The Prognostic Value of CT Angiography and CT Perfusion in Acute Ischemic Stroke

Abstract

Background: CT angiography (CTA) and CT perfusion (CTP) are important diagnostic tools in acute ischemic stroke. We investigated the prognostic value of CTA and CTP for clinical outcome and determined whether they have additional prognostic value over patient characteristics and non-contrast CT (NCCT).

Methods: We included 1,374 patients with suspected acute ischemic stroke in the prospective multicenter Dutch acute stroke study. Sixty percent of the cohort was used for deriving the predictors and the remaining 40% for validating them. We calculated the predictive values of CTA and CTP predictors for poor clinical outcome (modified Rankin Scale score 3-6). Associations between CTA and CTP predictors and poor clinical outcome were assessed with odds ratios (OR). Multivariable logistic regression models were developed based on patient characteristics and NCCT predictors, and subsequently CTA and CTP predictors were added. The increase in area under the curve (AUC) value was determined to assess the additional prognostic value of CTA and CTP. Model validation was performed by assessing discrimination and calibration.

Results: Poor outcome occurred in 501 patients (36.5%). Each of the evaluated CTA measures strongly predicted outcome in univariable analyses: the positive predictive value (PPV) was 59% for Alberta Stroke Program Early CT Score (ASPECTS) ≤7 on CTA source images (OR 3.3; 95% CI 2.3-4.8), 63% for presence of a proximal intracranial occlusion (OR 5.1; 95% CI 3.7-7.1), 66% for poor leptomeningeal collaterals (OR 4.3; 95% CI 2.8-6.6), and 58% for a >70% carotid or vertebrobasilar stenosis/occlusion (OR 3.2; 95% CI 2.2-4.6). The same applied to the CTP measures, as the PPVs were 65% for ASPECTS ≤7 on cerebral blood volume maps (OR 5.1; 95% CI 3.7-7.2) and 53% for ASPECTS ≤7 on mean transit time maps (OR 3.9; 95% CI 2.9-5.3). The prognostic model based on patient characteristics and NCCT measures was highly predictive for poor clinical outcome (AUC 0.84; 95% CI 0.81-0.86). Adding CTA and CTP predictors to this model did not improve the predictive value (AUC 0.85; 95% CI 0.83-0.88). In the validation cohort, the AUC values were 0.78 (95% CI 0.73-0.82) and 0.79 (95% CI 0.75-0.83), respectively. Calibration of the models was satisfactory.

Conclusions: In patients with suspected acute ischemic stroke, admission CTA and CTP parameters are strong predictors of poor outcome and can be used to predict long-term clinical outcome. In multivariable prediction models, however, their additional prognostic value over patient characteristics and NCCT is limited in an unselected stroke population.