Early Predictors of Post-stroke Motor Recovery

Thumbnail Image
Saltao da Silva, Mary Alice
Wheaton, Lewis A.
Borich, Michael
Associated Organization(s)
Organizational Unit
Organizational Unit
Supplementary to
Stroke is a leading cause of long-term disability worldwide. Despite robust spontaneous biological recovery mechanisms and provision of intensive rehabilitation therapies, most stroke survivors experience persistent loss of upper extremity function which is directly related to reduced independence in activities of daily living and diminished quality of life. Identification of clinical, anatomical, or neurophysiologic indices that accurately predict the capacity for recovery post-stroke is crucial to facilitate precision-based medicine approaches for clinical management, including targeted therapeutic interventions. The Predict Recovery Potential (PREP2) prediction tool uses a combination of clinical measurements and neurological biomarkers to predict paretic upper extremity (PUE) motor outcomes but has yet to be externally validated in the US healthcare system. The primary study objectives were to: 1) evaluate external validation feasibility of PREP2 in the US; 2) retrospectively assess current care practices to determine the routinely collected measures that are most predictive of PUE functional outcome post-stroke; 3) evaluate the prognostic merit of biomarkers isolated from clinical neuroimaging. I hypothesized that our data would demonstrate that PREP2 will be feasible for external validation in healthcare settings in the US and that a combination of clinical measures and biomarkers extracted from clinical magnetic resonance imaging (MRI) would accurately predict level of PUE motor recovery post-stroke in patients who underwent acute inpatient rehabilitation (AR) post-stroke. The studies were conducted via retrospective chart review for two cohorts of stroke patients over fiscal years 2016-2018. In Aim 1, I assessed prospective validation feasibility of the PREP2 prediction tool in acute care settings in the US using a cohort of all stroke admissions to Emory University and Grady Memorial Hospitals. In Aims 2 and 3, I assessed the ability of currently collected clinical measures and neurologic biomarkers isolated from clinical imaging to predict PUE motor outcomes post-stroke using a cohort of patients who remained within the Emory University Hospital system for acute hospitalization, AR, and outpatient care, allowing longitudinal assessment to track recovery and to estimate the level of PUE motor function return. Institutional electronic medical record systems were utilized to extract metrics including demographic data, stroke characteristics, longitudinal documentation of post-stroke motor function, and metrics of stroke care management along the post-stroke care continuum. Clinically diagnostic MRI was used to create lesion masks which were spatially normalized and processed to obtain corticospinal tract (CST) lesion overlap in both primary motor (M1) and non-M1 CST projections. Metric associations were investigated with correlation and cluster analyses, Kruskal-Wallis tests, classification and regression tree (CART) analyses. In Aim 1, we found that current stroke management allows for shoulder abduction finger extension manual muscle tests (SAFE score) to be obtained at therapy evaluations and for the National Institutes of Stroke Scale score to be extracted from the patient chart. On average, patients appropriate for CST integrity assessment remain in the acute care hospital setting at a time when CST function should be evaluated for PREP2 validation. In Aims 2 and 3, estimations of PUE strength extracted from the patient chart (E-SAFE) and clinical MRI-derived CST lesion overlap were associated with PUE functional outcome. Cluster analysis produced three distinct outcome groups and aligned closely to previous outcome categories. Outcome groups significantly differed in E-SAFE scores and lesion overlap on cortical projections within the CST, in particular those emanating from non-M1 cortical areas. Exploratory predictive models using clinical MRI metrics, either alone or in combination with clinical measures, were able to accurately identify recovery outcome category for patients using assessments made during both the acute and early subacute phases of post-stroke recovery. Results suggest that (1) prospective PREP2 validation studies are feasible in a US healthcare setting, (2) SAFE is an easy-to-acquire, readily implementable screening metric with high clinical utility for patients who undergo AR post-stroke, and (3) clinical MRI-derived biomarkers of both M1 and non-M1 contributions to CST integrity may offer unique insight into PUE motor outcome potential.
Date Issued
Resource Type
Resource Subtype
Rights Statement
Rights URI