Introduction
Adolescent idiopathic scoliosis (AIS) is a three-dimensional spinal deformity affecting 2%–3% of the population [1,2], and correction is a major clinical challenge for patients, families, and healthcare providers. Characterized by lateral spine curvature and vertebral rotation [3], AIS can significantly impair quality of life through pain [4,5], cosmetic concerns, and functional limitations that restrict participation in daily and athletic activities [6,7]. Posterior spinal fusion, long considered the “gold standard” of treatment for moderate to severe AIS [8,9], effectively corrects deformity and stabilizes spine curve. However, fusion reduces spinal mobility [10,11] and can potentially cause long-term complications, including adjacent segment degeneration [12], chronic pain [13], and neurological deficits [14], without improving preoperative respiratory deficits. These limitations have prompted the exploration of alternative motion-preserving surgical treatment [15].
Anterior vertebral body tethering (AVBT) is a promising nonfusion technique [16,17] in which growth modulation is used to achieve correction through a flexible tether attached to the convex side of the spine [18,19]. By restricting growth on the convex side and allowing continued growth on the concave side, the technique aims for gradual correction over time [20] while preserving motion [21] and potentially reducing fusion-related complications [22,23]. However, enthusiasm for AVBT has been tempered by reports of inconsistent outcomes, including overcorrection, inadequate correction, and frequent tether breakage, all of which necessitates revision surgery [24–26].
In parallel, the ApiFix system was developed as a posterior fusionless approach in which a semirigid rod is anchored to the concave side with pedicle screws [20,27]. The aim of this design is to provide controlled deformity correction with potentially fewer complications [23,28]. According to preliminary reports, short-term outcomes have been encouraging, and implant rupture rates might be reduced [29]; nonetheless, evidence is available only from small case series with short follow-up durations. Therefore, in this systematic review and meta-analysis, we independently evaluated the efficacy and safety of AVBT and ApiFix, synthesizing available data to clarify their true clinical value and inform treatment strategies for AIS.
Materials and Methods
Ethics approval and consent to participate
As this research is a meta-analysis, it does not involve any direct human or animal subjects and therefore does not require ethical approval from an institutional review board.
Eligibility criteria
The selection of the studies was guided by a PICO framework: the population (P) consisted of adolescents with idiopathic scoliosis; the interventions (I) were implantations of the ApiFix device and AVBT; no comparison (C) was performed; and outcomes (O) included radiological correction and related complications.
Information sources and search methods for the identification of eligible studies
We conducted a comprehensive literature search in PubMed, EMBASE, Scopus, and the Cochrane databases without any date or language restrictions to identify eligible studies published as late as October 2024. The search strategy included specific terms relevant to the study: “adolescent idiopathic scoliosis,” “tethering,” and “apifix.” Studies were selected by two independent reviewers, and for publications for which the reviewers disagreed, a third reviewer was consulted.
Data extraction and data items
Two independent reviewers extracted data from the selected studies; a third reviewer resolved any disagreements. Corrections of main thoracic curve, thoracolumbar/proximal thoracic curve, thoracic kyphosis, lumbosacral lordosis, and rib hump, as well as tether breakage and reoperations, were among the preoperative and postoperative (2-week and 1-, 2-, and 3-year) outcomes that were included in the extracted data, along with study characteristics.
Risk of bias assessment
Two reviewers assessed the risk of bias in the included studies by using the Methodological Items for Non-Randomized Studies (MINORS) criteria [30]. The quality of comparative studies (0–24 points) was classified as very low (0–4), low (5–7), fair (8–12), or high (13–16), as was that of noncomparative studies (0–16 points): very low (0–6), low (7–10), fair (11–15), and high (16–24).
Assessment of the quality of evidence
We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework to assess the evidence quality for each outcome noted in the meta-analysis [31]. Through this well-established technique, we thoroughly evaluated the quality of evidence supporting each outcome by considering variables such as research design, between-study consistency, precision of impact estimates, risk of bias, and clinical relevance.
Statistical analysis
For all statistical analyses, we used R software ver. 4.1.2 (The R Foundation for Statistical Computing, Vienna, Austria) and RStudio ver. 2024.09.0 (PBC, Boston, MA, USA) with the “metafor,” “meta,” and “dplyr” packages (The R Foundation for Statistical Computing). We conducted a random-effects meta-analysis to evaluate multiple spinal parameters at different postintervention times. The analysis included changes in the thoracolumbar curve, proximal thoracic curve, and rib hump documented 2 weeks, 1 year, 2 years, and 3 years after surgery, if data were available. We also examined alterations in thoracic kyphosis and lumbosacral lordosis 1 year after surgery.
We used the random-effects model (DerSimonian–Laird method) to account for between-study variance, a widely accepted approach in random-effects meta-analyses [32]. For continuous outcomes, effect sizes were calculated as mean differences between postintervention and preintervention measurements; for dichotomous outcomes, effect sizes were calculated as untransformed proportions, representing the proportion of events in each study. To avoid any publication bias, we did not exclude studies with zero events; instead, we applied a continuity correction of 0.5 to studies with zero events to ensure valid calculations of proportions and confidence intervals (CIs) [33].
To assess heterogeneity, we calculated τ2 (between-study variance), the Q statistic, and I2, which quantifies the percentage of variability caused by heterogeneity and not chance [34]. As appropriate, studies were organized in clinically relevant subgroups. We also conducted formal tests for subgroup differences to identify how the different subgroups varied in response.
Results
Study selection
The initial search yielded 289 studies. After removal of duplicate reports, the remaining 199 studies were assessed by title and abstract screening; this screening yielded 45 records, the full text of which was then screened. Ultimately, 32 studies (27 of AVBT and five of ApiFix) were evaluated in the meta-analysis. A flowchart depicting the Preferred Reporting Items for Systematic Reviews and Meta-Analyses selection process is shown in Fig. 1.
Study characteristics
Of the 27 studies of AVBT, three were conducted in Turkey [35–37], six in Germany [38–42], 12 in the United States [8,18,19,43–50], four in Canada [51–53], and two in France [54,55]. These studies were dated from as early as 2002 to as recent as 2024 and included a total of 1,263 patients; the mean follow-up time was 29.4 months. The mean ages in these studies ranged from 11.9 to 14.9 years, and the mean preoperative Cobb angle across the studies was 46.54°±5.33°. Of the patients, 1,040 (82%) were girls. The five studies of ApiFix [20,28,56–58] included 167 patients (33 boys and 90 girls) in multicenter settings across Israel, Canada, the United States, Greece, and The Netherlands, and the mean follow-up time was 29 months. The mean ages in these studies ranged from 14.2 to 15.1 years, and the preoperative Cobb angle across the studies averaged 45.35°±1.47°. All the included studies and the baseline characteristics of the patients are listed in Table 1.
Risk of bias assessment
To assess bias in the included studies, we used the MINORS criteria. Most studies received high scores in key areas such as clearly stated aims, inclusion of consecutive patients, and thorough data collection methods. In most studies, patients were monitored for adequate periods, and the attrition rate during follow-up was <5%. However, several studies lacked adequate control or contemporary comparison groups, which affected their overall MINORS scores. All studies’ scores ranged from 18–24, which indicated generally high methodological quality (see Table 2 for details).
GRADE summary
The GRADE assessment of our meta-analyses revealed that the overall certainty of evidence for both AVBT and ApiFix outcomes was low to very low. The main reasons for this low certainty are that studies were observational and heterogeneous, and many included small sample sizes for many of the outcome measures assessed, particularly ApiFix. Despite this, we considered most outcomes highly important, and those additional outcomes for curve changes and complication rates were also rated highly. Only thoracic kyphosis changes and rib hump corrections were rated as of moderate importance. AVBT results had moderate certainty in all except two outcomes; ApiFix results had very low certainty for all except one, the change in the minor curve. Details of the GRADE summary are listed in Table 3.
Outcomes of anterior vertebral body tethering
Changes in main thoracic curve
Two weeks after surgery, the main thoracic curve had reduced by 47.3% (95% CI, 33.4%–61.2%); the mean pooled preintervention thoracic curve was 50.33°±9.51°, the mean postintervention curve was 25.37°±8.86°, and the overall mean change was −23.81° (95% CI, −30.80° to −16.83°; I2=95.4%) (Fig. 2A). After 1 year, the main thoracic curve had reduced by 53.3% (95% CI, 47%−59.5%); the mean pooled preintervention curve was 46.81°±9.01°, the mean postintervention curve was 20.55°±9.37°, and the mean change was −24.95° (95% CI, −27.88° to −22.02°; I2=82.7%) (Fig. 2B). After 2 years, the main thoracic curve had reduced by 47.6% (95% CI, 38.8%−68.3%); the mean pooled preintervention curve was 42.95°±9.10°, the mean postintervention curve was 22.65°±10.47°, and the overall mean change was −22.50° (95% CI, −29.35° to −15.65°; I2=98.7%) (Fig. 2C). After 3 years, the main thoracic curve had reduced by 48.6% (95% CI, 41.2%−56%); the mean pooled preintervention curve was 48.47°±7.60°, the mean postintervention curve was 25.03°±8.59°, and the mean change was −23.56° (95% CI, −27.16° to −19.95°; I2=65.2).
In a subgroup analysis of the main thoracic curve change, curves were categorized according to their preoperative Cobb angles: mild (≤40°), moderate (41°−55°), and severe (>55°). After 2 weeks, the mean change in moderate curves was −19.94° (95% CI, −24.38° to −15.50°; I2=87%; for five studies), whereas the mean change in severe curves was −46.52° (95% CI, −51.76° to −41.29°; for one study). The subgroup differences after 2 weeks were significant (p<0.01) (Supplement 1A). After 2 years, the mean change in mild curves was −10.70° (95% CI, −12.54° to −8.86°; for one study), and the mean change in moderate curves was −23.82° (95% CI, −29.05° to −18.59°; I2=98%; for nine studies). The subgroup differences after 2 years were significant (p<0.01) (Supplement 1B); this indicates a difference in the magnitude of correction between mild and moderate curves.
Changes in proximal thoracic curve
Two weeks after surgery, the proximal thoracic curve had reduced by 47.17% (95% CI, 1.27%−93%); the mean pooled preintervention curve was 32.88°±9.92°, the mean postintervention curve was 17.41°±8.10°, and the overall mean change was −15.51° (95% CI, −30.59° to −0.42°; I2=98.3%). After 1 year, the proximal thoracic curve had reduced by 34.9% (95% CI, 21.6%−48.2%); the mean pooled preintervention curve was 38.84°±12.94°, the mean postintervention curve was 22.11°±10.62°, and the overall mean change was −13.55° (95% CI, −18.71° to −8.40°; I2=95.3). After 2 years, the proximal thoracic curve had reduced by 33.11% (95% CI, 17.9%−48.4%); the mean pooled preintervention curve was 34.34°±10.59°, the mean postintervention curve was 20.81°±10.46°, and the overall mean change was −11.37° (95% CI, −16.61° to −6.13°; I2=96.1%).
In a subgroup analysis of proximal thoracic curve change, curves were categorized according to their preoperative Cobb angles: mild (≤40°), moderate (41°−55°), and severe (>55°). Two weeks after surgery, the mean change in mild curves was −7.96° (95% CI, −11.27° to −4.64°; for two studies), whereas the mean change in moderate curves was −30.90° (95% CI, −34.30° to −27.50°; for one study). The subgroup differences after 2 weeks were significant (p<0.01) (Supplement 2A). After 1 year, the mean change in mild curves was −9.26° (95% CI, −11.22° to −7.29°; for five studies), the mean change in moderate curves was −9.28° (95% CI, −10.82° to −7.73°; for two studies), and the mean change in severe curves was −25.54° (95% CI, −28.12° to −22.97°; for two studies). The subgroup differences after 1 year were significant (p<0.01) (Supplement 2B). After 2 years, the mean change in mild curves was −7.23° (95% CI, −9.62° to −4.84°; for five studies), and the mean change in moderate curves was −22.11° (95% CI, −26.01° to −18.21°; for two studies). The subgroup differences after 2 years were significant (p<0.01) (Supplement 2C).
Changes in thoracolumbar curve
Two weeks after surgery, the thoracolumbar curve had reduced by 55.5% (95% CI, 34.8%–76.15%); the mean pooled preintervention curve was 40.21°±8.61°, the mean postintervention curve was 18.12°±8.62°, and the overall mean change was −22.31° (95% CI, −30.62° to −14.01°; I2=96.6%) (Fig. 3A). After 1 year, the thoracolumbar curve had reduced by 53.8% (95% CI, 37.2%−65.4%); the mean pooled preintervention curve was 42.04°±10.82°, the mean postintervention curve was 20.19°±10.96, and the mean change was −21.53° (95% CI, −27.43° to −15.63°; I2=97.5%) (Fig. 3B). After 2 years, the thoracolumbar curve had reduced by 48% (95% CI, 37.3%−58.7%); the mean pooled preintervention curve was 40.18°±9.78°, the mean postintervention curve was 20.42°±10.09°, and the mean change was −19.29° (95% CI, −23.59° to −14.99°; I2=96.3%) (Fig. 3C). After 3 years, the thoracolumbar curve had reduced by 56.2% (95% CI, 44%–68.3%); the mean pooled preintervention curve was 49.89°±8.15°, the mean postintervention curve was 20.51°±13.8°, and the mean change was −28.02° (95% CI, −34.09° to −21.95°; I2=87%) (Fig. 3D).
In a subgroup analysis of thoracolumbar curve change, curves were categorized according to their preoperative Cobb angles: mild (≤40°), moderate (41°–55°), and severe (>55°). Two weeks after surgery, the mean change in mild curves was −16.23° (95% CI, −24.12° to −8.34°; for two studies), whereas the mean change in moderate curves was −28.38° (95% CI, −39.26° to −17.50°; for two studies). The subgroup differences after 2 weeks were not significant (p=0.08) (Supplement 3A). After 1 year, the mean change in mild curves was −11.59° (95% CI, −14.64° to −8.53°; for four studies), and the mean change in moderate curves was −29.53° (95% CI, −34.33° to −24.74°; for five studies). The subgroup differences after 1 year were significant (p<0.01) (Supplement 3B). After 2 years, the mean change in mild curves was −13.22° (95% CI, −16.01° to −10.43°; for five studies), and the mean change in moderate curves was −24.14° (95% CI, −27.82° to −20.46°; for five studies). The subgroup differences at 2 years were significant (p<0.01) (Supplement 3C).
Corrections of rib hump
One year after surgery, the mean rib hump had decreased by 40% (95% CI, 34.2%−47.5%). The mean pooled preintervention rib hump was 14.37°±5.04°, the mean postintervention rib hump was 8.53°±4.67°, and the overall mean change was −5.87° (95% CI, −6.82° to −4.92°; I2=34.2%). After 2 years, the rib hump had decreased by 35.6% (95% CI, 30.3%–40.8%). The mean pooled preintervention rib hump was 14.37°±5.04°, the postintervention rib hump was 9.31°±5.08°, and the mean change was −5.11° (95% CI, −5.86° to −4.35°; I2=0.0%).
Changes in thoracic kyphosis
Two weeks after surgery, thoracic kyphosis had increased by 15.6% (95% CI, 56% increase to 24.8% decrease); the mean pooled preintervention curve was 20.79°±10.17°, the mean postintervention curve was 23.76°±9.34°, and the overall mean change was 3.25° (95% CI, −5.16° to 11.65°; I2=95.1%) (Supplement 4A). After 1 year, thoracic kyphosis had increased by 1.9% (95% CI, 6.54% increase to 2.7% decrease); the mean pooled preintervention curve was 31.45°±11.65°, the mean postintervention curve was 31.92°±11.19°, and the overall mean change was 0.60° (95% CI, −0.88° to 2.09°; I2=53.8%) (Supplement 4B). After 2 years, thoracic kyphosis had increased by 3.25% (95% CI, 20.04% increase to 13.6% decrease); the mean pooled preintervention curve was 24.90°±11.92°, the mean postintervention curve was 24.69°±12.53°, and the overall mean change was 0.81° (95% CI, −3.38° to 4.99°; I2=93.2%) (Supplement 4C).
In a subgroup analysis of thoracic kyphosis change, curves were categorized according to preoperative measurements: hypokyphosis (<20°), normal kyphosis (20°–50°), and hyperkyphosis (>50°), across multiple time points. After 1 year, the mean change in hypokyphosis was 0.84° (95% CI, −1.05° to 2.73°, for five studies); the mean change in normal kyphosis was 1.46° (95% CI, −1.13° to 4.06°; for five studies), and the mean change in hyperkyphosis was −1.69° (95% CI, −3.68° to 0.30°; for two studies). The subgroup differences after 1 year were not significant (p=0.09) (Supplement 5A). After 2 years, the mean change in hypokyphosis was 5.25° (95% CI, 1.08° to 9.41°; for four studies), the mean change in normal kyphosis was 0.52° (95% CI, −1.69° to 2.74°; for seven studies), and the mean change in hyperkyphosis was −14.50° (95% CI, −17.26° to −11.74°; 1 study). The subgroup differences after 2 years were significant (p<0.01) (Supplement 5B), which indicates a difference in the degree of correction among the three kyphosis categories at 2 years. The results of this combined analysis show that the correction was greater in patients with hyperkyphosis, especially after 2 years, than in those with normal kyphosis or hypokyphosis.
Changes in lumbosacral lordosis
Two weeks after surgery, lumbosacral lordosis had decreased by 8.9% (95% CI, 13.9% increase to 31.7% decrease); the mean pooled preintervention curve was 60.37°±16.18°, the mean postintervention curve was 54.91°±12.80°, and the overall mean change was −5.38° (95% CI, −19.15° to 8.38°; I2=92.4%). After 1 year, lumbosacral lordosis had decreased by 5.6% (95% CI, 5.3% increase to 16.5% decrease); the mean pooled preintervention curve was 51.12°±10.95°, the mean postintervention curve was 49.70°±10.67°, and the overall mean change was −2.87° (95% CI, −8.42° to 2.69°; I2=97.3%). After 2 years, lumbosacral lordosis had increased by 1.3% (95% CI, 3.4% increase to 0.83 decrease); the mean pooled preintervention curve was 53.22°±12.26°, the mean postintervention curve was 54.01°±11.95°, and the overall mean change was 0.68° (95% CI, −0.44° to 1.80°; I2=0.0%). After 3 years, lumbosacral lordosis had decreased by 23% (95% CI, 4.2% increase to 88% decrease); the mean pooled preintervention curve was 39.07°±10.98°, the postintervention curve was 36.63°±8.13°, and the overall mean change was −8.98° (95% CI, −34.46° to 16.50°; I2=99.5%).
In a subgroup analysis of lumbosacral lordosis change, curves were categorized according to their preoperative measurements as normal lordosis (30°–50°) and hyperlordosis (>50°), across multiple time points. One year after surgery, the mean change in normal lordosis was −6.06° (95% CI, −23.61° to 11.49°; four studies), and the mean change in hyperlordosis was −0.78° (95% CI, −1.91° to 0.35°; seven studies). The subgroup differences at 1 year were not significant (p=0.56). After 2 years, the mean change in normal lordosis was 3.29° (95% CI, −2.85° to 9.44°; three studies), and the mean change in hyperlordosis was 0.44° (95% CI, −0.80° to 1.67°; six studies). The subgroup differences were not significant (p=0.37); this indicates no significant variation in lumbosacral lordosis changes across subgroups.
AVDB-related complications
The pooled analysis revealed an overall implant-related complication rate of 26% (95% CI, 18%–33%; I2=95%) (Fig. 4A). We performed two leave-one-out meta-analyses: the one excluding the study by Baroncini et al. [42] in 2022 revealed the highest implant-related complication rate (26.8%), and the one excluding the study by Trobisch et al. [43] in 2024 revealed the lowest implant-related complication rate (22.6%) (Supplement 6A).
Tether breakage rate: According to the pooled analysis, the overall rate of tether breakage was 25% (95% CI, 17%–33%; I2=97%) (Fig. 4B) The analysis of tether breakage rates across the four follow-up periods revealed notable differences: the breakage rate was 20% (95% CI, 9%–31%) after 1 year, 6% (95% CI, 2%–9%) after 2 years, 48% (the highest rate; 95% CI, 25%–72%) after 3 years, and 3% (95% CI, 0%–6%) after 4 years. The subgroup differences were significant (p<0.01) (Supplement 6B).
Reoperation rate: According to the pooled analysis, the overall rate of reoperation was 7% (95% CI, 4%–10%; I2=58%) (Fig. 4C). The leave-one-out meta-analysis revealed that the overall reoperation rate was lowest (5%) when the study by Stein et al. [35] in 2024 was excluded (Supplement 6C).
Outcomes of ApiFix treatment
Changes in main thoracic curve after 2–3 years
According to the pooled analysis, the main thoracic curve was reduced by 47.2% (95% CI, 40.1%–54.3%) within 2–3 years; the mean pooled preintervention curve was 45.07°±6.58°, the mean postintervention curve was 22.74°±6.02°, and the overall mean change was −21.28° (95% CI, −24.46° to −18.09°; I2=85.7%). No outlier study was identified by sensitivity analysis.
Changes in minor curve after 2–3 years
According to the pooled analysis, the minor curve decreased by 29.1% (95% CI, 22.1%–36%) within 2–3 years; the mean pooled preintervention curve was 30.11°±8.85°, the mean postintervention curve was 21.43°±7.55°, and the overall mean change was −8.75° (95% CI, −10.83° to −6.66°; I2=0.0%).
Changes in thoracic kyphosis after 2–3 years
According to the pooled analysis, thoracic kyphosis decreased by 26.8%; the mean pooled preintervention curve was 37.41°±7.00°, the postintervention curve was 22.06°±8.00°, and the overall mean change was −10.04° (95% CI, −35.52° to 15.44°; I2=99.2%).
Changes in lumbar lordosis after 2–3 years
According to the pooled analysis, lumbar lordosis decreased by 37%; the mean pooled preintervention curve was 44.18°±7.00°, the mean postintervention curve was 22.94°±8.00°, and the overall mean change was −16.55° (95% CI, −39.09° to 5.99°; I2=99.0%).
Reoperation rate
According to the pooled analysis, the overall reoperation rate was 25% (95% CI, 9% to 42%; I2=81%). In a leave-one-out meta-analysis, the reoperation rate was highest (31.5%) when the study by Floman et al. [20] in 2021 was excluded and lowest (18.1%) when the study by Stadhouder et al. [57] in 2021 was excluded.
ApiFix-related complication rate
According to the pooled analysis, the overall rate of implant-related complications was 22% (95% CI, 8%–36%; I2=76%). In the leave-one-out meta-analysis, the implant-related complication rate was highest (27.3%) when the study by Floman et al. [20] in 2021 was excluded and lowest when the study by Stadhouder et al. [57] in 2021 was excluded.
Meta-regression analysis of predictive factors
We performed univariate meta-regression analysis to identify baseline patient characteristics associated with Cobb angle correction after AVBT. We identified 21 statistically significant associations across the outcomes analyzed (Table 4).
Skeletal maturity emerged as the strongest predictor of correction magnitude. Skeletal maturity according to the Sanders Skeletal Maturity Staging System had the highest explanatory power, accounting for 71.8% of between-study heterogeneity in proximal thoracic Cobb angle correction at the 1-year follow-up visit (β=−5.39°/stage; 95% CI, −7.71 to −3.06; p<0.0001). Similarly, the grade according to the Risser staging system was significantly related to thoracolumbar Cobb angle correction (β=−10.37°/grade; 95% CI, −17.25 to −3.49; p=0.003, R2=40.0%): lower Risser grades were associated with greater correction. Patient age at surgery was also a significant independent predictor (β=−7.84°/year for thoracolumbar Cobb angle; p=0.003): correction was substantially better in younger patients.
Baseline curve characteristics significantly influenced outcomes. The initial main Cobb angle was predictive of proximal thoracic curve correction (β=−1.17°/degree; 95% CI, −1.83 to −0.50; p=0.0006, R2=57.0%), which indicates that absolute correction was greater for larger curves. Baseline thoracic kyphosis was positively associated with correction magnitude across multiple outcomes (β ranged from −0.96° to −1.07° per degree; p<0.01). We found an unexpected positive association of preoperative curve flexibility with less correction (β=1.51°/%; 95% CI, 0.85–2.18; p<0.0001, R2=71.0%), which suggests that achieving optimal tether tension may be more challenging in highly flexible curves.
These findings have important implications for patient selection. The strong association between skeletal immaturity and correction success supports the growth modulation mechanism of AVBT and suggests that ideal candidates are patients with skeletal maturity equivalent to Sanders stages ≤4 and Risser grades 0–1 who possess sufficient remaining growth potential. The patients with moderately severe curves (45°–55°) demonstrated continued correction over time, whereas for larger curves, maximum correction was achieved early and followed by stability; this observation, which may indicate that optimal biomechanical tension was achieved in moderately severe, moderately flexible curves, is consistent with the Hueter-Volkmann law of asymmetric growth modulation.
Heterogeneity
Leave-one-out sensitivity analyses revealed notable changes in heterogeneity across multiple outcomes. For main thoracic curve change at 1 year, the exclusion of the study by Boudissa et al. [51] in 2017 eliminated heterogeneity completely (I2 from 82.7% to 0%) (Supplement 7A). For thoracic kyphosis 1 year after surgery, the exclusion of the study by Courvoisier et al. [46] in 2023 revealed the largest reduction in heterogeneity (I2 from 53.8% to 19.2%) (Supplement 7B). For lumbosacral lordosis 1 year after surgery, the exclusion of the study Boeyer et al. [54] in 2023 revealed the most substantial reduction in heterogeneity (I2 from 97.3% to 59.4%) (Supplement 7C). These findings suggest that the studies by Boudissa et al. [51], Courvoisier et al. [46], and Boeyer et al. [54] are potential sources of heterogeneity, and their results should be interpreted with caution.
Publication bias
To assess for publication bias, we used funnel plots for outcomes with at least 10 studies. For AVBT, funnel plots were created for changes in the main thoracic curve 2 weeks, 1 year, and 2 years after surgery; for changes in the thoracolumbar curve 1 and 2 years after surgery; for changes in the proximal thoracic curve, thoracic kyphosis, and lumbosacral lordosis 1 and 2 years after surgery; and for tether breakage, implant-related complications, and reoperation rates. The appearance of these funnel plots suggested potential asymmetry for some outcomes, particularly for the changes in the main thoracic curve after 2 years and in the thoracolumbar curve after 1 and 2 years (Supplement 8); this finding suggests possible publication bias. However, the funnel plots should be interpreted cautiously because of the high heterogeneity observed in most outcomes. For ApiFix, the limited number of studies (<10 for all outcomes) precluded a meaningful assessment of publication bias via funnel plots. The potential for publication bias in the ApiFix results should be considered when these findings are interpreted.
Discussion
In this meta-analysis, we integrated findings from 27 studies of AVBT and five studies of ApiFix, covering a substantial patient cohort of 1,263 patients over various follow-up periods (up to 3 years). This comprehensive and updated review of these emerging growth modulation techniques for AIS reveals that the initial enthusiasm, which centered on the potential for correcting spinal curves while preserving motion [16,17], might have been premature, in view of the insufficiency of long-term data. The long-term efficacy and safety of these techniques are still subjects of ongoing debate [20,22], and they appear not to have met initial expectations. The pursuit of alternative techniques is driven by the well-known long-term consequences of traditional spinal fusion, such as disc degeneration [59], diminished quality of life [38], elevated blood levels of metal ions [60], and potential obstetric complications. These concerns underscore the importance of meticulously evaluating the risk–benefit profile of each treatment approach to ensure the most favorable long-term outcomes for patients with AIS. Of importance is that short-term complications resulting from implant failures or insufficient correction of the deformity are very common, which accounts for the very high percentages of poor results in the short term.
To contextualize these findings, we compared our pooled findings of AVBT outcomes with published data about posterior spinal fusion (Table 5). Although posterior spinal fusion achieves superior coronal correction (60%–70% vs. 51.2%) [61,62], AVBT demonstrates distinct advantages with regard to sagittal plane preservation (thoracic kyphosis change: 0.8° vs. −5° to −10°) [63,64], reduced blood loss (150–300 mL vs. 500–1,500 mL), shorter hospitalization (2–4 days vs. 4–7 days), and, of greater importance, maintenance of spinal motion and growth potential. The higher overall complication rate with AVBT (26% vs. 10%–15%) [65,66] reflects predominantly implant-related events (tether breakage, anchor issues), which are often not clinically obvious; in contrast, with posterior spinal fusion, complications include primarily neurological injury (0.5%–1.3%) [67] and deep infection (1.2%–2.8%) [68]. The higher AVBT reoperation rate (8% vs. 3%–6%) reflects primarily conversion to fusion in cases of inadequate growth modulation. These data support AVBT as a complementary strategy rather than as an alternative to posterior spinal fusion, offering particular advantages for skeletally immature patients with moderate, flexible curves in whom motion preservation and growth potential are prioritized over maximum immediate correction.
Our analyses of both AVBT and ApiFix revealed incomplete reductions in the main thoracic and thoracolumbar curves, minimal to no improvement in sagittal plane deformities, and scarce rotation correction. The initial reduction in curves is lacking; the curves are not significantly corrected either at the onset of or after modulation, contrary to the claims made by the proponents of these techniques. The incomplete correction of scoliosis across one or more planes has severe consequences for these patients, not only cosmetically but also in terms of respiratory function and vertebral degeneration. In these systems, unidirectional corrective forces are used to address a three-dimensional vertebral deformity. Furthermore, it is impossible to predict the amount of correction during modulation because in most cases, correction does not occur and modulation sometimes even results in overcorrection. Moreover, tether breakage continues to be a significant concern, with an overall incidence of 22.2%, and increased breakage has been noted at the 3-year follow-up visit [42], which is worrisome.
In comparison, the evidence for ApiFix is more limited because of the smaller numbers of studies and patients (167 patients) and shorter follow-up periods. Initial findings suggest that ApiFix may also result in incomplete correction of the coronal plane, accompanied by a loading effect in the sagittal plane and minimal correction of rotation [27,28]. The serious issues stemming from inadequate correction of scoliosis across all three planes consequently persist. In both techniques, the implant remains in place for life, which means that complications could increase over the medium and long terms.
Studies have shown that anterior AVBT systems can increase thoracic kyphosis and decrease lumbar lordosis, whereas earlier posterior constructs were sometimes associated with hypokyphosis. However, modern posterior fusion techniques involving segmental pedicle screw constructs, optimized sagittal rod contouring, and patient-specific alignment targets are designed to restore or enhance thoracic kyphosis, thereby minimizing the risk of sagittal imbalance and flat back syndrome [19,24,36]. This incomplete three-dimensional correction can have significant functional and cosmetic implications for patients, particularly as aesthetic outcomes are paramount for patient satisfaction [69].
Furthermore, the purported motion preservation benefits of AVBT and ApiFix may come at the cost of compromised mobility in the instrumented segments, particularly in the T7–T10 thoracic region [70]. This region, which often encompasses the apex of the scoliotic curve, plays a crucial role in respiratory mechanics [70]. Maintaining mobility in this segment is particularly important during periods of high respiratory demand, and the vertebral implants used to promote modulation may compromise this function. This potential for respiratory compromise is further compounded by the risks inherent in anterior approaches, including segmental vessel bleeding, hemothorax, chylothorax, sympathetic trunk lesions, and pulmonary complications; pleural effusion is the most frequent [24,71]. Mathew et al. [59] and Buyuk et al. [45] documented some residual sagittal motion after AVBT; however, neither study focused on the effect of AVBT on respiratory function.
The high complication and reoperation rates, frequently reaching approximately 50% in large-scale AVBT studies [18,24,60], necessitate a critical reevaluation of the current use of vertebral growth modulation techniques and their potential long-term consequences. Newton et al. [60], despite being proponents of AVBT, reported a 41% reoperation rate in their study as a result of overcorrection, implant failure, and progression of the lumbar curve. A further concern from their analysis is that most curves at the final follow-up visit exhibited >20° of residual deformity. This finding is particularly alarming, inasmuch as Burgos et al. [72] demonstrated an association between residual curves of >20° and an increased risk of disc degeneration and diminished quality of life at long-term follow-up. This underscores the importance of achieving and maintaining significant correction, ideally <10°, to minimize long-term risks and optimize patient outcomes [73].
Although AVBT and ApiFix are currently the growth modulation techniques most widely studied, alternative modulation approaches are emerging. Burgos et al. [73] in 2023 presented a compelling alternative: utilizing a posterior approach with pedicle screws but without fusion in skeletally immature patients. The aim of this technique is a complete three-dimensional correction, and all of the patients by Burgos et al. [73] achieved curves of <10°. These corrections are maintained at the end of growth and after the removal of the implants. Because implants are removed when patients reach skeletal maturity, the long-term risks associated with permanent instrumentation are therefore mitigated. Early results almost 3 years after surgery included no loss of correction, no complications or reoperations, and satisfactory quality-of-life outcomes (according to Scoliosis Research Society-22 scores).
This meta-analysis builds upon previous research by encompassing a wider range of studies and patients [9,42,74]. It specifically addressed knowledge gaps identified by earlier research with short follow-up periods with regard to the limitations of three-dimensional correction, the potential effect on respiratory function, and the concerningly high reoperation rates associated with AVBT and ApiFix. Whereas earlier studies highlighted the promise of AVBT and, more recently, ApiFix, our findings offer a more nuanced and critical perspective. These techniques, which produced short-term complications in almost half the cases, produce an insufficient correction of scoliosis with known consequences of aesthetic deformity, respiratory impairment, and early degenerative changes. These outcomes may be more frequent and extensive than those produced by conventional vertebral fusion because they affect the twisted disc levels included in the instrumentation that maintain mobility, and the inability of these procedures to distribute loads in nonphysiological positions will lead to severe degenerative changes [42,75].
In addition, leaving metal implants inside the body has caused critical complications in many patients. These implants are not removed; nearly 50% of patients with these implants experience implant rupture, and 7% undergo reoperation. The reluctance to perform surgical intervention, despite the technical challenges involved, leads to worsening of scoliotic deformities and increased mechanical stress at the site of the rupture, which has even more severe consequences. These problems are expected to worsen over the long term and will probably be more severe than those resulting from conventional spinal instrumentation and fusion techniques.
To be effective, modulation techniques must initially correct scoliosis fully, they must be used during the growth period to allow vertebral bodies to correct their deformity, the material must be removed to prevent metalloids, and mobility should be preserved once the implant is removed and modulation has concluded. Neither AVBT nor ApiFix addresses any of these processes adequately. However, the technique developed by Burgos et al. [73] has the potential to meet these requirements.
Patient selection and indications
Appropriate patient selection is critical for achieving favorable outcomes with growth modulation techniques. Evidence suggests that AVBT yields the best results in skeletally immature patients (Risser grades 0–2, Sanders stage ≤5) who have curves ranging from 40° to 60° that are moderately flexible and who have the potential for further spinal growth [18,19,24,36]. In these cases, continued vertebral growth allows gradual correction and modulation across the tethered segments. However, in patients who are almost skeletally mature, correction may be incomplete, or tether breakage can result from excessive tension that occurs because of limited growth. Conversely, the ApiFix device appears more suitable for adolescents closer to skeletal maturity (Risser grades 3–5) with flexible, single curves of <60°, in whom correction depends on controlled dynamic adjustment rather than differential growth [20,27,28].
Both techniques have resulted in poorer outcomes and higher complication rates with rigid curves, double major deformities, or curves exceeding 65°. These findings emphasize that curve flexibility and growth potential (not merely chronological age) should guide the choice of fusionless correction. Therefore, careful preoperative assessment, including bending films and growth potential evaluation, remains essential for patient selection and surgical planning.
Limitations
Although this meta-analysis provides a comprehensive overview of the existing evidence on AVBT and ApiFix, it did have limitations. The follow-up period generally extended only to the end of growth or shortly thereafter, which restricted information about long-term outcomes, the durability of correction, the potential for late complications, and the overall effect on spinal health. The relatively short follow-up period in many studies, particularly those focusing on ApiFix, further constrained our ability to assess long-term effects.
In addition, publication bias, which favors studies with positive outcomes, could have skewed the overall depiction of AVBT and ApiFix effectiveness. It is also noteworthy that although some studies suggested that the incidence of tether breakage does not significantly increase with longer follow-up, inconsistencies in reporting the specific timing of these events hampered us in drawing definitive conclusions about the long-term risk of tether failure.
Conclusions
This meta-analysis revealed that AVBT and ApiFix provide meaningful but incomplete deformity correction: on average, coronal curves improved by approximately half, and residual deformity typically persisted at early to mid-term follow-up. These techniques achieve only partial reductions in the curvature—correcting nearly 50% of the curve on average—and significant modulation or alteration of the initially achieved correction was not confirmed in follow-up. In some cases, even reversals occurred, with a loss of the initial correction observed. Moreover, these techniques did not effectively address either sagittal plane or rotational deformities. Significant clinical concerns include the persistence of residual deformity and the occurrence of implant-related mechanical complications at the vertebral levels where failure occurs. Approximately 50% of patients treated with AVBT experienced tether breakage, and approximately 25% of those undergoing ApiFix required reoperation as a result of implant loosening, failure, or overcorrection. These complication rates are particularly noteworthy because the treated population consists of young patients with a long-life expectancy, which emphasizes the importance of long-term durability and safety. Because the current evidence reflects high rates of mechanical failure and the need for reoperation, AVBT and ApiFix should be used with caution, and patients should be selected with extreme caution until more robust long-term data become available.
PDF Links
PubReader
XML Download
Full text via DOI
Download Citation
Supplement
