A novel modified VBQ Score,Correlation with the Vertebral Bone Quality Score, independently assessing for bone quality in patients with Osteoporosis Vertebral Fracture

Yuxiang Wu, Jinhan Guo, Houjun Zhang, Qiang Wang, Lin Xu, Xiaohong Mu, Tianlin Wen, Huancai Lin, X. Ke, Haipeng Chen, Leqin Xu, Weiwei Ma, Bushuang Li, Qiwei ZhangResearch Square 2024 · DOI

Abstract

Traditional methods for assessing bone quality, such as Dual-Energy X-ray Absorptiometry (DEXA), often face limitations due to degenerative spinal changes. While the Vertebral Bone Quality (VBQ) score derived from MRI has emerged as a promising alternative, its requirement for specific lumbar imaging (L1–L4) limits its clinical utility in cases of thoracic fractures. This study introduces and validates a modified VBQ (m-VBQ) score that utilizes vertebrae adjacent to any site of interest. Our results demonstrate that the m-VBQ score strongly correlates with the standard VBQ (r=0.842) and maintains comparable predictive accuracy for osteoporotic vertebral fractures (OVF) and osteoporosis (OP), offering a more flexible and clinically operable tool for bone health assessment.

Introduction

Osteoporosis is a systemic skeletal disorder characterized by decreased bone mass and microarchitectural deterioration, frequently leading to osteoporotic vertebral fractures (OVF). Early and accurate assessment of bone quality is critical for fracture prevention and surgical planning. While DEXA remains the clinical gold standard, its accuracy is often compromised in elderly patients by osteophytes, vascular calcification, and other degenerative changes that artificially inflate T-scores.

To address these shortcomings, the Vertebral Bone Quality (VBQ) score was developed. By measuring the signal intensity of T1-weighted MRI—a proxy for fat infiltration in bone marrow—the VBQ score provides a non-invasive metric for bone fragility. However, the standard VBQ calculation is rigid, requiring the average signal intensity of the L1 through L4 vertebrae. In clinical practice, patients with thoracic fractures may only receive thoracic imaging, necessitating additional, costly lumbar scans solely for VBQ calculation. This study evaluates the m-VBQ score, a localized adaptation designed to overcome these geographic constraints.

Methodology

The researchers conducted a retrospective analysis of 206 adult patients (103 with OVF and 103 controls) treated at the Xiamen Hospital of Beijing University of Chinese Medicine between 2017 and 2022. The m-VBQ score was calculated using a "local neighborhood" approach centered on a specific vertebral scoring site (either a fractured or healthy vertebra).

The calculation involves two primary steps:

  1. Signal Intensity Measurement: The Median Signal Intensity (MSI) is calculated for the two vertebrae immediately above and the two vertebrae immediately below the chosen scoring site.
  2. Normalization: To account for inter-scanner variability, this MSI is divided by the signal intensity of the cerebrospinal fluid (CSF) within the spinal canal at the level immediately below the scoring site.
The formula for calculating the modified Vertebral Bone Quality (m-VBQ) score.

This methodology allows clinicians to derive a bone quality metric from any segment of the spine that has been imaged, eliminating the dependency on a full lumbar series.

Results and Statistical Validation

The study found a significant difference in m-VBQ scores between the fracture group (3.88 ± 0.86) and the non-fracture group (3.07 ± 0.59, P < 0.001). Statistical analysis confirmed the robustness of the modified metric:

  • Correlation with Standard VBQ: Pearson correlation analysis revealed a strong positive association (r=0.842, P<0.001) between the m-VBQ and standard VBQ scores. The consistency was further validated by an Intra-class Correlation Coefficient (ICC) of 0.841.
  • The "L2 Sweet Spot": Interestingly, the correlation between the two scores peaked when the scoring site was at the L2 level (r=0.951, ICC=0.950). As the scoring site moved further from L2, the correlation slightly decreased, though it remained statistically significant.
  • Predictive Performance: Receiver Operating Characteristic (ROC) analysis showed that the m-VBQ score is a reliable predictor of bone health. For OVF prediction, the m-VBQ achieved an Area Under the Curve (AUC) of 0.776, nearly identical to the standard VBQ’s 0.780. For osteoporosis detection, the m-VBQ (AUC 0.737) slightly outperformed the standard VBQ (AUC 0.730).
  • Reliability: Inter-rater reliability (ICC = 0.896) and intra-rater reliability (ICC = 0.81) were both excellent, indicating that the score is stable across different clinical observers.

Discussion

The primary insight of this research is that bone marrow fat infiltration—the biological basis for the VBQ score—is a systemic phenomenon that can be accurately sampled locally. The "local neighborhood" approach of the m-VBQ score effectively captures the bone quality of the surrounding environment, making it a highly practical tool for real-world clinical scenarios where imaging may be incomplete or focused on non-lumbar segments.

The discovery of the "L2 sweet spot" suggests that the mid-lumbar region may be the most representative area for overall spinal bone quality. However, the high AUC values for the m-VBQ across various segments (particularly its superior performance in the T3-T11 subgroup compared to standard VBQ) highlight its versatility. Furthermore, by assessing the vertebrae immediately adjacent to a fracture, clinicians may better predict the risk of subsequent "neighboring" fractures, a common complication in osteoporotic patients.

Conclusion and Future Directions

The m-VBQ score provides a flexible, reliable, and clinically accessible alternative to the standard VBQ score. By removing the requirement for specific lumbar imaging, it reduces patient costs and diagnostic delays. While the study notes challenges in measuring CSF signal intensity in the narrower thoracic canal, the overall accuracy remains high.

Future work should focus on automating the measurement process. Currently, the manual placement of "Regions of Interest" (ROIs) on MRI slices is time-consuming. The development of deep learning algorithms to automatically segment vertebrae and calculate m-VBQ scores could integrate this metric directly into clinical workflows, providing instantaneous bone quality assessments during routine spinal evaluations.

Summary updated

Abstract

<title>Abstract</title> Background :The VBQ Score, a formula based on magnetic resonance imaging (MRI) parameters, facilitates the assessment of bone quality in patients with osteoporotic vertebral fractures, offering substantial clinical convenience. However, the formula necessitates lumbar MRI imaging for data collection and evaluation, leading to practical challenges in clinical scenarios. Specifically, in the case of a thoracic vertebral fracture, additional lumbar spine MRI is necessary to obtain pertinent data, or when incomplete lumbar spine MRI images exist, the measurement of relevant data becomes challenging. Purpose: This study introduces a m-VBQ score and compares it with the standard VBQ score, investigating the feasibility and accuracy of the improved VBQ score for evaluating bone quality. Methods: A retrospective study was conducted, involving adult patients from January 2017 to December 2022. The study included a patient sample comprising 103 individuals with osteoporotic vertebral fractures (OVF) and 103 without OVF. METHODS: The collected data encompassed non-contrast T1-weighted MRI, DEXA, and demographic information from both the fracture group and the non-fracture group patients. The m-VBQ score was established by selecting the vertebral scoring site, which could be either the vertebral body of a single-segment fractured vertebra or the vertebral body of a single-segment non-fractured vertebra. This involved calculating the median signal intensity (MSI) of the two vertebrae above and the two vertebrae below the vertebral scoring site, followed by dividing this value by the signal intensity of the cerebrospinal fluid (CSF) in the vertebra immediately below the vertebral scoring site. The differences between the two m-VBQ score groups were evaluated using the Student's t-test. Additionally, the relationship between the m-VBQ score and VBQ score was examined through Pearson correlation analysis, univariate linear regression, and intra-group correlation analysis. Binary logistic regression was utilized to investigate the association of the VBQ score, m-VBQ score, and occurrence of OVF. Receiver Operating Characteristic (ROC) analysis was performed to determine the Area Under the Curve (AUC) for both scores as predictors of OVF and osteoporosis (OP). Furthermore, subgroup analyses were conducted based on the vertebral score location, with statistical significance determined at P&lt;0.05. Results:There was a significant difference in m-VBQ scores between the fracture and non-fracture groups (3.88±0.86 vs. 3.07±0.59, P &lt; 0.001). Pearson correlation analysis revealed a strong positive association between m-VBQ scores and VBQ scores (r=0.842, P&lt;0.001). As the scoring vertebral site approached L2 from above, the Pearson correlation coefficient increased, reaching its peak at L2 (r=0.951, P&lt;0.001) and subsequently decreasing for sites below. Linear regression analysis between m-VBQ and VBQ scores indicated that as the scoring vertebral site approaches L2 from above, the slope of the equation might increase, with the maximum slope observed at L2 (β1=0.9757), and then potentially decrease for lower sites. The consistency between VBQ scores and m-VBQ scores is excellent (ICC = 0.841). As the vertebral scoring site of the m-VBQ score progresses downward, nearing L2, the ICC coefficient potentially increases, reaching its peak at L2 (ICC = 0.950). Beyond this point, the ICC coefficient might decrease.Binary logistic regression highlighted that m-VBQ scores (OR=1.253-5.282, 95%CI:1.210-5.496, P=0.010) had a stronger significance in predicting OVF compared to VBQ scores (OR=2.579, 95%CI:1.210-5.496, P=0.014). Regarding OVF prediction, the AUC for m-VBQ was 0.776 (95%CI:0.713-0.838, P&lt;0.001), indicating good predictive capability, similar to VBQ. For specific vertebral sites, m-VBQ's predictive ability was superior to VBQ for the T3-T11 subgroup, equivalent for the L3-L5 subgroup, and inferior for the T12, L1, and L2 subgroups. In predicting osteoporosis (OP), the AUC for m-VBQ was 0.737 (95%CI:0.669-0.805, P&lt;0.001), suggesting a stronger predictive capacity compared to VBQ. Inter-rater reliability for m-VBQ scores among two reviewers yielded an ICC of 0.896 (95%CI 0.861-0.963, P&lt;0.001), and intra-rater reliability resulted in an ICC of 0.81 (95% CI 0.759-0.853, P&lt;0.001). Conclusion:The VBQ score and the m-VBQ score, derived from the adjacent vertebrae of OVF patients, are both effective indices for assessing bone quality in OVF patients. The m-VBQ score exhibits good predictive capability for OVF and OP, closely mirroring the VBQ score's predictive power. Compared to the conventional VBQ score, we believe the m-VBQ score offers more flexibility and clinical operability while maintaining similar reliability for evaluating OP and OVF. Furthermore, it enables the assessment of bone quality in vertebrae adjacent to fractures, potentially predicting subsequent fractures in neighboring vertebrae.
BibTeX
@inproceedings{wu2024novel,
  author = {Yuxiang Wu and Jinhan Guo and Houjun Zhang and Qiang Wang and Lin Xu and Xiaohong Mu and Tianlin Wen and Huancai Lin and X. Ke and Haipeng Chen and Leqin Xu and Weiwei Ma and Bushuang Li and Qiwei Zhang},
  title = {{A novel modified VBQ Score,Correlation with the Vertebral Bone Quality Score, independently assessing for bone quality in patients with Osteoporosis Vertebral Fracture}},
  booktitle = {Research Square},
  year = {2024},
  doi = {10.21203/rs.3.rs-4695580/v1},
}
Figure