451: Clinical Variables Outperform MEPs in Predicting Motor Recovery After Spinal Cord Injury

Topic: Clinical Science
Subtopic: Pain category (ICD-11), Chronic primary musculoskeletal pain, Artificial Intelligence/Machine Learning, Electrophysiology/Clinical neurophysiology (e.g. microneurography, EEG, MEG, OPMs, TMS) , Neuromodulation (e.g. TMS, TDCS, FUS, SCS, MCS, DBS, vagus nerve stimulation)

Background and Aims: Accurate early prediction of motor recovery after spinal cord injury (SCI) is essential for prognosis and rehabilitation planning. While motor-evoked potentials (MEP) have been proposed as objective biomarkers, their added value beyond clinical assessment remains uncertain.
To evaluate the predictive performance of clinical and MEP-derived variables for short-term motor recovery in patients with SCI.

Methods: n this retrospective cohort study, 17 patients with SCI and complete longitudinal clinical data were included, of whom 16 had available baseline MEP measurements. Lower Extremity Motor Score (LEMS) and Modified Ashworth Scale (MAS) were recorded at baseline (W0), week 4 (W4), week 8, and week 12 (W12). Baseline MEP features were extracted and summarized. Continuous prediction of W12 LEMS was performed using ridge regression with leave-one-out cross-validation (LOOCV). Binary classification of “good recovery” (LEMS ≥ median at W12) was evaluated using logistic regression and random forest models. Model performance was assessed using mean absolute error (MAE), coefficient of determination (R²), area under the receiver operating characteristic curve (AUC), accuracy, and F1 score.

Results: Baseline LEMS was the strongest predictor of outcome and was independently associated with W12 LEMS (β = 0.93, P < 0.001). Ridge regression using baseline clinical variables demonstrated strong performance (MAE, 3.49; R², 0.93; Spearman r, 0.90). Logistic regression achieved high discrimination for good recovery (AUC, 0.97; accuracy, 0.82), which improved further with inclusion of W4 clinical data (AUC, 0.99; accuracy, 0.88). In contrast, models incorporating baseline MEP features showed lower performance (AUC, 0.71–0.76). No significant longitudinal changes or between-group differences were observed in LEMS or MAS (all P > 0.05).

Conclusions: Baseline motor function is the dominant predictor of short-term recovery after SCI, and early clinical reassessment further improves prediction. Baseline MEP-derived features did not provide additional predictive value in this cohort.

References: 1. K. L. Bunday and M. A. Perez, “Motor recovery after spinal cord injury enhanced by strengthening corticospinal synaptic transmission,” Current Biol., vol. 22, no. 24, pp. 2355–2361, Dec. 2012, doi: 10.1016/j.cub.2012.10.046.
2. K. A. Moxon, A. Oliviero, J. Aguilar, and G. Foffani, “Cortical reorganization after spinal cord injury: Always for good? ” Neuroscience, vol. 283, pp. 78–94, Dec. 2014, doi: 10.1016/j.neuroscience.2014.06.056.
3. R. N. Lemon, “Descending pathways in motor control,” Annu. Rev. Neurosci., vol. 31, no. 1, pp. 195–218, Jul. 2008, doi: 10.1146/annurev.neuro.31.060407.125547.
4. V. Dietz and K. Fouad, “Restoration of sensorimotor functions after spinal cord injury,” Brain, vol. 137, no. 3, pp. 654–667, Mar. 2014, doi: 10.1093/brain/awt262.
5. J.-C. Lamy, C. Ho, A. Badel, R. T. Arrigo, and M. Boakye, “Modulation of soleus h reflex by spinal DC stimulation in humans,” J. Neurophysiol., vol. 108, no. 3, pp. 906–914, Aug. 2012, doi: 10.1152/jn.10898.2011.

Presenting Author: Vu Pham Thao Vy
Submitting Author: Pham Thao Vy Vu

Senior Author: Wing P. Chan

IASP 2026 Global Year theme of Neuropathic Pain: Yes

• Wing P. Chan: PhD, MD - Taipei Medical University
• Bor-Shing Lin: PhD, MD - National Taipei University
• Zhao Zhang: PhD - National Taipei University
• Chih-Wei Peng - Taipei Medical University
• Shih-Hsuan Chen - National Taipei University
• Chien-Hung Lai - Taipei Medical University