Abstract TP158: Neural Function and Injury Combined Best Predict Behavioral Gains During Inpatient Stroke Rehabilitation

Abstract

Background: A better understanding of the mechanisms of recovery during rehabilitation could inform treatment decision-making. We tested two hypotheses: [1] a combination of neural function and injury measures is better than either measure alone for predicting motor gains during inpatient rehabilitation facility (IRF) admission; and [2] performance of prediction measures varies according to severity of baseline impairment. Methods: Fifteen patients with subacute stroke (56±12 yr, 16 days post-stroke) admitted to an IRF underwent EEG [3-min, resting-state, dense-array (256-lead)] and MRI [anatomical and diffusion tensor] at IRF admission; and serial behavioral testing. Neural function was assessed using EEG measures of coherence and power from electrodes overlying ipsilesional (M1 i ) and contralesional (M1 c ) primary motor cortex, in the Delta (1-3 Hz) and high Beta (20-30 Hz) frequency bands. Neural injury was assessed as integrity of white matter in corpus callosum (CC). Change in arm Fugl-Meyer (FM) and Functional Independent Measurement motor (FIM-m) scores served as primary and secondary behavioral recovery metrics, respectively. Results: In subjects with moderate or severe impairment (FM <55, N=11), neither neural function (M1 i -M1 c Delta coherence) nor neural injury (CC integrity) alone significantly predicted FIM-m score change. However, when combined into a single model, these measures did significantly predicted FIM-m score change (R 2 = 0.85, p=0.024); note that baseline behavior was not a significant predictor. An identical neural injury+function model approached significance at predicting FM score change (R 2 = 0.72 p=0.08). These models failed, however, when applied to all 15 patients (R 2 = 0.06, p=0.81). Conclusions: Results thus far in this ongoing study suggest that recovery during inpatient rehabilitation is best predicted by combining neural function and injury measures, and not by behavioral assessments. Performance of recovery predictors varies according to severity of baseline deficits, as adding mild strokes to moderate/severe strokes increased the sample size but diluted the model. These findings could potentially inform patient selection, treatment decisions, and discharge planning in an IRF setting.