The Utility of Cervical Vertebral Maturation Method for Staging Skeletal Growth and Curve Progression in Patients with Adolescent Idiopathic Scoliosis

Abstract

Background:

Cervical vertebral maturation (CVM) is a skeletal maturity method that can be assessed routinely on whole spine radiographs to minimize radiation exposure. Originally used in orthodontics, its role in staging adolescent growth spurt and curve progression in adolescent idiopathic scoliosis (AIS) remains unclear. The aim of this study was to investigate growth rates across CVM stages, its cutoff for indicating peak growth (PG) versus growth cessation (GC), and its relationship with coronal curve progression.

Methods:

One hundred forty-two AIS patients were prospectively followed from Risser stage 0, until growth completion. Longitudinal data collected included arm span (AS), body height (BH), sitting height (SH), coronal Cobb angle, and maturity assessments. CVM was evaluated through its relationship with growth rates and curve progression rates. A total of 1107 spine radiographs corresponding to longitudinal growth rates were analyzed to detect PG and GC in each patient, with predictive accuracy assessed using receiver operating characteristic curve analysis. Curve progression rate of each CVM stage in treatment-naïve patients was plotted against timing to peak curve progression.

Results:

CVM correlated most with Proximal Femur Maturity Index (PFMI) (τ_b = 0.662, p < 0.001). CVM stage 3 and 6 showed the respective highest and lowest mean growth rates in SH and AS. CVM stage 3 predicted PG with an area under the curve (AUC) of 0.711 to 0.720. CVM stage 5 predicted GC with AUC of 0.840 to 0.850. CVM stage 3 had the highest curve progression rate (0.45° per month). Peak curve progression occurred 5.8 months after CVM 3 and 9.1 months before CVM 4, lagging behind PG by 6.5 months.

Conclusions:

CVM stage 3 indicates peak growth, while stage 6 marks growth cessation. In this cohort of AIS patients, GC is more accurately predicted than PG by CVM. Peak curve progression occurred between CVM stage 3 and CVM stage 4.

Clinical Relevance:

This study highlights CVM method’s ability in indicating timing of growth cessation. CVM can be used to indicate curve progression beyond peak growth, especially until the point of growth cessation.

Level of Evidence:

Prognostic Level II. See Instructions for Authors for a complete description of levels of evidence.

Introduction

Background Rationale

In adolescent idiopathic scoliosis (AIS), accurate skeletal maturity assessment is essential in determining the timing of peak growth and risk of curve progression^1-3. Risser staging is commonly used, but can both underestimate⁴ and overestimate⁵ skeletal maturity, and may mismatch with other indices⁴.

Cervical vertebral maturation (CVM) is a skeletal maturity method that can be assessed on lateral spine radiographs, with its orthodontic use dating back to 1975⁶. Peak height velocity (PHV) was linked⁷ to CVM stage 3, but the roles of CVM 1 in relation with peak growth (PG), and CVM 5 and 6 to growth cessation (GC) remain unclear. Also, there is an absence of studies⁸ on the sensitivity and specificity of CVM method in predicting PG and GC. This longitudinal study addresses these gaps, alongside CVM method’s prediction for timing to peak curve progression^4,9-11.

Objectives

The aim of this study was to (1) assess growth rate variations across CVM stages and their association with other skeletal maturity indicators, (2) determine the predictive accuracy of CVM for peak growth (PG) and growth cessation (GC), and (3) examine the relationship between CVM stages and the degree of curve progression, and timing of peak curve progression in AIS patients.

Methods

Study Design

Skeletally immature (Risser stage 0) patients of Chinese descent who were referred to our tertiary scoliosis clinic with the diagnosis of AIS between September 2015 and July 2020 were screened for recruitment (Appendix 1). Inclusion required a left hand-wrist radiograph and an EOS whole spine radiograph (biplanar stereoradiography) on the same day at initial consultation. Patients were followed till growth completion, defined as a body height growth rate of <1 cm/year¹² between 2 consecutive visits. Skeletal maturity was assessed at 4- to 6-month intervals using multiple radiographic measures, including Risser staging, Sanders Staging (SS), Distal Radius and Ulna Classification (DRU)^13-16, Proximal Femur Maturation Index (PFMI)¹⁷, Thumb Ossification Composite Index (TOCI)¹⁸, and Proximal Humerus Ossification System (PHOS)¹⁹. Ethics committee approval and informed consent of the parents and patients were obtained.

Data Collection

Patient demographic information collected including age, sex, curve type, menarche status, bracing, and surgical treatment status. Standing body height (BH), sitting body height (SH) and arm span (AS) were measured by an allied healthcare individual, and coronal major curve magnitude (in Cobb angles) was measured by the attending orthopaedic surgeon at every visit. Both had no prior knowledge of the study and were not involved in the data analysis. Most follow-up appointments were scheduled at an interval of 4 to 6 months²⁰.

Definitions of Cervical Vertebral Maturation

CVM²¹ consist of stages 1 to 6 (Fig. 1). Stage 1: inferior borders of cervical vertebral bodies C (cervical vertebrae) 2 to C4 are flat; Stage 2: the notch is present along the inferior border of the odontoid process (C2); Stage 3: there is visible notching of the inferior borders of C2 and C3; Stage 4: all 3 C2, C3, and C4 bodies have obvious concavities along their inferior surfaces, with C3 and C4 possessing rectangular horizontal shapes; Stage 5: at least 1 of C3 or C4 vertebral bodies become square shaped. Stage 6: at least 1 of the C3 or C4 has assumed a rectangular vertical morphology, with the length of the posterior border being longer than the inferior border.

Fig. 1.

Schematic diagrams and definitions of the CVM method. CVM = cervical vertebral maturation.

Reliability Test

Inter-rater and intrarater reliability for CVM were evaluated using the weighted kappa coefficient, given 6 ordinal categories. Minimum sample size was 72 radiographs^22-24, calculated by the formula²⁵ 2k² (k = number of categories = 6). Hence, all lateral spine radiographs from 10 randomly selected patients were included, yielding a total of 89 radiographs. The selection was performed by an investigator (J.P.Y.C.) not involved in the grading process. Two readers (S.T.Y.C. and P.W.H.C.), blinded to the patient's demographic and growth data, independently assessed the CVM stages. Intrarater reliability was assessed 4 weeks later. Weighted kappa coefficient (κ_w) of ≥0.75 indicates excellent agreement, κ_w between 0.40 and 0.75 represents fair-to-good agreement beyond chance, and κ_w of ≤ 0.40 represents poor agreement²⁶. Good-to-excellent reliability was found for CVM, with intraobserver κ_w = 0.76 (95% CI: 0.67-0.85) and interobserver κ_w = 0.69 (95% CI: 0.59-0.79).

Statistical Analysis

Growth rates (cm/month) were calculated by the interval change of bodily growth divided by the exact time interval. The correlation between CVM stages, growth rates, and other indices was examined using the Kendall rank correlation. Growth rates across stages were analyzed using Kruskal-Wallis one-way analysis of variance with Bonferroni correction for multiple comparisons. Cross-tabulation showed the distribution of other indices within CVM stages.

To evaluate predictive accuracy for PG and GC, receiver operative characteristic (ROC) curve and area under the curve (AUC) analysis were performed. PG was defined as the highest growth rate, and GC was defined as the following: BH and AS ≤ 0.15 cm/month¹³, and SH ≤ 0.10 cm/month. The optimal CVM stage for PG and GC was determined based on the best sensitivity and specificity.

For subgroup analysis of curve progression, only treatment-naïve patients and those whose follow-ups occurred before the initiation of bracing or surgery were included, to exclude confounding effects from treatment. Curve progression was calculated by rate of Cobb angle change (degrees/month). The visit displaying the largest curve progression was identified as the peak curve progression. The time difference (months) between each visit and peak curve progression was calculated.

A p-value of <0.05 was considered statistically significant. Descriptive statistics were reported in means ± standard deviations, with 95% confidence intervals (CIs) when applicable. Analyses were conducted using SPSS (version 29.0.1.0; IBM).

Results

Participants and Descriptive Data

A total of 142 patients (83.8% female) with a major coronal Cobb angle of 22.3° ± 9.0° were recruited (Table I). Curve distribution was thoracic (51.4%), thoracolumbar (23.9%), and lumbar (24.6%). 81% were eventually braced, 16.2% were observed until skeletal maturity, and 2.8% underwent surgery.

Table I.

Demographics of Patients

	Whole cohort	Girls	Boys
No. (%) of patients	142 (100)	119 (83.8)	23 (16.2)
No. (%) of radiographs	1,107 (100)	952 (86.0)	155 (14.0)
Chronological age at first presentation *(yr)	11.2 ± 0.8	11.0 ± 0.7	11.9 ± 0.8
Chronological age at menarche *(yr)		12.5 ± 0.9
Postmenarche † (no.[%]) (n = 952)		505 (53.0)
Months after menarche *		3.9 ± 19.0
Standing body height *(cm)	155.7 ± 10.3	154.1 ± 9.5	164.6 ± 10.4
Arm span *(cm)	154.9 ± 11.1	153.0 ± 10.1	164.9 ± 11.4
Sitting height *(cm)	83.3 ± 5.3	82.7 ± 5.0	87.0 ± 5.7
Coronal Cobb angle *(major curve) (deg)
No bracing	23.3 ± 11.6	23.7 ± 11.8	22.0 ± 11.1
Out of brace measurement at time of radiograph	27.3 ± 10.4	27.1 ± 10.1	29.4 ± 12.3
In brace measurement at time of radiograph	19.0 ± 9.2	18.8 ± 8.9	19.7 ± 10.8

The values are given as the mean and the standard deviation.

^*cm = centimeter, deg = degree, and yr = year.

^†The values are based on the number of radiographs.

Relationship of CVM with Other Indices and Growth Rates

CVM showed significant correlations (p < 0.001 for all) with other skeletal maturity indices (τ_b = 0.61-0.662), except for PHOS (τ_b = 0.558) (Appendix 2). PFMI and TOCI showed the highest correlation with CVM (τ_b = 0.662, p < 0.001), followed by Sanders staging (τ_b = 0.652, p < 0.001). Mean growth rates at peak growth are presented in Appendix 3. CVM was significantly correlated (p < 0.001 for all) with chronological age (τ_b = 0.573), growth rates based on BH (τ_b = -0.436), SH (τ_b = -0.288), AS (τ_b = -0.445), and timing from peak curve progression (τ_b = 0.486).

Growth Rate Variations

CVM 3 saw the highest overall growth rate in AS and SH (Fig. 2). Gender-specific analysis showed that CVM 3 corresponded to peak SH and AS growth rates (Appendix 4). Significant differences in growth rates were found across CVM stages (χ² (6) = 88.004 [p < 0.001] for BH, χ² (6) = 50.629 [p < 0.001] for SH, and χ² (6) = 74.279 [p < 0.001] for AS). Pairwise comparisons revealed that CVM stage 3 had positive mean differences in growth rates in AS, where there was a significant stepwise decrease across CVM stages 4, 5, and 6 (p < 0.001), but no significant differences when CVM 3 was compared with CVM 1 or 2 (p > 0.99) (Appendix 5).

Fig. 2.

Mean growth rate at each CVM stage, along with a cross-reference of CVM stages to Risser stages, SS, distal radius (R) and ulna (U) classification, PFMI, TOCI, and PHOS; the most prevalent stage(s) (in percentages) within each skeletal maturity index is stated for both boys and girls. CVM = cervical vertebral maturation, PFMI = proximal femur maturation index, PHOS = proximal humerus ossification system, SS = Sanders staging, and TOCI = thumb ossification composite index.

CVM 4 was the most prevalent for PG of BH and SH in girls (Appendix 6), which PG occurred 9.2 months after the last recorded CVM 3 stage and 18.0 months before the appearance of CVM 5.

Peak Growth Prediction

AUC analysis (Table II) indicated that PG could be predicted by CVM, with a cutoff of CVM 3.5.

Table II.

Receiver Operating Characteristic (ROC) Curve Analysis for Peak Growth and Growth Cessation (≤0.15 cm/Month for BH and AS), (≤0.10 cm/Month for SH)^*

Peak Growth Based on Growth Parameters (BH, SH, and AS)	AUC	95% CI	p	Cutoff Stage of CVM (≥) †	Sensitivity (%)	Specificity (%)
Girls
BH	0.721	0.678-0.765	<0.001	3.5	79.0	50.0
SH	0.702	0.657-0.748	<0.001	3.5	78.2	48.1
AS	0.72	0.676-0.765	<0.001	3.5	77.6	55.0
Boys
BH	0.699	0.607-0.792	<0.001	3.5	63.8	59.1
SH	0.737	0.653-0.821	<0.001	3.5	67.7	69.6
AS	0.696	0.604-0.787	<0.001	3.5	63.3	60.9
Overall
BH	0.720	0.680-0.758	<0.001	3.5	76.8	51.5
SH	0.711	0.670-0.750	<0.001	3.5	76.7	51.9
AS	0.718	0.677-0.757	<0.001	3.5	75.3	56.0

Growth Cessation Based on Growth Parameters (BH, SH, and AS)	AUC	95% CI	p	Cutoff Stage of CVM (≥)†	Sensitivity (%)	Specificity (%)
Girls
BH	0.847	0.817-0.877	<0.001	4.5	86.5	69.9
SH	0.832	0.799-0.865	<0.001	4.5	85.6	68.3
AS	0.832	0.799-0.865	<0.001	4.5	82.5	70.8
Boys
BH	0.870	0.797-0.943	<0.001	5.5	77.8	87.7
SH	0.889	0.824-0.955	<0.001	5.5	76.9	87.5
AS	0.957	0.915-0.999	<0.001	5.5	83.3	97.2
Overall
BH	0.849	0.821-0.877	<0.001	4.5	86.8	69.3
SH	0.840	0.810-0.870	<0.001	4.5	87.0	68.1
AS	0.850	0.821-0.879	<0.001	4.5	84.0	71.6

^*AUC = area under the curve, AS = arm span, BH = standing body height, SH = sitting height, CI = confidence interval, and CVM = cervical vertebral maturation.

^†Cutoff values are the averages of 2 consecutive ordered observed test values.

Growth Cessation Prediction

AUC analysis suggested that GC could also be predicted by CVM. The optimal cutoff for GC was CVM 4.5 (Table II), though a CVM cutoff of 5.5 displayed the highest specificity (Appendix 7). The mean BH, SH, and AS at CVM 3 and 6, along with their respective gains are presented in Appendix 8.

Relationship of CVM stages with curve progression

Coronal curve progression was highest at CVM 3 (0.45° per month) and lowest at CVM 6 (0.05° per month) (Fig. 3). Peak curve progression occurred between CVM 3 and CVM 4, 5.8 months after CVM 3, and 9.1 months before CVM 4.

Fig. 3.

Curve progression rate by CVM stage in treatment-naïve patients. The plot compares the timing to peak curve progression with the mean curve progression rate and mean growth rate, highlighting the relationship between the CVM stage and the timing of peak curve progression. CVM = cervical vertebral maturation.

Discussion

Two-thirds of pubertal growth and PHV occur before iliac apophysis ossification²⁷, and curve progression continues beyond Risser stage 4 and 5²⁸. Alternative maturity indices are warranted²⁹.

CVM method’s potential in predicting growth and curve progression in AIS has not been studied. With the increasing availability of EOS imaging in scoliosis centers, low-dose whole spine radiographs are obtained routinely for monitoring. Our results demonstrated that CVM can be reliably determined on lateral EOS radiographs, with adequate anatomical details³⁰.

Correlation of CVM with Other Skeletal Maturity Indicators

CVM was more strongly correlated with indices more effective in indicating growth cessation than Risser staging. Concordance in distribution of CVM with PFMI was evident at latter stages (e.g. 92% of CVM stage 6 corresponds closely to advanced PFMI grades 5-6)—suggest the potential for combining both to avoid mismatch in growth cessation seen in using Risser staging alone.

Growth Rate Variations

CVM 3 was the stage of peak growth. With the ROC analysis cutoff value for PG being equal to or greater than 3.5, clinicians should be made aware that imminent PG is about to occur when the patient is in CVM stage 3. However, no statistically significant differences in growth rates were found among stage 1 to 3 (Appendix 5), also evidenced by the overlap in other maturity indicators. For example, at CVM stage 1, PFMI grades 1, 2, and 3 were simultaneously present substantially (Fig. 2), suggesting that earlier CVM stages may not differentiate growth potential that well in AIS. Ideally, distribution of other maturity indicators within CVM stage 1 should display less heterogeneity. The simultaneous appearance of concavities in the inferior borders of cervical vertebrae C2 and C3 during follow-up may blur stage distinctions³¹.

CVM stage 4, though the most prevalent stage of PG in BH and SH, showed significantly lower growth rates than CVM stages 1 to 3. For the PG occurred at CVM stage 4, the timing was typically closer to CVM stage 3 than to stage 5, highlighting CVM stage 4 as a transitional stage encompassing both PG patients and those decelerating. CVM stage 6, on the other hand, can be regarded as a surrogate end point for growth¹².

Predictive Accuracy for PG and GC

CVM had higher sensitivity and specificity for GC than PG. Since minimizing false positives for GC is crucial to avoid premature bracing discontinuation¹, CVM stages 5 and 6 provide clinically relevant thresholds for growth potential that remains. CVM stage 5 had ∼70% specificity for predicting GC, increasing to >90% specificity at CVM stage 6. A recent machine learning study³² also revealed that CVM 6 had the best precision-recall value, further reinforcing CVM stage 6’s potential in guiding clinical decisions such as brace weaning or discharge from regular growth surveillance.

CVM and Timing of Peak Curve Progression

Peak curve progression occurred between CVM stage 3 and stage 4, lagged behind PG by 6.5 months, similarly found in DRU classification¹. Despite both growth and curve progression followed a significant decreasing trend from CVM stage 4 onward, growth deceleration occurred sooner than curve magnitude deceleration (Fig. 3), nevertheless, highest degree of curve progression and growth rate occurred simultaneously in CVM stage 3. This direct assessment from the cervical spine may correspond more to spinal growth patterns than distal skeletal maturity indices. Notably, peak AS was more consistently identified by DRU classification, conversely peak SH was better indicated by CVM method (Appendix 9).

Limitations and Future Directions

COVID-19–related follow-up delays could have underestimated peak growth and curve progression rates. Defaulted appointment may have led to skeletal maturity advancement without prompt capture of consecutive CVM stages. Moreover, Cobb angle correction is traditionally not applied in AIS height assessment³³, with the rationale that coronal deformity accounts for less than 5 cm of height³⁴, and the final height was not different³⁵. Also, coronal curve magnitude was only calculated from patients who were under observation or before their surgery/bracing commencement. While this eliminated treatment confounders, the period of peak curve progression could be masked in patients coincidentally treated. However, by only including patients presenting at Risser stage 0, we focused on those most at risk of curve progression. Since CVM stage 4 occurs ∼9 months after peak curve progression, future studies could develop bracing protocol incorporating CVM staging toward growth deceleration and cessation.

Conclusion

CVM method demonstrated greater accuracy in identifying GC than PG, with CVM stage 6 reliably indicated GC, while CVM stage 3 marked the onset of PG. Peak curve progression occurred between CVM stage 3 and stage 4. This CVM method provides a framework for assessing both the timing and risk of curve progression, including as patients grow beyond the growth spurt toward growth cessation.

Appendix

Supporting material provided by the authors is posted with the online version of this article as a data supplement at jbjs.org (http://links.lww.com/JBJSOA/A910). This content was not copyedited or verified by JBJS.