Axial rotation predicts coronal correction in vertebral body tethering: a retrospective three-dimensional study in Hong Kong

Article information

Asian Spine J. 2026;.asj.2025.0413

Publication date (electronic) : 2026 January 12

doi : https://doi.org/10.31616/asj.2025.0413

Kai Chun Augustine Chan, Anjaly Saseendran, Kenny Yat Hong Kwan

Department of Orthopaedics and Traumatology, School of Clinical Medicine, The University of Hong Kong, Hong Kong SAR, China

Corresponding author: Kenny Yat Hong Kwan, Department of Orthopaedics and Traumatology, The University of Hong Kong, 5/F Professorial Block, Queen Mary Hospital, 102 Pokfulam Road, Pokfulam, Hong Kong SAR, China, Tel: +852-2255-4341, Fax: +852-2255-4341, E-mail: kyhkwan@hku.hk

Received 2025 July 11; Revised 2025 September 15; Accepted 2025 October 6.

Abstract

Introduction

Traditionally, adolescent idiopathic scoliosis (AIS) was managed either by bracing or surgery, most commonly posterior spinal fusion (PSF). Although PSF was well-proven to produce favorable long-term results, drawbacks of PSF include compromising spinal growth and decreasing spinal mobility [1]. Optimizing spinal growth is crucial to the development of lung volume and pulmonary function [2], whereas unfused vertebral segments were subjected to greater mechanical stress, predisposing to spinopelvic pain long-term [3]. The newest surgical approach is vertebral body tethering (VBT), a fusionless surgical technique.

VBT utilizes intrinsic spinal growth as a corrective mechanism [1,4,5], with immediate correction achieved with compression screws and further curve correction achieved by growth modulation [5]. Previous literature has mainly investigated on coronal plane correction in VBT patients, with varying degrees of success [5–10]. Surgical outcomes following VBT could be unpredictable and subject to complications, such as tether breakage [11,12]. Deformity in the axial plane was often neglected, and clinicians still lacked knowledge on the role of axial-plane parameters and how they might influence postoperative outcomes [13–16].

Previous studies often characterized radiographic outcomes with regard to plain radiographs but lacked accuracy in assessing the axial plane. Utilizing three-dimensional (3D) reconstruction models tailored to individual patients, our study aims to investigate the changes in spinal axial, coronal, and sagittal profiles after VBT. Furthermore, we (1) investigated the relationship between preoperative axial parameters and postoperative coronal and sagittal outcomes, (2) determined the effect of intraoperative axial derotation on deformity correction, and (3) assessed the presence of VBT-specific complications, such as tether breakage and overcorrection. Ultimately, we sought to determine if preoperative rotation severity plays a prognostic role in postoperative success and provide insights into the biomechanics of VBT.

Materials and Methods

Study design and ethics approval

This is a retrospective study involving consecutive patients receiving VBT in Queen Mary Hospital and Duchess of Kent Children’s Hospital, Hong Kong from February 2019 to April 2024. Ethical approval from the Institutional Review Board of HKU/HA HKW (the University of Hong Kong and Hospital Authority Hong Kong West Cluster) was obtained before study commencement (UW 19–002). Informed consent from individual patients was omitted due to retrospective nature of study.

In our institution, VBT is indicated in patients with skeletal immaturity (Risser ≤3 or Sanders stage ≤6) [17], progressive curves of 45°–80° and flexible curves (fulcrum flexibility ≥50%) [18]. We included only patients with at least 1-year follow-up and operated for AIS. The exclusion criteria were (1) conversion to spinal fusion within 1 year and (2) nonidiopathic scoliosis. Biplanar radiographic images were obtained using a low-dose X-ray device EOS (EOS imaging, Paris, France) in both posteroanterior and lateral standing views. EOS images were subsequently used to create 3D spinal reconstruction for individual patients with the sterEOS software (EOS imaging). Two investigators (authors 1 and 2) blinded to clinical information were trained to perform all image reconstructions independently. The time points for curve analysis were preoperatively, immediately postoperatively, 1 year postoperatively, and 2 years postoperatively (Consolidated Standards of Reporting Trials [CONSORT] study flowchart) (Fig. 1). Fig. 2 shows an example of a 3D reconstructed image of one of our patients.

Fig. 1

CONSORT (Consolidated Standards of Reporting Trials) flowchart for vertebral body tethering (VBT) patient identification, inclusion, and final analysis. PSF, posterior spinal fusion; 3D, three-dimensional; Preop, preaoperative; Postop, postoperative.

Fig. 2

Three-dimensional reconstruction of one of our patients at 2 years postoperatively in posteroanterior view, lateral view, and overhead view.

Study outcomes

Primary outcomes were changes in coronal, axial, and sagittal profiles from preoperatively to immediately postoperatively, 1 year and 2 years postoperatively. Coronal deformity was defined by the major Cobb angle, and correction was assessed by the magnitude of curve regression and correction rate (%). Curve regression was defined as the absolute difference between post- and preoperative curve magnitude, whereas correction rate was further defined as (preoperative–postoperative curve magnitude)/preoperative curve magnitude×100%.

Axial deformity was assessed by apical vertebral rotation (AVR), maximal vertebral rotation (MVR), and torsion index. AVR and MVR were two different entities, thus were assessed separately [19]. The curve apex was automatically identified using the SterEOS software. The precise degree of vertebral rotation was assessed from the overhead view using SterEOS and defined as the angle between the patient’s sagittal plane and each vertebra’s axis of symmetry (Fig. 3). Clockwise/counterclockwise rotations were denoted as positive/negative values by using the software. We used absolute values for axial parameters to indicate rotation severity. Torsion index was defined as the mean of the two sums of intervertebral axial rotations from the lower junction to the curve apex, and from the apex to the upper junction [14,15]. MVR and AVR denote rotation severity, whereas the torsion index describes the net twisting of the whole curve toward one side. Sagittal parameters assessed included T1–T12 kyphosis and L1–L5 lordosis for thoracic and thoracolumbar/lumbar curves, respectively. A 3D analysis was performed for each instrumented curve, and both curve segments were investigated for a double VBT (presence of thoracic and thoracolumbar/lumbar tethers).

Fig. 3

Diagram illustrating calculation of vertebral rotation (0) from overhead view, defined by patient’s sagittal axis and individual vertebrae’s axis of symmetry.

The secondary outcome was the relationship between axial parameters (AVR, MVR, and torsion index) with short-to-medium term changes in coronal/sagittal profiles (up to 2 years). Axial derotation was defined as the difference in rotation preoperatively to immediately postoperatively, indicating intraoperative correction. Preoperative axial parameters and derotation were correlated with curve regression, correction rate, and changes in kyphosis/lordosis at 1- and 2-year follow-up, respectively. We performed subgroup analysis for thoracic and thoracolumbar/lumbar VBTs. The presence of long-term complications, including tether breakage, overcorrection, and revision procedures, was documented at the final follow-up.

Results

Patient demographics

We included 44 patients (seven males, 37 females) with 58 instrumented curves. We excluded six patients due to inadequate follow-up (n=1), no preoperative EOS scan (n=3), and fusion within 1 year (n=2). The average follow-up for patients was 36.3±17.1 months (range, 12–74 months), with 36 curves with EOS radiographs at the 2-year follow-up for analysis. Seventeen females (45.9%) had menarche at the time of VBT. Preoperatively, all patients were skeletally immature, and the mean curve flexibility was 77.8%±15.4%. The mean pre- and final postoperative Sanders stages were 3.0±0.9 (range, 1–5) and 6.8±0.7 (range, 5–8), respectively, with a mean stage progression of 3.9±1.0. Regarding Risser staging, the pre- and postoperative stages were 0.8±1.1 (range, 0–3) and 4.5±0.8 (range, 3–5), respectively, with a mean progression of 3.8±1.1 per patient. Skeletal maturation over the follow-up period was accompanied by a mean spinal height increase of 3.2±1.7 cm preoperatively to the final follow-up (Table 1).

Table 1

Patient demographics

Changes in 3D parameters

The mean Cobb angle, AVR, and MVR improved from 48.0°±10.7°, 9.1°±5.7°, and 13.4°±5.7° preoperatively to 22.3°±8.9°, 6.2°±4.8°, and 9.8°±4.3° postoperatively, respectively, and correction was maintained at 24.1°±11.9°, 6.0°±5.7°, and 11.2°±5.8°, respectively, at 2 years. The torsion index remained similar preoperatively at 2 years (4.1°±3.3° vs. 4.5°±3.1°, p=0.631) despite initial improvement (4.5°±3.1° to 2.0°±2.0° immediately postoperatively). Sagittal profiles, including T1–T12 kyphosis and L1–L5 lordosis, remained similar (Table 2).

Table 2

Coronal, axial, and sagittal profiles at preoperative, immediate postoperative, and 1- and 2-year postoperative follow-ups

Relationship between preoperative parameters and 3D changes

Preoperative AVR (Pearson correlation coefficient [r]=0.26; p=0.048) and MVR (Pearson correlation r=0.43; p<0.001) correlated with the preoperative Cobb angle. Preoperative AVR, MVR together with AVR and MVR derotation correlated significantly with 1-year curve regression (Pearson correlation r=0.40–0.56; p<0.001 to p=0.002), 1-year correction rate (Pearson correlation r=0.37–0.46; p<0.001 to p=0.004), and 2-year curve regression (Pearson correlation r=0.35–0.49; p=0.002 to 0.038). Preoperative AVR (Pearson correlation r=0.44; p=0.024) and AVR derotation (Pearson correlation r=0.44; p=0.025) in thoracic curves also correlated with 2-year correction rate. The correlations were more significant in thoracic VBT (Pearson correlation r=0.44–0.62; p<0.001 to p=0.024). Preoperative curve flexibility correlated with 1-year correction rate (Pearson correlation r=0.40; p=0.003) and curve regression (Pearson correlation r=0.31; p=0.030), but the correlations weaned off at 2 years (p>0.05) (Table 3).

Table 3

Relationships between axial plane parameters with coronal and sagittal profile changes at 1 year and 2 years postoperatively

With univariate analysis, preoperative AVR and MVR, AVR and MVR derotation all predicted correction rate and curve regression at 1- and 2-year curve regression (p<0.001 to p=0.038) (Table 4). Stepwise multivariate analysis showed preoperative MVR (unstandardized B=0.842; p=0.022) and AVR derotation (B=0.675; p=0.041) as significant predictors for 1-year correction rate and preoperative MVR for 1- and 2-year curve regression. Hence, increasing intraoperative axial correction was associated with better Cobb correction. After adjusting for other axial parameters, curve flexibility did not fit into our multivariate model (p>0.05) (Table 5).

Table 4

Univariate analysis for predictive value of individual axial parameters

Table 5

Multivariate regression model for predictive value of axial plane parameters

Discussion

We reported that coronal Cobb and axial rotation greatly improved at index operation, and deformity correction was maintained at 2 years. Greater preoperative axial rotation corresponds to more coronal correction, whereas more aggressive intraoperative derotation could contribute to better correction long-term.

Axial deformity was partially corrected intraoperatively, and correction was maintained at 2 years. Previous literature mostly assessed axial deformity with AVR, but a single parameter could not fully describe axial deformation [14,19]. Hence, axial deformity was comprehensively assessed using AVR, MVR, and the torsion index [13,14]. AVR and MVR improved more significantly than torsion, suggesting that VBT reduced rotation severity, but rotational deformity of the entire curve remained. Tether tensioning in VBT prevents severe rotation of individual vertebrae. However, the whole curve could still have unpredictable net twisting toward one side during spinal growth. Positioning of screws impacts correction, with posterior screws at the apex achieving more derotation [21]. Previous studies reported axial correction [9], but Newton et al. [8] also reported that VBT successfully corrected coronal and sagittal deformities but lacked effect on axial correction. A recent 3D study reported progressive axial deterioration after the index procedure [10]. Severe axial-plane deformity resulted in abnormal rotational range of motion and increased the risk of long-term pulmonary dysfunction in patients with AIS [22,23]. Hence, it is of significance that VBT could achieve axial derotation; however, whether progressive correction of overall axial deformation in terms of geometric torsion remains to be investigated.

Coronal curve correction was mostly achieved through the index operation, demonstrating >50% correction. Further growth modulation during spinal growth and skeletal maturation was not significant, but correction was maintained at 2 years, which was consistent with other studies [5–10]. Catanzano et al. [24] reported that growth modulation only occurred in 54% of patients with progressive thoracic curve improvement, and deformity correction was maintained at 2 years. In contrast, longer-term follow-up suggested progressive decrease in postoperative success, with revision rate increasing to 24% (mean follow-up, 5.7 years) [25]. Despite minimal modulation effect in subsequent stages, long-term operative success in VBT could be up to 75% [5–10]. Previous studies reported increased thoracic kyphosis following VBT, with no changes in lumbar lordosis, relating to the anterior compression effect of instrumentation [21]. However, kyphosis showed no significant changes.

Surprisingly, we found increased postoperative coronal correction in patients with more severe axial rotations. This association was not previously reported and could be explained by the relationship between the two planes of deformity. Usually, patients with more severe rotations have more severe curves; hence, achieving a higher percentage correction is easier compared with smaller curves with the same magnitude of corrected Cobb. Another hypothesis is that higher preoperative rotational severity might indicate greater spinal axial mobility, rendering curves more receptive to axial corrective forces, and this could be true in curves where modulation occurs. Coupling of axial and lateral forces within a motion segment is essential for deformity correction [16,26]. Previous finite element analysis underscored the significance of unrestricted axial rotatory motion in VBT mechanics, but there has been uncertainty regarding the adequacy of the mechanical forces generated by the tether to produce a coupling effect [27]. The correlations were especially significant in thoracic curves, which might be due to intrinsic differences between the thoracic and lumbar spine in axial mobility. Axial motion is generally restricted by the positioning and orientation of the pivoting facet joints [28]. Facets are coronally oriented in the thoracic spine, favoring rotation rather than flexion–extension, and the reverse is true for the lumbar spine, with its sagittally oriented facets [21]. Hence, favorable surgical outcomes in VBT could also be achieved in patients with more severe axial-plane deformities. Axial rotation is best assessed by 3D analysis, but this method is time-consuming and not widely assessable. Apical rotation could be estimated using the Nash–Moe method with plain radiographs, and newer methods have emerged with CT/magnetic resonance imaging assistance [29]. Despite our preliminary analysis, future research should explore VBT biomechanics and the true nature of these relationships [27].

Moreover, increasing intraoperative derotation plays a role in coronal correction. Although rod derotation and direct vertebral derotation are commonly used in spinal fusion, derotation in VBT largely involves tether tensioning and screw orientation. Greater tensioning could achieve more derotation by the cord-tensioning effect. As the force exerted by the tether is mainly vertical along the curve instead of transverse [27], placing screws more posteriorly at the apex would help enhance derotation [21]. Placing screws cranial and caudal to the apex more anteriorly with 200 N tension is associated with better rotational correction [1]. Adding a second nonpretensioned tether on the ipsilateral side to a pretensioned first tether could potentially create an endpoint for vertebral rotation [30]. However, excessively tensioning the tether above 300 N could hinder lateral bending and rotational movement [27]. Careful positioning of screws at the apical and peripheral segments with appropriate tensioning is crucial in achieving adequate derotation to improve surgical success.

Although statistically significant relationships could only be observed with AVR and MVR, the clinical significance of the torsion index could not be underestimated. The apical vertebra is often, or adjacent to, the maximally rotated vertebra. Vertebral rotation increases progressively from the distal end of the curve toward the curve apex [31]. Hence, AVR and MVR represent the rotational severity of the curve and have been shown to correlate well with coronal parameters. Intervertebral rotations were maximal at the cranial and caudal ends of the scoliotic curve and decreased toward the apex [14,32], and the mechanical torsion index represents the additive effects of individual intervertebral rotations, indicating theextent to which the vertebrae are continuously rotated within the whole curve [15,33]. As opposed to AVR/MVR, which could potentially denote rotatory severity, the torsion index describes the overall axial deformation of the curve, an intrinsic property of scoliosis. The lack of associations between coronal plane correction and changes in torsion index indicates that mechanical torsion is a distinct property, representing intrinsic rotation. Although surgical correction could yield considerable axial correction in reducing the magnitude of individual vertebral rotations, inherent rotational deformity is not easily amendable, and patients could still experience the detrimental effects of residual curve rotation despite surgical intervention.

Literature reported tether breakage to have an estimated 2-year cumulative rate of 19%, increasing to 50% at 36 months [11,12], and we reported a comparable rate of 28.9% at the final follow-up. Although studies have reported that patients with tether breakage had a higher progression risk [11,12], the correction rate was not significantly affected in our patients with tether breakage, and none required revision surgery. In contrast, two patients with overcorrection required implant removal. Hence, VBT patients should be followed up more frequently to look out for any early signs of overcorrection to the contralateral side [6,9]. These should be identified at early stages, and revision surgery should be contemplated, considering skeletal maturity and residual growth potential of the individual. VBT failure leading to PSF conversion was uncommon, which occurred in only two patients (4.5%). The majority of patients achieved a considerable correction rate at 2 years while avoiding definitive fusion.

This is the first study to investigate the predictive relationship of axial parameters with the greatest number of curves available (n=58) for 3D reconstructive analysis. Individual 3D analysis of VBT curves was an accurate and comprehensive way to assess postoperative outcomes compared with plain radiographs. However, our study was still limited by the sample size and follow-up. Although a single-center study could minimize baseline differences in patient demographics and surgeon factors, a multicenter study could definitely strengthen our findings by incorporating a larger patient sample. Moreover, longer-term follow-up (>5 years) would be needed in a subsequent analysis to determine if observed relationships are still valid as patients mature. The biomechanics of VBT remains uncertain, and whether the predictive role of axial parameters on operative success will persist or diminish long-term still requires extensive investigation.

Conclusions

VBT effectively corrected and maintained coronal and axial deformity at 2 years. Further, axial parameters showed a predictive value in determining postoperative success, with increased preoperative rotation and intraoperative derotation associated with greater coronal correction. To improve surgical success, clinicians should aim for adequate derotation by appropriate tensioning and screw positioning. More research is warranted to investigate the biomechanical nature of these relationships and whether using axial predictors could be applied clinically.

Data Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Article information Continued