Comparison of predictive performance for cage subsidence between CT-based Hounsfield units and MRI-based vertebral bone quality score following oblique lumbar interbody fusion

Objective To compare the predictive performance between CT-based Hounsfield units (HU) and MRI-based vertebral bone quality (VBQ) for cage subsidence (CS) following oblique lumbar interbody fusion combined with anterolateral single-rod screw fixation (OLIF-AF). Methods A retrospective study was performed on consecutive patients who underwent OLIF-AF at our institution from 2018 to 2020. CS was determined by CT according to the change in the midpoint intervertebral space height. The VBQ score and HU value were measured from preoperative MRI and CT, respectively. Then, we evaluated the predictive performance of those two parameters by comparing the receiver operating characteristic (ROC) curves. Results The mean global and segmental VBQ scores were significantly higher in the CS group, and the mean global and segmental HU values were significantly lower in the CS group. The area under the curve (AUC) of CS prediction was higher in the operative segments’ VBQ score and HU value than the measurement in the global lumbar spine. Finally, the combined segmental VBQ score and segmental HU value demonstrated the highest AUC. Conclusion Both MRI-based VBQ score and CT-based HU value can achieve accurate CS prediction. Moreover, the combination of those two measurements indicated the best predictive performance. Clinical relevance statement Both MRI-based VBQ score and CT-based HU value can be used for cage subsidence prediction, in order to take preventive measures early enough. Key Points • Osteoporosis is a risk factor for CS, both MRI-based VBQ score and CT-based HU value are important predictors during vertebral bone quality evaluation. • The VBQ score and HU value measured in the operative segments are better predictors of CS than the measurement in the global lumbar spine. • Combined segmental VBQ score and segmental HU value achieved the best predictive performance for CS.