Multivariate Prognostic Model of Acute Stroke Combining Admission Infarct Location and Symptom Severity: A Proof-of-Concept Study

Seyedmehdi Payabvash; John C Benson; Andrew E Tyan; Shayandokht Taleb; Alexander M McKinney

doi:10.1016/j.jstrokecerebrovasdis.2017.10.034

Multivariate Prognostic Model of Acute Stroke Combining Admission Infarct Location and Symptom Severity: A Proof-of-Concept Study

J Stroke Cerebrovasc Dis. 2018 Apr;27(4):936-944. doi: 10.1016/j.jstrokecerebrovasdis.2017.10.034. Epub 2017 Nov 29.

Authors

Seyedmehdi Payabvash¹, John C Benson², Andrew E Tyan³, Shayandokht Taleb², Alexander M McKinney²

Affiliations

¹ Department of Radiology, University of Minnesota, Minneapolis, Minneapolis; Department of Radiology and Biomedical Imaging, University of California, San Francisco, California. Electronic address: [email protected].
² Department of Radiology, University of Minnesota, Minneapolis, Minneapolis.
³ Department of Radiology, University of Minnesota, Minneapolis, Minneapolis; Department of Radiology and Radiological Sciences, Johns Hopkins University, Baltimore, Maryland.

PMID: 29198948
DOI: 10.1016/j.jstrokecerebrovasdis.2017.10.034

Abstract

Background: The information on topographic distribution of acute ischemic infarct can contribute to prediction of functional outcome. We aimed to develop a multivariate model for stroke prognostication, combining admission clinical and imaging variables, including the infarct topology.

Methods: Acute ischemic stroke patients without baseline functional disability who had magnetic resonance imaging within 24 hours of onset or last-seen-well were included. The admission stroke severity was determined using the National Institutes of Health Stroke Scale (NIHSS) score. The relation between infarct location and outcome was assessed using both voxel-based and visual atlas-based analyses. The disability/death was defined by a modified Rankin Scale score greater than 2 at 3-month follow-up.

Results: Among 198 patients included in this study, higher admission NIHSS score (P < .001), larger infarct volume (P < .001), and major arterial occlusions (P < .001) were associated with disability/death in univariate analyses. On voxel-based analysis, infarcts in the middle centrum semiovale, insula, and midbrain/pons were associated with higher rates of disability/death. In multivariate analysis, admission NIHSS score (P < .001), infarction of insula (P = .005), and midbrain/pons (P = .006) were independent predictors of disability/death. In receiver operating characteristics analysis, a simple 0-to-3 scoring system using these 3 variables had an area under the curve of .812 for prediction of disability/death (P < .001).

Conclusions: Admission symptom severity, infarction of insula, and midbrain/pons were independent predictors of clinical outcome in acute ischemic stroke patients. The methodology of this hypothesis-generating study can help conceive quantitative population-based probabilistic models for prognostication or treatment triage in stroke patients, combining admission clinical and imaging findings-including infarct topography.

Keywords: Ischemic stroke; brainstem; insula; prognosis; topography.

MeSH terms

Aged
Aged, 80 and over
Area Under Curve
Brain / diagnostic imaging*
Brain / physiopathology*
Brain Ischemia / diagnosis*
Brain Ischemia / mortality
Brain Ischemia / physiopathology
Brain Ischemia / therapy
Brain Mapping / methods*
Databases, Factual
Decision Support Techniques*
Diffusion Magnetic Resonance Imaging*
Disability Evaluation*
Female
Humans
Logistic Models
Male
Mesencephalon / diagnostic imaging
Mesencephalon / physiopathology
Middle Aged
Multivariate Analysis
Odds Ratio
Patient Admission*
Pons / diagnostic imaging
Pons / physiopathology
Predictive Value of Tests
Proof of Concept Study
ROC Curve
Reproducibility of Results
Retrospective Studies
Risk Factors
Severity of Illness Index
Stroke / diagnosis*
Stroke / mortality
Stroke / physiopathology
Stroke / therapy
Time Factors
Treatment Outcome