ABSTRACT
BACKGROUND AND PURPOSE: Hemorrhagic transformation (HT) is a complication of stroke that can occur spontaneously or after treatment. We aimed to assess the inter- and intrarater reliability of HT diagnosis. METHODS: Studies assessing the reliability of the European Cooperative Acute Stroke Study (ECASS) classification of HT or of the presence (yes/no) of HT were systematically reviewed. A total of 18 raters independently examined 30 post-thrombectomy computed tomography scans selected from the Aspiration versus STEnt-Retriever (ASTER) trial. They were asked whether there was HT (yes/no), what the ECASS classification of the particular scan (0/HI1/HI2/PH1/PH2) (HI indicates hemorrhagic infarctions and PH indicates parenchymal hematomas) was and whether they would prescribe an antiplatelet agent if it was otherwise indicated. Agreement was measured with Fleiss' and Cohen's κ statistics. RESULTS: The systematic review yielded four studies involving few (≤3) raters with heterogeneous results. In our 18-rater study, agreement for the presence of HT was moderate [κ = 0.55; 95% confidence interval (CI), 0.41-0.68]. Agreement for ECASS classification was only fair for all five categories, but agreement improved to substantial (κ = 0.72; 95% CI, 0.69-0.75) after dichotomizing the ECASS classification into 0/HI1/HI2/PH1 versus PH2. The inter-rater agreement for the decision to reintroduce antiplatelet therapy was moderate for all raters, but substantial among vascular neurologists (κ = 0.70; 95% CI, 0.57-0.84). CONCLUSION: The ECASS classification may involve too many categories and the diagnosis of HT may not be easily replicable, except in the presence of a large parenchymal hematoma.
