Objective To compare the efficacy of MRI-based cervical vertebral bone quality score (C-VBQ) and CT-based cervical Hounsfield units (C-HU) in predicting cage subsidence after anterior cervical discectomy and fusion (ACDF), and to explore related risk factors. Methods A total of 125 patients (209 surgical segments) who underwent 1- to 3-level ACDF in our institution from January 2019 to December 2023 were retrospectively enrolled. Baseline data, radiographic parameters (C-VBQ, C-HU, T1 slope, cervical Cobb angle, segmental angle, etc.), and surgical information of the patients were collected. Cage subsidence was defined as a loss of surgical segment height ≥ 3 mm (ΔH ≥ 3 mm) at the last follow-up compared with immediately after surgery. Differences between the subsidence group and non-subsidence group were compared. Binary Logistic regression analysis was used to identify independent risk factors for patient-level subsidence, while generalized estimating equations and linear regression were applied to analyze factors related to segment-level subsidence height. The predictive efficacy of C-VBQ and C-HU was evaluated using ROC curves. Results The incidence of patient-level subsidence was 36%. Patients in the subsidence group were older (P < 0.0001), with significantly higher C-VBQ scores (P < 0.0001) and significantly lower C-HU values (P < 0.0001) compared to the non-subsidence group. Multivariate Logistic regression showed that C-VBQ (OR = 1.61) and number of surgical levels (OR = 2.45) were independent predictors of patient-level subsidence, with a model prediction accuracy of 83.2%. C-HU (OR = 0.98) and number of surgical levels (OR = 1.97) were also independent predictors, with a prediction accuracy of 76.8%. ROC curve analysis revealed that the AUC of C-VBQ for predicting subsidence (0.873) was superior to that of C-HU (0.833). Segmental-level analysis showed that the intraoperative traction height of subsided segments was significantly greater than that of non-subsided segments (P = 0.009). Linear regression model indicated that C-VBQ (β = 1.974, P < 0.001) and traction height (β = 0.166, P < 0.001) were significant positive predictors of subsidence height ΔH (R²=0.427). A significant negative correlation was observed between C-VBQ and C-HU (r=-0.800, P < 0.001). Conclusion Both C-VBQ score and C-HU value are significantly associated with cage subsidence after ACDF. C-VBQ is a robust indicator for predicting subsidence after multi-level ACDF, with better independent predictive efficacy and higher specificity than C-HU. An increased number of surgical levels, particularly three-level procedures, significantly elevated the risk of subsidence. Excessively high intraoperative traction height is an independent risk factor for subsidence magnitude. C-VBQ score provides a radiation-free and low-cost method for evaluating bone quality, and when combined with C-HU, it can serve as an important tool for preoperative risk stratification in ACDF.

C- VBQ versus CT Hounsfield units in predicting cage subsidence following ACDF: a retrospective cohort analysis

期刊论文
发表日期：

2025年10月11日

查看全文

Yuanpeng Li,
Lu Chen,
Sinuo Shen,
Hang Shi,
Zanli Jiang,
Lei Zhu

摘要

Objective

To compare the efficacy of MRI-based cervical vertebral bone quality score (C-VBQ) and CT-based cervical Hounsfield units (C-HU) in predicting cage subsidence after anterior cervical discectomy and fusion (ACDF), and to explore related risk factors.

Methods

A total of 125 patients (209 surgical segments) who underwent 1- to 3-level ACDF in our institution from January 2019 to December 2023 were retrospectively enrolled. Baseline data, radiographic parameters (C-VBQ, C-HU, T1 slope, cervical Cobb angle, segmental angle, etc.), and surgical information of the patients were collected. Cage subsidence was defined as a loss of surgical segment height ≥ 3 mm (ΔH ≥ 3 mm) at the last follow-up compared with immediately after surgery. Differences between the subsidence group and non-subsidence group were compared. Binary Logistic regression analysis was used to identify independent risk factors for patient-level subsidence, while generalized estimating equations and linear regression were applied to analyze factors related to segment-level subsidence height. The predictive efficacy of C-VBQ and C-HU was evaluated using ROC curves.

Results

The incidence of patient-level subsidence was 36%. Patients in the subsidence group were older (P < 0.0001), with significantly higher C-VBQ scores (P < 0.0001) and significantly lower C-HU values (P < 0.0001) compared to the non-subsidence group. Multivariate Logistic regression showed that C-VBQ (OR = 1.61) and number of surgical levels (OR = 2.45) were independent predictors of patient-level subsidence, with a model prediction accuracy of 83.2%. C-HU (OR = 0.98) and number of surgical levels (OR = 1.97) were also independent predictors, with a prediction accuracy of 76.8%. ROC curve analysis revealed that the AUC of C-VBQ for predicting subsidence (0.873) was superior to that of C-HU (0.833). Segmental-level analysis showed that the intraoperative traction height of subsided segments was significantly greater than that of non-subsided segments (P = 0.009). Linear regression model indicated that C-VBQ (β = 1.974, P < 0.001) and traction height (β = 0.166, P < 0.001) were significant positive predictors of subsidence height ΔH (R²=0.427). A significant negative correlation was observed between C-VBQ and C-HU (r=-0.800, P < 0.001).

Conclusion

Both C-VBQ score and C-HU value are significantly associated with cage subsidence after ACDF. C-VBQ is a robust indicator for predicting subsidence after multi-level ACDF, with better independent predictive efficacy and higher specificity than C-HU. An increased number of surgical levels, particularly three-level procedures, significantly elevated the risk of subsidence. Excessively high intraoperative traction height is an independent risk factor for subsidence magnitude. C-VBQ score provides a radiation-free and low-cost method for evaluating bone quality, and when combined with C-HU, it can serve as an important tool for preoperative risk stratification in ACDF.