Preoperative metabolite ratio improves prediction of postoperative recovery in degenerative cervical myelopathy: a...

Degenerative cervical myelopathy (DCM) is the most common cause of spinal cord dysfunction worldwide. Despite surgical decompression being the standard of care, postoperative recovery remains highly variable. Conventional MRI provides anatomical information yet correlates poorly with functional outcomes. Magnetic resonance spectroscopy (MRS) provides metabolic insights into neuronal integrity, gliosis, and energy metabolism; however, its prognostic utility in DCM remains incompletely validated.

To evaluate the predictive value of preoperative MRS-derived metabolite ratios, alongside clinical and imaging parameters, and to develop and validate a simplified Predictive Recovery Score (PRS) for prognostication. A comparative model analysis was performed to determine the incremental contribution of MRS beyond clinical and structural parameters.

Prospective observational cohort study.

Sixty-nine consecutive patients with cervical myelopathy undergoing surgical decompression were enrolled and followed for 2 years. No patients were lost to follow-up after enrolment; all 69 patients completed the 24-month assessment and were included in the final analysis.

Neurological recovery was assessed using the modified Japanese Orthopaedic Association (mJOA) score and Hirabayashi's recovery rate formula. Patients were stratified into good-recovery (>50%) and poor-recovery (<50%) groups. A secondary analysis using a minimum clinically important difference (MCID ≥2 mJOA points) was also performed.

Demographic, clinical, and imaging data were collected, including comorbidities, stenosis grade, compression ratio, and diffusion tensor imaging (DTI) metrics. Single-voxel MRS at the C2 level quantified various metabolites and metabolite ratios. Group comparisons, correlation analyses, and ROC curves were performed to assess the correlation between the metabolites and recovery rate. Logistic regression identified predictors of recovery. Incremental model performance was assessed by comparing 3 nested models. Internal validation was performed using bootstrap resampling and leave-one-out cross-validation.

Thirty-five patients achieved good recovery, while 34 had poor recovery. Comorbidities were significantly more frequent in the poor recovery group (64.7% vs 34.3%, p=.012). Preoperative mJOA scores were lower (10.91±1.64 vs 14.31±1.51, p=.001), and grade 3 stenosis was more prevalent (82.4% vs 48.6%, p=.003) in poor recovery patients. MRS revealed elevated Cho/NAA (1.28±0.45 vs 0.94±0.26, p=0.001), Cr/NAA (1.26±0.52 vs 0.88±0.21, p=.001), and MIn/NAA (1.33±0.88 vs 0.89±0.37, p=.008) ratios in poor-recovery patients, all of which were inversely correlated with recovery rate. Logistic regression analysis identified 6 parameters as predictors of postoperative recovery: 2 clinical (preop mJOA score, comorbidity), 1 MRI-based parameter (stenosis grade), and 3 metabolite ratios (Cho/NAA, Cr/NAA, and MIn/NAA). A full PRS model incorporating all 6 predictors demonstrated superior discrimination compared with clinical-only and clinical-plus-MRI models (AUCs of 0.74, 0.78, and 0.92, respectively). Bootstrap-corrected AUC was 0.87, and MCID-based secondary analysis yielded AUC 0.89, confirming robustness.

MRS metabolite ratios are significant independent predictors of recovery in DCM. The PRS integrates metabolic, clinical, and structural parameters into a single bedside tool, demonstrating high discriminative accuracy on internal validation. Prospective external validation is required before clinical implementation. The PRS preoperative counselling enables risk stratification and has the potential to support individualized patient counselling and risk stratification in DCM.