Agreement in endovascular thrombolysis patient selection based on interpretation of presenting CT and CT-P changes in ischemic stroke patients

Background To evaluate the agreement in patient selection based on computed tomography (CT) and CT-perfusion (CT-P) imaging interpretation between stroke specialists in stroke patients considered for endovascular treatment. Methods All endovascular-treated acute ischemic stroke patients were identified through a prospective database from two comprehensive stroke centers; 25 consecutively treated patients were used for this analysis. Initial CT images and CT-P data were independently interpreted by five board eligible/certified vascular neurologists with additional endovascular training to decide whether or not to select the patient for endovascular treatment. The CT/CT-Pimages were evaluated separately and used as the sole imaging decision making criteria, 2 weeks apart from each other (memory wash-out period). For each set of imaging data inter-rater and intra-rater agreement scores were obtained using Cohen's kappa statistic to assess the proportion of agreement beyond chance. Results Kappa values for the treatment decisions based on CT images was 0.43 (range 0.14-0.8) (moderate agreement), and for the decisions based on CTP images was 0.29 (range 0.07-0.67) (fair agreement) among the five subjects. There was substantial variability within the group and between images interpretation. Observed agreement on decision to treat with endovascular therapy was found to be 75% with CT images and 59% with CT-P images (with no adjustment for chance). Kappa values for intra-rater agreement were -0.14 (ranged -0.27-0.27) (poor agreement). Conclusions There is considerable lack of agreement, even among stroke specialists, in selecting acute ischemic stroke patients for endovascular treatment based on CT-Pchanges. This mandates a careful evaluation of CT-P for patient selection before widespread adoption.