Comparison of Predictive Performance for Pedicle Screw Loosening Between Computed Tomography-Based Hounsfield Units and Magnetic Resonance Imaging-Based Vertebral Bone Quality Score After Lumbar Surgery.

Abstract

To compare predictive performance for pedicle screw loosening between computed tomography (CT)-based Hounsfield units (HU) and magnetic resonance imaging (MRI)-based vertebral bone quality score (VBQ) after lumbar surgery. A retrospective study was conducted on patients who received transforaminal lumbar interbody fusion continuously at our institution from May 2018 to September 2020. On the basis of 12 months' follow-up lumbar radiographs, screw loosening was defined as a clear zone of minimal thickness of ≥1 mm around the pedicle screw on radiography. VBQ score and HU value were measured using preoperative MRI and CT, respectively. Then, we evaluated the predictive performance of these 2 parameters by comparing the receiver operating characteristic curve. In all patients, area under the curve (AUC) of the VBQ score (AUC = 0.752; 95% confidence interval [CI] 0.663-0.841; P < 0.001) was larger than those of the CT HU value (AUC = 0.652; 95% CI 0.558-0.746; P = 0.005), but there was no significant difference between them (P_AUC = 0.076). In patients with lumbar spinal stenosis, AUC of VBQ score (AUC = 0.863; 95% CI 0.764-0.961; P < 0.001) was larger than those of the CT HU value (AUC = 0.673; 95% CI 0.513-0.833; P = 0.043), with significant difference (P_AUC = 0.003). MRI-based VBQ score and CT-based HU value have similar performance in predicting pedicle screw loosening after lumbar surgery. Furthermore, in patients with lumbar spinal stenosis, VBQ score demonstrated better predictive ability than HU value.