Predicting recovery after intracerebral hemorrhage – An external validation in patients from controlled clinical trials

• Published in: Journal of neurology 2009-03, Vol.256 (3), p.464-469
• Main Authors: Weimar, C., Ziegler, A., Sacco, R. L., Diener, H. C., König, I. R.
• Format: Article
• Language: English
• Subjects: Accuracy; Age; Age Factors; Aged; Aged, 80 and over; Algorithms; Archives; Archives & records; Area Under Curve; Biological and medical sciences; Cerebral Hemorrhage - diagnosis; Cerebral Hemorrhage - mortality; Clinical trials; Consciousness; Controlled Clinical Trials as Topic; Datasets; Fatalities; Female; Hemorrhage; Humans; Hypothesis testing; Ischemia; Logistic Models; Male; Medical sciences; Medicine; Medicine & Public Health; Middle Aged; Models, Theoretical; Multiple sclerosis and variants. Guillain barré syndrome and other inflammatory polyneuropathies. Leukoencephalitis; Neurology; Neuropsychological Tests; Neuroradiology; Neurosciences; Original Communication; Probability; Prognosis; Recovery of Function; Regression analysis; Severity of Illness Index; Stroke; Variables
• ISSN: 0340-5354; 1432-1459
• DOI: 10.1007/s00415-009-0115-z

Background An early and reliable prognostic indication in stroke patients is potentially useful for initiation of individual treatment and for informing patients and relatives. We recently developed a regression model as well as a simple 11-point predictive score (Essen ICH score) for functional recovery within three months after acute intracerebral hemorrhage (ICH) based on age and the National Institutes of Health Stroke Scale (NIH-SS). Here, we demonstrate the applicability of our models in an independent sample of ICH patients from controlled clinical trials. Methods The prognostic models were used to predict functional recovery in 564 patients from the Virtual International Stroke Trials Archive (VISTA). Furthermore, we tried to improve the accuracy by re-calibration and estimating new model parameters. Findings The logistic regression model and the Essen ICH score were able to correctly classify 77.5 % and 76.4 % of patients, respectively. Re-calibration and novel estimation of parameters yielded only a slight improvement of overall predictive accuracy. Interpretation For acute ICH patients included in controlled trials, our predictive models based on age and the NIH-SS correctly predict functional recovery after three months and could be useful for future trial design.