Three-dimensional gait analysis following surgical correction of proximal femoral type IV fibrous dysplasia with varus deformity

Abstract

This study aimed to analyse postoperative gait characteristics in patients with fibrous dysplasia (FD) of the proximal femur with varus deformity and to comprehensively assess limb function. We retrospectively reviewed patients who underwent surgical correction between January 2013 and October 2017. Limb length discrepancy and mechanical axis alignment were evaluated at follow-up. Gait parameters of both lower limbs were assessed using a three-dimensional (3D) gait analysis system, and the affected side was compared with the contralateral limb. Limb alignment and length were largely restored, with residual discrepancies of ≤ 2 cm in most patients. Functional outcomes were excellent, with high Musculoskeletal Tumour Society (MSTS) and Harris Hip Score results. However, gait asymmetries persisted postoperatively. Compared with the contralateral limb, the affected limb demonstrated significantly lower gait speed (0.98 ± 0.13 vs. 1.15 ± 0.15 m/s) and cadence (99.07 ± 5.97 vs. 110.32 ± 7.21 steps/min). Single-limb support time was shorter (33.47 ± 2.63 vs. 39.59 ± 3.69%) and double-limb support time was prolonged (16.20 ± 2.67 vs. 11.85 ± 3.49%). Both peaks of vertical ground reaction force were lower on the affected side. Hip extension and total hip range of motion were significantly reduced on the affected side, whereas knee and ankle joint parameters showed no significant differences between limbs. Despite correction of deformity and satisfactory clinical scores, residual gait abnormalities persisted after surgery. Three-dimensional gait analysis provides an objective means of functional evaluation and can complement traditional functional assessment tools in guiding targeted postoperative rehabilitation.

Introduction

Fibrous dysplasia (FD), also known as osteodysplasia fibrosa, is a non-hereditary, tumour-like bone lesion in which normal medullary bone is progressively replaced by fibrous tissue.1 The proximal femur is among the most commonly affected sites, where the resulting structural compromise reduces mechanical strength and frequently leads to coxa vara or femoral varus deformity. Clinically, FD of the proximal femur typically presents with limb length discrepancy (LLD), limping, and varying degrees of joint pain [1, 2]. Long-term sequelae include an elevated risk of hip osteoarthritis [3]. Corrective osteotomy remains the standard surgical treatment for deformity correction, limb length restoration, and functional reconstruction, with generally favourable outcomes reported in the literature [4, 5]. However, long-term follow-up reveals that, despite restoration of mechanical alignment and leg length, gait improvement remains suboptimal in some patients.

Conventional postoperative outcome measures including the Musculoskeletal Tumour Society (MSTS) score and Harris Hip Score primarily reflect pain, range of motion, and patient-reported function, but do not characterise the temporal, spatial, or kinetic features of walking. Patients may achieve high scores while adopting compensatory gait strategies that mask underlying biomechanical deficits, highlighting the need for objective dynamic functional assessment. Bipedal locomotion is a defining characteristic of human physical function, and gait analysis provides a systematic, objective measure of lower limb performance during walking [6].

Leg length discrepancy a hallmark of untreated or incompletely corrected proximal femoral FD is known to produce a range of compensatory gait adaptations, including pelvic obliquity, altered knee flexion in terminal stance, and asymmetric loading patterns [7, 8]. Even after surgical correction, residual or functional LLD may persist and contribute to ongoing gait asymmetry. Three-dimensional (3D) gait analysis provides a comprehensive, objective means of quantifying these biomechanical deviations, complementing conventional assessment, and guiding individualised rehabilitation. Using 3D gait analysis, this study aimed to characterise postoperative gait in patients who had undergone corrective osteotomy for proximal femoral FD with varus deformity, and to identify residual functional deficits that may inform rehabilitation strategies.

Materials and methods

Patient population

We retrospectively included consecutive patients who underwent proximal femoral osteotomy for FD with varus deformity between January 2013 and October 2017 at West China Hospital, Sichuan University. Patients were eligible for inclusion if they were between 10 and 35 years of age, regardless of sex, and had a diagnosis of unilateral proximal femoral FD with varus deformity, including either coxa vara or femoral varus, without involvement of the contralateral femur. Patients were additionally required to have undergone surgical correction at least one year prior to enrolment.

Patients were excluded if they had congenital or benign femoral deformities unrelated to FD, congenital pelvic or spinal deformities, or severe postoperative complications such as internal fixation failure or osteotomy site infection. The study was approved by the Ethics Committee of West China Hospital, Sichuan University. Written informed consent was obtained from all participants.

Instrumentation for gait analysis

Three-dimensional gait analysis was conducted using a motion capture system comprising ten infrared cameras (Oqus 300, Qualisys, Sweden) and two force plates (80 × 40 × 10 cm; Bertec, USA), each equipped with strain gauge-based load cells for measurement of ground reaction forces. Data acquisition and processing were performed using Qualisys Track Manager (QTM) and Visual3D (C-Motion, USA) for motion modelling, trajectory tracking, parameter calculation, and report generation (Fig. 1).

Fig. 1.

Gait Analysis Laboratory Setup and Data Processing Software Output. Panel A shows the three-dimensional skeletal model generated by Visual3D (C-Motion, USA), illustrating real-time marker trajectory reconstruction during a gait trial. Panel B displays the full suite of biomechanical output waveforms available from the Visual3D system, including hip, knee, and ankle joint torques in the coronal and sagittal planes, and lateral, longitudinal, and vertical ground reaction force (GRF) curves across the gait cycle. The present study focused on vertical GRF and sagittal plane joint kinematics as primary outcome parameters; the additional waveforms shown in Panel B illustrate the full analytical capability of the system and represent targets for future investigation. Data acquisition was performed using Qualisys Track Manager (QTM; Qualisys, Sweden) synchronised with two Bertec force plates and a ten-camera infrared motion capture system (Oqus 300, Qualisys, Sweden)

Data collection protocol

Anthropometric measurements

Standard physical measurements, including body height, body weight, knee width, ankle width, and foot width, were obtained manually. Radiographic evaluation was used to measure the mechanical axis, limb length, and femoral collodiaphyseal angle (CDA).

Three-dimensional gait analysis

Gait analysis was performed in the 3D Gait Laboratory using a ten-camera infrared motion capture system (Oqus 300, Qualisys, Sweden) synchronised with two force plates (80 × 40 × 10 cm; Bertec, USA), each equipped with strain gauge-based load cells. Before testing, the laboratory environment was cleared of reflective surfaces, and system calibration was performed using dynamic control rods and horizontal alignment tools. Reflective markers were attached following a modified Helen Hayes protocol, including 12 dynamic markers placed on the metatarsal heads, calcaneal tuberosities, anterior and posterior superior iliac spines, and the thigh and shank segments, and 10 static markers positioned on the greater trochanter, femoral condyles, and malleoli bilaterally. Static data were recorded with participants standing barefoot on the force plates in an upright posture with feet shoulder-width apart and arms relaxed (Fig. 2). After removing the static markers, dynamic trials were conducted as participants walked at a self-selected speed along a 6-metre walkway; data were considered valid when full foot contact occurred on a force plate. Each participant completed four valid walking trials, and the mean values across trials were used for analysis.

Fig. 2.

Marker placement for static calibration on the lower limbs and pelvis. Positions of the 12 dynamic and 10 static reflective markers according to the modified Helen Hayes protocol, shown on the anterior and posterior aspects

All trials were time-normalised to 101 data points representing 0–100% of the gait cycle. Ground reaction force data were normalised to body weight (BW) to facilitate inter-subject and inter-limb comparison. Spatiotemporal parameters including walking speed and cadence were calculated on a per-stride basis for each limb independently, using the force plate contact data for each limb’s successive ground contacts. This per-limb approach is a standard method in clinical 3D gait analysis and enables detection of between-limb asymmetries in stride timing and speed. Marker trajectories and ground reaction forces were processed using Qualisys Track Manager and Visual3D (C-Motion, USA) to calculate the following parameters: spatiotemporal variables (gait speed, cadence, step length, stride length, single-limb support time, double-limb support time); kinematic variables (sagittal plane hip, knee, and ankle joint angles); and vertical ground reaction force (vGRF). Vertical GRF was selected as the primary kinetic outcome because it is the most robust and clinically interpretable kinetic measure for characterising asymmetric loading in small retrospective samples. Anterior-posterior and mediolateral GRF components, as well as joint moments, are acknowledged as important mechanistic measures and are identified as priorities for future prospective investigation. All data collection and processing were performed by an experienced gait analysis engineer.

Statistical analysis

Given the small sample size (n = 20) and the paired within-subject study design, the Wilcoxon signed-rank test was used for all comparisons between the affected and contralateral limbs. A p-value of less than 0.05 was considered statistically significant. Effect sizes were calculated using Cohen’s d, derived from group means and pooled standard deviations, with d values interpreted as small (< 0.5), medium (0.5–0.8), or large (> 0.8). All statistical analyses were performed using SPSS software (IBM, USA).

Isokinetic muscle strength assessment

Isokinetic muscle strength testing was performed on all 20 patients using the ISOMED2000 isokinetic dynamometer (D&R Ferstl GmbH, Germany). Peak torque values were measured bilaterally for hip abduction, adduction, flexion, and extension. Testing was conducted with patients in the supine position for flexion and extension, and in the lateral decubitus position for abduction and adduction. A standardised angular velocity of 60°/sec was applied in active mode. Each patient performed four sets of ten repetitions per movement direction per side, with the healthy side tested first following familiarisation practice. The maximum peak torque value across the four sets was recorded for each direction and side. Results are reported as absolute peak torque (N·m) for affected versus healthy side comparison.

Results

Demographic characteristics and follow-up outcomes

Demographic characteristics are detailed in (Table 1). All 20 patients completed the follow-up, which ranged from 13 to 65 months (mean: 35.6 months). Postoperative limb length discrepancy was ≤ 1 cm in 15 patients and 1–2 cm in the remaining five. Mechanical axis was satisfactory in 13 patients; seven showed deviations of 3–5 mm from the knee joint centre. The postoperative collodiaphyseal angle (CDA) improved significantly from 92.5 ± 3.14° preoperatively to 128.6 ± 2.56° postoperatively, indicating substantial deformity correction.

Table 1.

Demographic characteristics

Indicator	Value / Range
Sample size	20
Sex (M/F)	11/9
Age (years)	10–35 (20.1 ± 4.6)
BMI (kg/m²)	17.2–29.5 (23.5 ± 1.6)
Affected side (L/R)	10/10
Time since surgery (months)	13–65 (35.6 ± 5.2)

Spatiotemporal parameters

The affected limb demonstrated significantly reduced gait speed and cadence compared with the contralateral limb (Table 2). Effect sizes for these differences were large (Cohen’s d = 1.21 and 1.70, respectively). Step length and stride length did not differ significantly between sides. Single-limb support time (SLST) was significantly shorter on the affected side, and double-limb support time (DLST) was significantly prolonged, both with large effect sizes (d = 1.91 and 1.40, respectively), indicating adoption of a more guarded gait pattern.

Table 2.

Spatiotemporal gait parameters

Parameter	Affected side	Healthy side	P	Cohen’s d	Effect
Step length (m)	0.56 ± 0.14	0.59 ± 0.09	0.621	0.26	Small
Stride length (m)	1.12 ± 0.11	1.18 ± 0.23	0.473	0.33	Small
Cadence (steps/min)*	99.07 ± 5.97	110.32 ± 7.21	0.033*	1.70	Large
Gait speed (m/s)*	0.98 ± 0.13	1.15 ± 0.15	0.046*	1.21	Large
DLST (GC%)*	16.20 ± 2.67	11.85 ± 3.49	0.027*	1.40	Large
SLST (GC%)*	33.47 ± 2.63	39.59 ± 3.69	0.041*	1.91	Large

Data are mean ± SD. DLST  Double-limb support time, SLST  Single-limb support time, GC Gait cycle. *P < 0.05.

Kinetic parameters

Both the first and second peaks of vertical ground reaction force were significantly lower on the affected side (Table 3). Effect sizes were medium for the first peak (d = 0.63) and large for the second peak (d = 0.87). The valley (minimum vGRF) did not differ significantly between limbs, suggesting relatively preserved buffering capacity during mid-stance.

Table 3.

Vertical ground reaction force (vGRF, normalised to body weight [BW])

Gait Cycle Phase	Affected Side (BW)	Healthy Side (BW)	P	Cohen’s d	Effect
1st peak*	0.98 ± 0.23	1.19 ± 0.41	0.024*	0.63	Medium
Valley (mid-stance)	0.75 ± 0.11	0.72 ± 0.07	0.418	0.33	Small
2nd peak*	1.06 ± 0.23	1.24 ± 0.18	0.032*	0.87	Large

Data are mean ± SD. *P < 0.05.

Kinematic parameters

Hip extension was significantly reduced on the affected side: the maximum hip extension angle was − 8.1 ± 1.5° compared with − 12.9 ± 2.3° on the contralateral side (P = 0.023; d = 2.47, very large effect). Total hip sagittal range of motion was also significantly lower on the affected side (33.9 ± 5.2° vs. 40.1 ± 4.7°; P = 0.038; d = 1.25). Knee and ankle joint parameters showed no significant differences between limbs (Table 4).

Table 4.

Sagittal plane ranges of motion of hip, knee, and ankle joints

Joint	Parameter	Affected side (°)	Healthy side (°)	P	Cohen’s d
Hip	Max flexion angle	25.8 ± 5.4	27.7 ± 6.1	0.245	0.33
	Max extension angle*	−8.1 ± 1.5	−12.9 ± 2.3	0.023*	2.47
	ROM*	33.9 ± 5.2	40.1 ± 4.7	0.038*	1.25
Knee	Max flexion angle	59.1 ± 6.5	58.6 ± 6.9	0.584	0.07
	Max extension angle	−1.8 ± 4.8	−1.6 ± 4.4	0.157	0.04
	ROM	61.0 ± 5.5	59.7 ± 6.3	0.263	0.22
Ankle	Max flexion angle	−16.0 ± 6.0	−15.5 ± 6.9	0.152	0.08
	Max extension angle	13.2 ± 4.3	12.9 ± 3.9	0.486	0.07
	ROM	29.3 ± 5.0	28.5 ± 6.1	0.571	0.14

Data are mean ± SD (degrees). ROM  Range of motion.*P < 0.05

Isokinetic muscle strength

Peak torque values on the affected side were significantly lower than the healthy side for hip abduction, adduction, and flexion, with abduction showing the greatest deficit. Hip extension peak torque did not differ significantly between sides (Table 5).

Table 5.

Hip joint peak torque— affected versus healthy side

Direction	Affected side (N·m)	Healthy side (N·m)	P
Abduction	29.2 ± 3.7	48.1 ± 2.3	0.013
Adduction	39.3 ± 5.8	61.2 ± 5.4	0.026
Flexion	40.3 ± 2.8	58.2 ± 3.2	0.032
Extension	68.5 ± 3.5	70.8 ± 4.2	0.231

Data are mean ± SD

Discussion

Gait speed and cadence on the affected limb were significantly reduced after corrective osteotomy, while step length and stride length did not differ significantly from the contralateral limb. This pattern is consistent with findings in other populations with hip pathology: Bennett et al. reported that patients ten years after hip replacement exhibited persistently reduced gait speed and shorter step lengths at long-term follow-up [9]. Walking speed is a well-recognised composite measure of gait function and is strongly associated with fall risk, particularly in populations with musculoskeletal impairment [10, 11]. A gait speed threshold of approximately 1.0 m/s has been proposed as a clinically meaningful benchmark for fall risk stratification,11 and the mean affected-side walking speed of 0.98 m/s in our cohort falls below this threshold, suggesting that residual functional vulnerability may persist despite surgical success.

Given that all participants had satisfactory functional scores and were more than a year post-surgery, acute pain is unlikely to fully account for the altered gait. The reduced walking speed and cadence may instead reflect a combination of residual neuromuscular deficits, incomplete motor relearning, and a cautious gait strategy adopted to manage perceived instability.

The affected limb exhibited shorter SLST and prolonged DLST, a pattern consistent with reduced single-limb stability and compensatory weight-sharing strategies [12]. Winter et al. demonstrated that slower speed and prolonged DLST improve dynamic balance in individuals with reduced postural stability [13]. Our findings are consistent with this compensatory mechanism: patients appear to sacrifice temporal efficiency to enhance walking stability, even when gross spatial parameters such as step and stride length remain relatively preserved.

The first and second peak vGRF were both significantly lower on the affected side, consistent with findings in hip pathology literature where reduced peak forces reflect diminished loading capacity during stance [14]. The reduced vGRF peaks are consistent with impaired force generation during the propulsive and weight-acceptance phases, which may relate to altered hip muscle function a hypothesis requiring confirmation through objective strength testing. The preserved mid-stance vGRF valley suggests that buffering capacity during single-limb support was relatively maintained, which may reflect the more conservative joint changes associated with corrective osteotomy compared with arthroplasty, where capsular and ligamentous disruption may alter loading patterns more profoundly [15, 16].

Hip extension and total hip sagittal ROM were significantly reduced on the affected side, with very large effect sizes, whereas knee and ankle kinematics were comparable between limbs. This pattern is consistent with studies of hip pathology showing reduced hip extension during late stance [17]. The preservation of step and stride lengths despite reduced hip extension likely reflects compensatory strategies involving increased contribution from the knee and ankle on the affected side and greater reliance on the contralateral limb, enabling spatial gait metrics to remain within normal limits even when joint-level motion is restricted.

Restoration of mechanical alignment is a prerequisite for, but not a guarantee of, functional recovery. Our findings demonstrate that even patients with satisfactory radiographic correction and good clinical scores may exhibit persistent gait asymmetries. Postoperative rehabilitation should therefore extend beyond range-of-motion restoration and pain management to include regular gait assessments during follow-up and structured gait training targeting temporal symmetry and limb loading.

Isokinetic muscle strength testing confirmed objective deficits in hip abductor, adductor, and flexor strength on the affected side, with abduction showing the greatest reduction — approximately 39% lower than the contralateral limb. Hip extension strength was the only direction that did not differ significantly between sides. These findings provide objective confirmation for the gait patterns observed in this study. Reduced hip abductor strength directly impairs single-limb stability during the stance phase, consistent with the shortened SLST and prolonged DLST observed in our cohort. Reduced hip flexor strength contributes to diminished propulsive force generation, which is consistent with the lower second vGRF peak and reduced cadence on the affected side. The relatively preserved hip extension strength may partly explain why step and stride lengths remained comparable between limbs despite the kinematic restriction in hip extension angle — extension torque capacity was sufficient to maintain spatial parameters even when the angular range was limited. The severity of the abductor deficit in particular underscores the need for targeted hip abductor strengthening as a core component of postoperative rehabilitation in this population.

Clinical implications

Our findings highlight that even after successful corrective osteotomy, patients may exhibit subtle but clinically meaningful gait abnormalities including reduced walking speed, cadence, hip extension, and ground reaction forces despite comparable step and stride lengths between limbs. These alterations suggest residual biomechanical deficits and compensatory loading patterns that may increase fall risk and contralateral joint overload over time. Postoperative rehabilitation should therefore emphasise hip-specific strengthening, dynamic balance training, and gait retraining alongside standard mobility restoration. Regular gait assessments are recommended to monitor long-term recovery and guide individualised rehabilitation strategies.

Limitations

Several limitations of this study should be acknowledged. First, the absence of a healthy normative control group limits the interpretation of between-limb comparisons, particularly given that leg length discrepancy may itself induce compensatory gait adaptations on the contralateral limb, as documented in recent literature.

Second, isokinetic muscle strength testing was performed as part of this study and confirmed significant deficits in hip abductor, adductor, and flexor strength on the affected side. However, preoperative strength measurements were not available, precluding assessment of the dynamic change in muscle function following surgery. Future studies should incorporate pre- and postoperative isokinetic assessment to characterise the trajectory of muscle strength recovery.

Third, individual leg length discrepancy values at the exact time of gait analysis were not prospectively recorded; reported values are derived from the most recent clinical follow-up records prior to gait assessment.

Fourth, individual patient MSTS and Harris Hip Score data were not available in a format suitable for correlation analysis with gait parameters, precluding examination of the relationship between clinical scores and functional gait metrics.

Fifth, the per-limb calculation of spatiotemporal parameters reflects asymmetric stride timing and is a recognised approach in gait analysis; however, trial-level walking speed would provide an additional complementary measure.

Finally, mean ± standard deviation waveform curves for joint kinematics and kinetics across the gait cycle were not generated for this retrospective analysis; future prospective studies should incorporate waveform-level visualisation and statistical parametric mapping to enable more comprehensive between-limb comparison.

Future directions

Future prospective studies in this population should incorporate: (1) age- and sex-matched healthy control participants for normative comparison; (2) prospective correlation analysis between clinical outcome scores and gait parameters; (3) joint moment analysis and AP/ML ground reaction force assessment to provide mechanistic insight into loading asymmetries; and (4) statistical parametric mapping for waveform-level comparison of kinematic and kinetic data between limbs.

Conclusions

Traditional measures of alignment and limb length cannot fully capture functional recovery following corrective surgery for proximal femoral FD. In this cohort, postoperative gait was characterised by slower speed, reduced cadence, prolonged double-limb support, shorter single-limb support, and significantly reduced hip extension all with large effect sizes despite satisfactory radiographic correction and good clinical scores. These findings underscore that restoration of mechanical alignment is a prerequisite for functional recovery but does not guarantee gait normalisation. Targeted postoperative rehabilitation incorporating regular gait assessment during follow-up and structured gait training is essential to address residual biomechanical deficits and reduce long-term fall and overload risk. Three-dimensional gait analysis provides an objective and responsive complement to traditional outcome measures in this population.

Author contributions

Saif Abdulraqeb: Study design, statistical analysis, writing—review & editing. Zhigang Lang: Study design, writing—review & editing. Fan Wu: Study design, writing—review & editing. Yan Xiong: Study design, writing—review & editing. Wenli Zhang: Study design, writing—review & editing. Duan Hong: Study design, Supervision, reviewed and revised the manuscript.

Funding

This study was not supported by any funding.

Data availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of West China Hospital, Sichuan University and was conducted in accordance with the Declaration of Helsinki.

consent for publication

Written informed consent for publication was obtained from all patients.

Competing interests

The authors declare no competing interests.

Written informed consent to participate was obtained from all patients.