Accuracy of predictive scores of hemorrhagic transformation in patients with acute ischemic stroke

• Autor(a) principal: Andrade,João Brainer Clares de
• Data de Publicação: 2022
• Outros Autores: Mohr,Jay Preston, Ahmad,Muhammad, Lima,Fabricio Oliveira, Barros,Levi Coelho Maia, Silva,Gisele Sampaio
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Arquivos de neuro-psiquiatria (Online)
• Texto Completo: http://old.scielo.br/scielo.php?script=sci_arttext&pid=S0004-282X2022000500455
• topic: Ischemic Stroke; Precision Medicine

ABSTRACT Background: Hemorrhagic transformation (HT) is a complication in ischemic strokes, regardless of use of reperfusion therapy (RT). There are many predictive scores for estimating the risk of HT. However, most of them include patients also treated with RT. Therefore, this may lead to a misinterpretation of the risk of HT in patients who did not undergo RT. Objective: We aimed to review published predictive scores and analyze their accuracy in our dataset. Methods: We analyzed the accuracy of seven scales. Our dataset was derived from a cohort of 1,565 consecutive patients from 2015 to 2017 who were admitted to a comprehensive stroke center. All patients were evaluated with follow-up neuroimaging within seven days. Comparison of area under the curve (AUC) was performed on each scale, to analyze differences between patients treated with recombinant tissue plasminogen activator (tPA) and those without this treatment. Results: Our dataset provided enough data to assess seven scales, among which six were used among patients with and without tPA treatment. HAT (AUC 0.76), HTI (0.73) and SEDAN (0.70) were the most accurate scores for patients not treated with tPA. SPAN-100 (0.55) had the worst accuracy in both groups. Three of these scores had different cutoffs between study groups. Conclusions: The predictive scores had moderate to fair accuracy for predicting HT in patients treated with tPA. Three scales were more accurate for predicting HT in patients not treated with tPA. Through standardizing these characteristics and including more patients not treated with RT in a large multicenter series, accurate predictive scores may be created.