, talus, and calcaneus were included together to preserve inter-osseous articular geometry. Generalized Procrustes analysis was applied in ShapeWorks to remove scale effects, allowing shape and alignment differences to be compared independent of overall bone size. ShapeWorks employs a particle-based modeling framework that distributes correspondence points across each bone surface through an energy optimization process. These particles act as anatomically consistent landmarks, each representing the same relative location across all subjects, allowing direct, one-to-one comparison of shape and alignment between individuals and groups. This approach captures global morphology while preserving local geometric detail. After optimization, mean shape models were generated for each group (low-grade and high-grade PTOA) at each time point (6M, 12M, 18M).

Joint Distance Analyses

Using correspondence particle locations, a validated MATLAB toolbox, Joint Measurement Analysis, ¹⁶ was employed to compute joint distances across the tibiotalar, tibiofibular, talofibular, and subtalar joints for each subject and time point. For each joint, the minimum 3D distance between opposing subchondral bone surfaces was calculated at each correspondence particle, serving as a surrogate for cartilage thickness and joint congruency. Particle-level joint distances were then compared both longitudinally (across time points within each OA group) and cross-sectionally (between the 2 OA groups at each time point). Statistical analyses were performed using an analysis of variance (ANOVA) or a repeated measures ANOVA for cross-sectional or longitudinal analysis, respectively, with post hoc t tests and false discovery rate (FDR) correction (α = .05).

Shape Model Analyses

Differences between mean shapes were calculated and visualized as surface distance maps, which are color maps showing where one mean bone model is more prominent than another (CloudCompare, cloudcompare.org). To determine regions of statistically significant shape or alignment variation, Hotelling T² tests, a multivariate statistical test used to compare corresponding 3D point locations between groups, were applied at each correspondence particle, followed by an FDR correction (α = .05).^17-20 Analyses were performed on both the 4-bone combined model and the individual bones. Significant particles were localized to identify anatomical regions showing morphology-only, alignment-only, or combined variation between 2 groups. Similar to the joint analyses, comparisons were conducted both cross-sectionally (between OA groups at each time point) and longitudinally (between time points within each OA group).

Radiologic Measurement Analyses

Radiologic measurements were calculated using an automated MATLAB toolbox that leverages a previously validated bone coordinate system from the Automatic Anatomical Foot and Ankle Coordinate Toolbox (AAFACT). ²¹ This workflow uses AAFACT-defined anatomical vectors, calculated directly from the segmented WBCT bone models, to reproduce standard clinical radiologic measurements on 3D bone models derived from the previously described WBCT reconstructions. The following angles were computed for each subject and time point: sagittal and axial talocalcaneal angles, calcaneal inclination angle, tibiotalar tilt angle, 0° and 20° hindfoot alignment angle, hindfoot moment arm, and sagittal and axial tibiocalcaneal angles. All measurements were statistically compared longitudinally (repeated measures ANOVA) and cross-sectionally (ANOVA) to determine significance (α = .05).

Results

Patient Characteristics

Thirty-two patients were split into 2 groups based on their OA scores at 18M, yielding 16 with low-grade PTOA and 16 with high-grade PTOA. Demographic variables are included in Table 1. Age, body mass index, sex distribution, and fracture classification did not differ significantly between groups.

Table 1.

Patient Characteristics.

Characteristics	Low-grade PTOA	High-grade PTOA	P Value
Ankles, n	16	16	–
Age, y, mean ± SD	46.6 ± 14.8	48.3 ± 14.1	.7433
Body mass index, mean ± SD	28.8 ± 4.3	29.6 ± 5.5	.6498
Sex, M/F	8/8	11/5	.2802
Fracture type, 43B/43C	4/12	4/12	1.0000

Abbreviation: PTOA, post-traumatic osteoarthritis.

Shape Model Analyses

Longitudinal variation: Low-grade PTOA

From 6M to 12M, scattered shape differences were observed across the calcaneus and talus without a consistent pattern (Figure 1). The fibula translated slightly posteriorly, but no other alignment changes. For 12M to 18M, there was no perceivable shape significance (Hotelling T², P < .05; Figure 1), but the calcaneus shifted more laterally and inferiorly, the talus shifted more laterally and superiorly, and the fibula shifted laterally. Between 6M and 18M, shape changes became a bit more apparent, with the talus becoming more anteriorly pronounced. Similar to the 12M to 18M comparison, the calcaneus shifted more laterally and inferiorly, the talus shifted more laterally and superiorly, and the fibula shifted laterally.

Figure 1.

Longitudinal statistical comparison between low-grade mean shapes of the multidomain statistical shape models. Pairwise comparisons are organized by column as 6-month (6M) vs 12-month (12M), 12M vs 18-month (18M), and 6M vs 18M. For each comparison, the left panels show Hotelling T² results and the right panels show surface distance maps. Hotelling T² tests identify particles that have significantly different shape (pink), alignment (blue), or a combination of both (purple) between each time point. Surface distance maps are plotted on the earlier time point listed in each column heading; red indicates regions where the earlier time point is more prominent, and blue indicates regions where the later time point is more prominent. Distances are reported in millimeters.

Longitudinal variation: High-grade PTOA

From 6M to 12M, only the fibula translated slightly posteriorly (Figure 2). For 12M to 18M, only the fibula translated slightly inferiorly. Between 6M and 18M, shape changes were similar to the low-grade group with the talus becoming more anteriorly pronounced. However, the calcaneal and talar pitch decreased and the tibia and fibula shift posteriorly and inferiorly.

Figure 2.

Longitudinal statistical comparison between high-grade mean shapes of the multidomain statistical shape models. Pairwise comparisons are organized by column as 6-month (6M) vs 12-month (12M), 12M vs 18-month (18M), and 6M vs 18M. For each comparison, the left panels show Hotelling T² results and the right panels show surface distance maps. Hotelling T² tests identify particles that have significantly different shape (pink), alignment (blue), or a combination of both (purple) between each time point. Surface distance maps are plotted on the earlier time point listed in each column heading; red indicates regions where the earlier time point is more prominent, and blue indicates regions where the later time point is more prominent. Distances are reported in millimeters.

Cross-sectional variation

At all 3 time points, the low-grade group demonstrated a more valgus hindfoot while the high-grade group had a flatter calcaneal pitch (Figure 3). Additionally, there were no other significant shape changes (Hotelling T², P < .05; Figure 3) in any of the single bone analyses, except for the region on the calcaneus bridging the anterior and medial subtalar facets, which remained consistently pronounced in the high-grade group.

Figure 3.

Cross-sectional statistical comparison between low-grade and high-grade mean shapes of the multi-bone statistical shape models. Pairwise comparisons at each post-operative time point included 6-month (6M), 12-month (12M), and 18-month (18M). Two analyses are shown: Hotelling T² tests and surface distances. Hotelling T² tests identify particles that have significantly (P < .05) different shape (pink), alignment (blue), or a combination of both (purple) between each group. Surface distance maps are plotted on the low-grade shape; red indicates regions where the low-grade shape is more prominent, and blue indicates regions where the high-grade shape is more prominent. Distances are reported in millimeters.

Joint Distance Analyses

Longitudinal variation: Low-grade PTOA

In the low-grade group, from 6M to 12M, the medial region of the tibiotalar joint widened over time, with no other patterns in the other joints (Figure 4). For 12M to 18M, the subtalar posterior facet and sinus tarsi widen over time, whereas the anterior facet narrows over time. Between 6M and 18M, the medial region of the tibiotalar joint and parts of the center of the joint widen over time and the subtalar posterior facet widens over time.

Figure 4.

Longitudinal joint distance statistical comparison within the low-grade group for the subtalar, tibiotalar, tibiofibular, and talofibular joints. Pairwise comparisons are organized by column as 6-month (6M) vs 12-month (12M), 12M vs 18-month (18M), and 6M vs 18M. Particles outlined in a pink ring highlight particles that are significantly different between time points. For each comparison, the top row displays the joint distance map plotted on the earlier time point listed in the column heading, whereas the bottom row displays the same comparison plotted on the later time point.

Longitudinal variation: High-grade PTOA

In the high-grade group, from 6M to 12M, the posterior region of the tibiotalar joint narrows over time with no other changes (Figure 5). For 12M to 18M, there were no patterns in any joint variation. Between 6M and 18M, the posterior region of the tibiotalar joint narrows over time, and the medial region of the tibiotalar joint widens over time. The talofibular joint also widens over time.

Figure 5.

Longitudinal joint distance statistical comparison within the high-grade group for the subtalar, tibiotalar, tibiofibular, and talofibular joints. Pairwise comparisons are organized by column as 6-month (6M) vs 12-month (12M), 12M vs 18-month (18M), and 6M vs 18M. Particles outlined in a pink ring highlight particles that are significantly different between time points. For each comparison, the top row displays the joint distance map plotted on the earlier time point listed in the column heading, while the bottom row displays the same comparison plotted on the later time point.

Cross-sectional variation

At 6M, in the subtalar joint, the low-grade group had a narrower sinus tarsi and medial facet, whereas the high-grade group had a narrower anterior-medial facet link (Figure 6). At 12M, there were similar, less pronounced, patterns in the subtalar joint compared to the 6M analysis. Additionally, the anterior tibiotalar joint was narrower in the high-grade group. At 18M, subtalar findings were consistent with earlier time points. The high-grade group also had a narrower central tibiotalar joint and a wider talofibular joint.

Figure 6.

Cross-sectional joint distance statistical comparison between low-grade and high-grade groups for the subtalar, tibiotalar, tibiofibular, and talofibular joints. Pairwise comparisons at each post-operative time point included 6-month (6M), 12-month (12M), and 18-month (18M). Particles outlined in a pink ring highlight particles that are significantly different between time points. For each comparison, the top row displays the joint distance map plotted on the low-grade shape, whereas the bottom row displays the same comparison plotted on the high-grade shape.

Radiologic Measurements

There were no significant changes for any of the radiologic measurements over time in the low-grade group (all P > .05; Table 2). The high-grade group had a significant increase in the tibiocalcaneal angle from 70.73° ± 6.19° at 6M to 73.03° ± 4.26° at 18M (P = .03), but no other significant findings (all other P > .05; Tables 2 and 3). For the cross-sectional analysis, at 6M, the low-grade group had a significantly smaller tibiotalar tilt angle (−3.77° ± 3.38° vs −0.60° ± 4.96°, P = .04) and trended toward a larger hindfoot alignment angle (9.69° ± 6.70° vs 5.12° ± 8.03°, P = .09). At 12M, the low-grade group was trending to have a smaller tibiotalar tilt angle and a larger calcaneal inclination angle. And at 18M, the low-grade group continued to trend toward a smaller tibiotalar tilt angle. Although most of the hindfoot measurements were not significantly different (P = .07-.09; Tables 2 and 3), those measurements did show the high-grade group was consistently more varus at all time points.

Table 2.

Radiographic Measurements for Each Group at Each Time Point. ^a

Measurement	Low-Grade 6M	Low-Grade 12M	Low-Grade 18M	High-Grade 6M	High-Grade 12M	High-Grade 18M
TCA (Sag.)	29.29° ± 5.99°	29.99° ± 5.46°	29.58° ± 5.37°	27.42° ± 4.25°	27.48° ± 4.49°	27.16° ± 5.30°
TCA (Ax.)	20.63° ± 5.98°	20.90° ± 6.08°	20.62° ± 5.92°	19.10° ± 5.19°	19.27° ± 5.95°	19.36° ± 5.55°
CIA	21.03° ± 6.30°	21.41° ± 5.58°	20.99° ± 5.91°	17.95° ± 4.79°	18.21° ± 4.48°	17.97° ± 5.02°
TTA	−3.77° ± 3.38°	−3.31° ± 3.48°	−3.20° ± 2.77°	−0.60° ± 4.96°	−0.47° ± 5.13°	−0.49° ± 5.06°
HAA (0°)	9.69° ± 6.70°	9.74° ± 5.95°	8.75° ± 6.12°	5.12° ± 8.03°	5.72° ± 8.53°	5.29° ± 8.92°
HAA (20°)	8.86° ± 6.04°	9.00° ± 5.48°	8.10° ± 5.62°	4.76° ± 7.50°	5.23° ± 7.93°	4.88° ± 8.25°
HMA (mm)	14.52 ± 8.33	14.99 ± 7.49	13.82 ± 7.81	10.72 ± 10.72	11.41 ± 11.48	10.58 ± 11.0
TiCA (Sag.)	72.09° ± 7.40°	71.39° ± 6.83°	71.26° ± 6.87°	70.73° ± 6.19°	72.00° ± 5.09°	73.03° ± 4.26°
TiCA (Ax.)	3.65° ± 7.39°	4.12° ± 6.65°	3.46° ± 6.63°	2.76° ± 6.69°	2.48° ± 6.43°	3.00° ± 7.49°

Abbreviations: Ax., axial; CIA, calcaneal inclination angle; HAA, hindfoot alignment angle; HMA, hindfoot moment arm; Sag., sagittal; TCA, talocalcaneal angle; TiCA, tibiocalcaneal angle; TTA, tibiotalar tilt angle.

^aMeasurements include mean ± SD. Bolded values indicate measurements that are significantly different, see Table 3 for related statistical values.

Table 3.

Statistical Comparisons. ^a

Measurement	Low-Grade	High-Grade	6M	12M	18M
TCA (Sag.)	.32	.81	.32	.17	.21
TCA (Ax.)	.51	.85	.44	.45	.54
CIA	.54	.76	.13	.08	.13
TTA	.30	.88	.04	.08	.07
HAA (0°)	.29	.50	.09	.13	.21
HAA (20°)	.32	.61	.10	.13	.21
HMA (mm)	.40	.49	.27	.31	.34
TiCA (Sag.)	.73	.03	.58	.78	.39
TiCA (Ax.)	.72	.80	.72	.48	.86

Abbreviations: Ax., axial; CIA, calcaneal inclination angle; HAA, hindfoot alignment angle; HMA, hindfoot moment arm; Sag., sagittal; TCA, talocalcaneal angle; TiCA, tibiocalcaneal angle; TTA, tibiotalar tilt angle.

^aP values from a repeated measures ANOVA for the longitudinal comparisons and from an ANOVA for the cross-sectional comparisons with α = .05. Bold values indicate significance.

Discussion

This study characterized longitudinal 3D morphologic, alignment, and radiologic adaptations of the ankle-hindfoot complex during the first 18M following surgical fixation of pilon fractures. Patients who developed high-grade PTOA demonstrated progressive flattening of the calcaneal pitch, posterior and inferior translation of the tibia and fibula, and tibiotalar joint narrowing, features indicative of restricted compensatory motion and increased articular loading. In contrast, patients with low-grade PTOA showed more valgus hindfoot alignment, subtle tibiotalar joint widening, and greater hindfoot adaptability over time. These findings partially supported our hypothesis, as the high-grade PTOA group demonstrated distinct alignment differences and a more varus hindfoot configuration, although the expected cavovarus pattern was not fully observed because of flattening of the calcaneal pitch.

Normal tibiotalar biomechanics depend on close articular congruity and balanced coronal alignment to distribute load across the talar dome. Prior cadaveric and computational studies have shown that small changes in talar position or ankle inversion/eversion can substantially alter tibiotalar contact area and pressure distribution.^22-24 In this context, the progressive tibiotalar narrowing, flatter calcaneal pitch, and more varus hindfoot configuration observed in the high-grade PTOA group may reflect a less favorable load-distribution environment. Clinically, restoring articular congruity, talar centering, and coronal plane alignment when planning fixation constructs maintain high importance. Although this study did not directly evaluate osteosynthesis strategy, these patterns suggest that durable restoration of the ankle mortise and avoidance of residual varus malalignment may be important for preserving post-operative load distribution.

The divergent trajectories observed between low-grade and high-grade PTOA groups align with prior evidence implicating joint malalignment and altered load distribution as key drivers of PTOA. ⁶ By combining multi-bone statistical shape modeling and WBCT-based quantitative joint analyses, this study reveals that mechanical adaptations occur not only at the tibiotalar joint but also across the subtalar complex, highlighting the mechanical coupling of these articulations in PTOA. ⁷ The consistently pronounced anterior-medial calcaneal facet in the high-grade group may represent an anatomic constraint that limits subtalar eversion, thereby reducing the hindfoot’s ability to dissipate impact forces and increasing stress transmission through the tibiotalar joint (Figure 7).

Figure 7.

The top row is an 18-month cross-sectional statistical comparison between low-grade and high-grade mean shapes of the single-bone calcaneal statistical shape model. Surface distances show how different, in millimeters, one model is to the other. The bottom row indicates representative patients in each group, highlighting the anterior-medial facet differences between groups.

Clinically, these findings underscore the importance of evaluating hindfoot alignment and joint mechanics following pilon fracture fixation. Traditional radiographic measures may underestimate 3D malalignment or compensatory adaptation that influence load transmission through the ankle complex. In this study, radiographic-style scalar measurements derived from WBCT 3D bone reconstructions detected few significant differences between groups, whereas 3D shape modeling and joint-based metrics revealed distinct patterns. ²⁵ This suggests that even when calculated from 3D bone models, scalar measurements may not fully capture the complexity of post-traumatic morphologic remodeling. WBCT combined with shape-based joint metrics could enable earlier detection of malalignment trajectories predictive of PTOA. ²⁶ Identifying patients who demonstrate limited subtalar adaptations or loss of valgus alignment within the first post-operative year may help guide targeted rehabilitation strategies, orthotic interventions, or closer imaging surveillance to mitigate progression toward end-stage degeneration.

This study has several limitations. The sample size, although larger than previous 3D longitudinal shape analyses, remains modest and may limit subgroup statistical power. Soft tissues were not modeled, which may influence articular contact and compensatory mechanistic hypotheses. Pre-injury foot morphology was not available in this cohort, as patients were first imaged after their injury. Therefore, subtle pre-existing deformities could not be directly accounted for. However, the observed differences between groups suggest that foot type may play an important role in recovery. Additionally, static WBCT captures only the loaded configuration at 1 point in time; dynamic imaging or in vivo kinematic analysis could provide further insight into functional adaptations. Future work should integrate dynamic functional assessments, contralateral limbs, and prospective follow-up beyond 18M to determine whether early 3D alignment changes predict long-term clinical outcomes.

Conclusion

Overall, this work identifies divergent morphologic and alignment trajectories associated with early PTOA development following pilon fracture fixation. Quantitative 3D metrics derived from WBCT and SSM offer new insight into how hindfoot alignment and subtalar adaptability influence post-traumatic joint health. These tools may ultimately enable predictive modeling of PTOA risk and guide personalized management strategies aimed at preserving ankle function.