139. Quantitative microstructural MRI accurately predicts the likelihood of patients with degenerative cervical myelopathy requiring surgical intervention

Abstract

<h3>BACKGROUND CONTEXT</h3> The definitive management of degenerative cervical myelopathy (DCM) is surgical decompression in moderate to severe cases. The evidence for early surgical intervention in mild DCM patients is less definitive and therefore, current guidelines are less clear on the relative merits of operative vs nonoperative intervention in this group of patients. According to the best available evidence, anywhere from 20%-62% of patients with mild DCM will deteriorate within 6 years; however, there is no reliable way to predict such a deterioration. However, advances in microstructural MRI may afford a novel approach to address this knowledge gap since these imaging techniques allow one to assess pathological changes such as demyelination, scarring and axonal/neuronal loss. <h3>PURPOSE</h3> Using prospectively collected data, we investigate whether quantitative MRI (qMRI) metrics can predict which patients with DCM are likely to deteriorate clinically and require surgery within one year of presentation. <h3>STUDY DESIGN/SETTING</h3> Prospective cohort study. <h3>PATIENT SAMPLE</h3> This study included 58 DCM patients (33 mild, 15 moderate and 10 severe). <h3>OUTCOME MEASURES</h3> Quantitative MRI metrics and mJOA. <h3>METHODS</h3> Data were stratified into patients who required surgery within one year of follow-up and those who remained stable and continued in the nonoperative group. MRI Acquisition and Image Processing: All patients underwent a multiparametric qMRI scan protocol including diffusion tensor (DTI), magnetization transfer (MT) and T2* weighted imaging at the initial visit. Quantitative metrics including cross sectional area (CSA), fractional anisotropy (FA), magnetization transfer ratio (MTR), and T2* white matter to gray matter ratio (T2*WI WM/GM) were measured in the cervical spine at the rostral, caudal and most compressed levels (MCL). Statistical analysis multivariable logistic regression was performed using baseline FA, MTR and T2*WI WM/GM in the rostral spine and CSA at the MCL as predictor variables. Surgical decompression at one year was used as the dependent variable. Model discrimination and reliability were measured using c-index and Brier's score respectively. Validation and calibration were performed using bootstrap method with 100 repetitions. <h3>RESULTS</h3> A total of 33 mild (mJOA 15-17), 15 moderate (mJOA 12-14) and 10 severe (mJOA 0-11) DCM patients were included in the study. Twenty-eight patients had undergone surgical decompresion at One-year follow-up, whereas the other 28 remained in nonoperative managment. Baseline qMRI was able to identify patients requiring surgery, with an accuracy of 81.6% (AUC or c-index). The model showed good reliability with a Brier score of 0.18 and Somers' Dxy rank correlation of 0.63. Calibration using bootstrap method showed overall good predictive ability with slight over prediction at higher observed probability of surgery. <h3>CONCLUSIONS</h3> Baseline quantitative MRI is able to accurately predict the likelihood of deterioration and surgical intervention in DCM patients. It can potentially play an important role in screening patient populations for surgical candidates or those who are at high risk of deterioration. <h3>FDA DEVICE/DRUG STATUS</h3> This abstract does not discuss or include any applicable devices or drugs.