A novel method of C-VBQ score and CT-HU as risk predictors for cage subsidence after short level ACDF

Abstract

We used a novel technique to measure subsidence after short-level anterior cervical discectomy and fusion (ACDF) with plate and screw fixation. And we aimed to evaluate cage subsidence by using cervical vertebral bone quality (C-VBQ) score and CT-HU. Quantitative assessment of vertebral bone quality was performed through two complementary methods: C-VBQ scoring and CT Hounsfield unit measurements, with their correlation established using Pearson’s analysis. The research methodology incorporated multiple statistical techniques, beginning with linear regression to examine demographic influences on C-VBQ values. For subsidence risk evaluation, all imaging datasets were rigorously compiled and subjected to multivariate regression analysis to identify significant contributing factors. The predictive validity of these factors was subsequently verified through ROC curve analysis, providing a robust assessment of their clinical utility. 112 patients who underwent short-level (≤ 2) ACDF were enrolled in this study, and 29 patients (25.89%) with cage subsidence were found. The subsidence group demonstrated significantly higher C-VBQ values (3.10 ± 0.22) compared to the non-subsidence cohort (2.70 ± 0.36, P < 0.001). Multivariate analysis identified the C-VBQ score as the sole independent predictor of cage subsidence following ACDF, achieving an 83.9% predictive accuracy. CT-HU value exhibited a significant inverse relationship with C-VBQ scores (P < 0.001). Univariate regression analysis identified age as the only parameter significantly correlated with C-VBQ values (P < 0.001). The inverse relationship between C-VBQ and CT-HU suggests its validity as a bone quality measure. Importantly, C-VBQ proved clinically valuable for identifying subsidence risk in 1–2 level ACDF procedures.

Introduction

Anterior cervical discectomy and fusion (ACDF) has been well-established as an effective surgical intervention for restoring physiological curvature and alleviating spinal cord compression of degenerative cervical diseases¹. However, ACDF procedures involving internal fixation systems are associated with several implant-related complications, most notably screw or plate loosening, hardware fracture, and cage subsidence (CS)². Reported incidence rates of CS in the literature vary considerably, ranging from 9% to 79.7% across different study populations³.

CS significantly influenced biomechanics and clinical efficacy resulting in intervertebral height reduction, progressive cervical kyphosis, and spinal cord compression, potentially causing recurrence of neurological symptoms^2,4–7. Studies have explored the fact that internal fixation system failure, CS, and pseudarthrosis development have emerged as leading indications for revision surgery following ACDF procedures^7–10. Current clinical practice reveals significant heterogeneity in CS assessment methodologies following spinal fusion procedures. The absence of standardized measurement protocols has led to diverse evaluation approaches, ranging from intervertebral height quantification to fusion segment analysis and implant position assessment, each employing distinct diagnostic thresholds¹¹. While technical factors such as imaging parameters may introduce measurement variability, clinical evidence consistently identifies compromised bone quality—particularly diminished bone mineral density (BMD) and established osteoporosis—as the predominant determinants of subsidence risk¹².

Despite Dual-energy X-ray absorptiometry’s (DEXA) represented the most important role in BMD assessment, its practical implementation for cervical spine evaluation remains challenging in routine clinical practice¹³. Quantitative computed tomography (QCT) provides accurate volumetric BMD measurements, offering three-dimensional assessment of variations in bone architecture, but its clinical applicability is limited due to equipment and costs¹⁴. Preoperative cervical CT Hounsfield units (CT-HU) measurements emerged as a significant tool to assess osteoporosis and a independent predictor of postoperative CS in ACDF patients, suggesting its potential utility in surgical planning¹⁵. In response to radiation safety considerations and the predominance of MRI in spinal diagnostics, the vertebral bone quality score (VBQ) had developed by Ehersman et al. as a promising radiation-free alternative for bone quality assessment¹⁶. The work of Soliman et al. established cervical VBQ (C-VBQ) scoring as a clinically useful preoperative predictor for CS in single-level ACDF procedures¹⁷. However, few studies have used VBQ scores to evaluate CS after short-segment ACDF or performed comprehensive measurements beyond single-level subsidence. This study aimed to apply a novel evaluation integrating radiographic assessment to investigate the clinical utility of the C-VBQ score in predicting CS following short-level ACDF.

Materials and methods

Patients population

Consecutive patients treated by using ACDF for short-level degenerative cervical diseases in our center from October 2020 to March 2023 retrospectively. The postoperative observation time was at least 12 months. Ethical approval was granted by the local Institutional Review Board of Tianjin Medical University General Hospital. The research conformed to the principles of the Declaration of Helsinki, with documented informed consent was obtained from all subjects and/or their legal guardian(s).

The inclusion criteria were: (1) Participants received instrumented short level (≤ 2) ACDF for degenerative symptomatic cervical conditions (including myelopathy or radiculopathy) persisting despite appropriate non-operative therapies; (2) Complete imaging data before and after surgery: plain X-ray pictures, CT scanning, magnetic resonance imaging (MRI); (3) all cases maintained follow-up for at least 12 months after surgery. Conversely, the exclusion criteria were: 1) patients with cervical spine trauma, neoplastic lesions, infectious pathologies, congenital malformations, or systemic disorders (including rheumatoid arthritis and neurodegenerative diseases) that could compromise surgical outcomes; (2) prior cervical spine surgical interventions; (3) insufficient follow-up data or documented history of substance use disorders.

Data collection

Demographic and clinical data were collected, including age, sex, body mass index (BMI), smoking status, alcohol consumption history, comorbidities (diabetes mellitus and hypertension), operative details (duration and surgical levels), and follow-up duration. All patients were followed at regular intervals with clinical and radiographic evaluation. Preoperative and postoperative imaging parameters were systematically evaluated using our institution’s picture archiving and communication system (PACS). Preoperative clinical evaluations were systematically performed using standardized metrics: Visual Analogue Scale (VAS) for pain intensity, modified Japanese Orthopaedic Association (mJOA) score for myelopathy severity, and Neck Disability Index (NDI) for functional impairment.

C-VBQ score calculation

The C-VBQ scoring methodology was adapted from the technique described by Soliman et al.¹⁷ as illustrated in Fig. 1). The C-VBQ score was calculated by dividing the median signal intensity (SI) of C3-C6 vertebral bodies by the CSF SI (Formula 1).

1

Fig. 1.

The cervical T1-MRI region of interest used to calculate VBQ score.

All C-VBQ measurements were conducted using institutional PACS software by two independent, trained spinal surgeons who do not interfere with each other and blinded to the clinical data. Inter-rater discrepancies exceeding 10% triggered third-observer adjudication, with final values determined by averaging all concordant measurements for subsequent analysis.

CT-HU value calculation

CT-HU values were calculated per Schreiber et al.‘s validated methodology in Fig. 2¹⁸. The mean HU value across these four levels was calculated to enable precise comparison with corresponding C-VBQ scores. (Formula 2).

Fig. 2.

The CT of cervical spine used to calculate the HU value.

2

Cage subsidence calculation

Cage subsidence measurements were performed using the method previously reported by Li et al.³ Operational segment height (OSH) was quantified on neutral lateral radiographs as the vertical distance between the midpoints within the fusion. Comparative measurements were obtained from non-operated segments (NOS) between post-operation and final follow-up. (Fig. 3). Subsidence was defined when the proportion was ≥ 3 mm between the postoperative and final follow-up images or when the cage was obviously to penetrate the vertebral endplate. The formula for cage subsidence is Formula 3.

Fig. 3.

The systemic method used to calculate cage subsidence. (A) The operational segment height and non-operated segment height after ACDF at post-operational time in one interval. (B) The operational segment height and non-operated segment height after ACDF at last follow-up in one interval. The single interval of subsidence value was 0.42 cm.(C).The operational segment height and non-operated segment height after ACDF at post-operational time in two intervals. (D) The operational segment height and non-operated segment height after ACDF at at last follow-up in two intervals. The two intervals of subsidence value were 0.32 cm.

3

Surgical technique

The surgical strategy employed a consistent right anterior cervical approach performed by an experienced surgical team. Initial exposure was followed by intervertebral space distraction using Caspar distractors, facilitating thorough discectomy and posterior longitudinal ligament removal. Particular attention was given to endplate preparation, where cartilage was carefully prepare while preserving the structural integrity of the bony endplates. After determining optimal implant dimensions through trial sizing, allograft-filled cages (Medtronic Sofamor Danek, Memphis, TN, USA) were precisely positioned and secured with anterior plating systems from the same manufacturer. Postoperative care included a 4-week cervical collar immobilization period with progressive activity reintroduction under medical supervision.

Statistical analyses

All statistical procedures were conducted in SPSS (version 26; IBM Corp., Armonk, NY) with a predetermined significance threshold of P < 0.05 (two-tailed). Data characterization followed conventional approaches: continuous variables as mean ± SD (analyzed via Student’s t-test) and categorical variables as frequency distributions (assessed using χ² or Fisher’s exact tests). Analytical methods were stratified by data type and research objectives: initial univariate linear regression screened demographic associations with C-VBQ scores, while multivariate logistic regression incorporating comprehensive clinical variables identified independent predictors of cage subsidence. For imaging parameter correlations, Pearson’s coefficient quantified the C-VBQ/CT-HU relationship. Diagnostic performance evaluation involved ROC curve analysis with Youden index-derived optimal cutoffs, ensuring rigorous assessment of model discrimination capacity.

Results

Patient demographic characteristics

This retrospective analysis evaluated 112 consecutive patients (60 male, 52 female) undergoing cervical spine surgery, stratified by postoperative cage subsidence status. Demographic comparisons revealed comparable baseline characteristics between the subsidence (n = 29) and non-subsidence (n = 83) cohorts regarding sex distribution, comorbidities, surgical parameters, and preoperative clinical scores (all p > 0.05, Table 1). However, the subsidence group demonstrated significantly older age (60.93 ± 8.69 vs. 52.31 ± 11.42 years, P < 0.001) and distinct imaging biomarkers. Quantitative analysis showed markedly elevated C-VBQ scores (3.10 ± 0.22 vs. 2.70 ± 0.36, P < 0.001) and reduced CT-HU values (316.5 ± 27.09 vs. 413.45 ± 27.96 HU, P < 0.001) in patients. (Table 1.)

Table 1.

Clinical characteristics of anterior cervical discectomy and fusion patients.

Patient characteristics	Subsidence group(n = 29)	Non-Subsidence group(n = 83)	P-value
Age	60.93 ± 8.69	52.31 ± 11.42	<0.001
Males	14(23.3%)	46(76.7%)	0.507
Females	15(28.8%)	37(71.2%)	0.507
BMI	24.38 ± 2.75	25.14 ± 3.73	0.319
Smoking	9(31.0%)	16(19.3)	0.191
Alcoholism	6(20.7%)	10(12.0%)	0.354a
Hypertension	6(20.7%)	21(25.3)	0.617
Diabetes	2(6.9%)	8(9.6%)	1.000a
Involved segments
Single-level	19(27.5%)	50(72.5%)	0.615
Non-single-level	10(23.3%)	33(76.7%)	0.615
C-VBQ	3.10 ± 0.22	2.70 ± 0.36	<0.001
CT-HU	316.53 ± 27.09	413.45 ± 27.96	<0.001
VAS	5.72 ± 1.58	5.81 ± 1.46	0.797
mJOA	9.41 ± 1.57	9.35 ± 1.66	0.856
NDI	29.21 ± 4.02	28.20 ± 4.67	0.306
Blood loss	83.45 ± 27.39	83.39 ± 24.18	0.991
Follow-up	15.72 ± 2.36	16.24 ± 2.70	0.362

BMI: Body Mass Index; C-VBQ: Cervical Vertebral Bone Quality; CT-HU: Computed Tomography Hounsfield Units; VAS: Visual Analogue Scale; mJOA: Modified Japanese Orthopaedic Association; NDI: Neck Disability Index; a:Fisher’s exact test.

The "bold" values represented the P value < 0.05.

C-VBQ score-a cage subsidence predictive way

To elucidate key determinants of postoperative cage subsidence in ACDF procedures, we performed comprehensive multivariate analysis of preoperative parameters. The logistic regression model identified C-VBQ score as the sole independent risk factor, demonstrating a robust association with subsidence risk (odds ratio[OR] = 26.254, 95% CI=[3.93-175.28], P = 0.001), as detailed in Table 2. Subsequent linear regression analysis further validated this relationship, with C-VBQ scores showing moderate but statistically significant positive correlation with subsidence severity (Pearson’s r = 0.38, P < 0.001; Fig. 4). These consistent findings across different analytical approaches underscore the pivotal role of vertebral bone quality assessment in predicting surgical outcomes.

Table 2.

Multivariate logistic regression analysis for potential risk factors of subsidence after anterior cervical discectomy and Fusion.

Variable	Odds ratio	95% CI	P-value
Age	1.045	0.975–1.120	0.211
Males	4.036	0.954–17.073	0.058
BMI	0.899	0.737–1.098	0.297
Smoking	2.633	0.543–12.766	0.229
Alcoholism	4.952	0.811–30.253	0.083
Hypertension	0.475	0.104–2.159	0.335
Diabetes	1.726	0.216–13.797	0.607
Single-level	0.193	0.018–2.111	0.178
C-VBQ	26.254	3.930−175.282	0.001
VAS	0.837	0.559–1.252	0.386
mJOA	1.036	0.732–1.466	0.844
NDI	1.074	0.944–1.222	0.279
Blood loss	1.020	0.976–1.065	0.381
Follow-up	0.962	0.770–1.202	0.736

BMI: Body Mass Index; C-VBQ: Cervical Vertebral Bone Quality; VAS: Visual Analogue Scale; mJOA: Modified Japanese Orthopaedic Association; NDI: Neck Disability Index.

The "bold" values represented the P value < 0.05.

Fig. 4.

C-VBQ scores showed significant positive correlation with subsidence severity.

C- VBQ score correlation with the CT-HU and age

Correlational analyses revealed two clinically important relationships regarding vertebral bone quality assessment. First, an inverse association was observed between C-VBQ scores and CT Hounsfield units (Pearson’s r =-0.43, P < 0.001), as visually represented in Fig. 5. This moderate negative correlation suggests that higher C-VBQ values, indicative of poorer bone quality, consistently corresponded with lower CT-HU measurements. Second, univariate regression demonstrated that advancing age showed a stronger positive association with elevated C-VBQ scores (r = 0.531, P < 0.001), with detailed results presented in Table 3. These findings collectively highlight that both imaging parameters and demographic factors significantly influence vertebral bone quality assessments.

Fig. 5.

C-VBQ scores showed significant negative correlation with CT-HU.

Table 3.

Univariate linear regression analysis between vertebral bone quality score and patient-related characteristic.

Variable	β	Standard error	P-value
Age	0.018	0.003	<0.001
BMI	−0.004	0.009	0.628
VAS	−0.015	0.021	0.476
mJOA	−0.016	0.019	0.389
NDI	0.008	0.007	0.214
Blood loss	0.002	0.001	0.071
Follow-up	−0.009	0.012	0.440

BMI: Body Mass Index; VAS: Visual Analogue Scale; mJOA: Modified Japanese Orthopaedic Association; NDI: Neck Disability Index.

The "bold" values represented the P value < 0.05.

Predictive value assessment

The predictive performance of the C-VBQ scoring system was rigorously assessed through receiver operating characteristic (ROC) analysis. Demonstrating excellent diagnostic capability, the model achieved an area under the curve of 0.839 (95% CI = 0.768–0.911), as illustrated in Fig. 6. Optimal threshold determination using the Youden index yielded a C-VBQ cutoff of 2.80, which balanced high sensitivity (96.6%) for detecting subsidence risk with reasonable specificity (69.9%). These robust metrics suggest the clinical utility of this threshold for preoperative risk stratification in cervical fusion procedures.

Fig. 6.

(A) The ROC curves illustrate that the VBQ scores areas under AUC was 0.839. (B) The ROC curves illustrate that the CT-HU areas under AUC was 0.927.

Discussion

ACDF has employed a frequent surgical procedure for addressing degenerative cervical myelopathy and radiculopathy proposed by Smith and Cloward^1,19. It can decompress the compressive tissues and supply stability after fusion through the cervical vertebral body by using cages²⁰. However, cage subsidence represents one of the most prevalent implant-related complications following ACDF, potentially leading to segmental height loss and subsequent adverse outcomes including kyphotic deformity and hardware failure^4,21. In this study, we introduced a novel scoring approach using cervical spine MRI, which has been shown to significantly forecast cage subsidence after ACDF surgery. Moreover, a new systemic method is used to measure cage subsidence, which avoids errors from imaging scales and equipment differences.

According to the results of earlier research, several factors correlate with cage subsidence after ACDF, such as lower cervical spine and multi-level operating segments, incorrect size of cage and over-distraction^5,17,22–24. In addition, the extent of osteoporosis is also a recognized element that influences the occurrence of cage subsidence^{3,7,17,23–25}. In this study, age, C-VBQ, and CT-HU were significantly different between the two groups, but no notable differences were found in the surgical segments. Patients undergoing ACDF are typically older and may experience reduced bone mineral density. At present, DEXA imaging is considered the gold-standard technique to evaluate bone density^3,17,26, which may not accurately reflect bone quality in the cervical spine due to regional variations in trabecular architecture²⁷. Consequently, DEXA has limited clinical utility for cervical spine assessment due to fundamental anatomical and technical constraints.

Another technique for measuring bone mineral density is QCT. The outcomes from QCT are considered more reliable than those from DEXA T-scores because they can eliminate the superimposition effects that DEXA struggles with^17,28,29. However, the expense and significant radiation exposure from QCT restrict its application. Recent studies have increasingly demonstrated the potential of opportunistic computed tomography as a screening method for osteopenia and osteoporosis^17,18,30. However, routine preoperative assessment rarely includes these specialized diagnostic evaluations in current clinical practice^7,29. A new method needed to be developed by the researchers to utilize MRI in assessing cage subsidence.

The evolution of vertebral bone quality assessment through MRI has progressed through several key developments. Initially introduced for lumbar spine evaluation by Ehresman et al., the VBQ scoring system derived from T1-weighted MRI has gained validation through numerous studies confirming its strong correlation with DEXA measurements in osteoporosis diagnosis³¹. Subsequent research has extended these findings to investigate associations between VBQ scores and various osteoporosis-related complications, including implant failure and revision surgery risks^{3,6,16,17,32–38}. Building upon these foundation, Soliman et al. pioneered the cervical adaptation (C-VBQ) specifically for ACDF outcome prediction¹⁷. While these scoring systems show considerable promise, methodological refinements remain necessary - particularly regarding population diversity and subsidence measurement precision. Our contribution addresses these limitations through an innovative measurement approach that standardizes assessments across imaging platforms, thereby enhancing consistency and clinical applicability.

Several limitations should be taken into account in relation to our study. Firstly, the data were collected retrospectively, which means that not all patients had follow-up images available. Small sample sizes and single-center designs restrict the universe of our research findings. Further external validation studies are required to verify the efficacy of this novel system in predicting cage subsidence. Our analysis identified the C-VBQ score as a significant independent predictor of cage subsidence following ACDF, demonstrating strong correlation with CT Hounsfield units. However, the absence of preoperative DEXA measurements and laboratory biomarkers in our dataset necessitates further validation to comprehensively evaluate C-VBQ’s relationship with systemic bone metabolism.

Conclusion

Our findings establish the C-VBQ scoring system as a dual-purpose clinical tool with important applications in anterior cervical surgery. The metric demonstrated robust validity as a preoperative bone quality assessment, while simultaneously serving as a powerful prognostic indicator for postoperative cage subsidence. Notably, the observed correlation patterns - direct association with subsidence incidence and inverse relationship with CT Hounsfield units - provide compelling evidence for the scoring system’s construct validity. These consistent directional relationships across multiple analytical approaches reinforce the clinical relevance of C-VBQ assessment in surgical planning and risk stratification.

Author contributions

Qiang Zhang wrote this manuscript. Zeque Zhang and Rongzhi Ma collected data. Yuan Xue revised the manuscript. All authors reviewed the manuscript.

Data availability

The datasets used and analyzed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.