Predicting a post‐thrombolysis intracerebral hemorrhage: a systematic review

Summary Objectives Fear of an intracerebral hemorrhage (ICH) is a deterrent to the uptake of thrombolytic therapy, an evidence‐based treatment for an acute ischemic stroke. Several characteristics associated with post‐thrombolysis ICH have been identified, but their combined utility for risk stratification has yet to be clarified. We critically examined risk models to predict post‐thrombolysis ICH, and evaluated their potential clinical utility. Methods MEDLINE and EMBASE (Inception to October 2012) were searched and bibliographies of retrieved articles examined. Inclusion and exclusion criteria of each study were reviewed. Eligible studies had to report on the development, validation or impact assessment of a model constructed to predict the occurrence of ICH after thrombolysis for an ischemic stroke. Results We identified 10 publications reporting on 13 post‐thrombolysis ICH risk models, each developed from logistic regression. The number of participants/outcomes/predictors per model ranged from 89/45/4 to 27 804/558/7; age, glycemia and the NIH stroke scale score were common predictors. Models had modest‐to‐acceptable discriminatory ability (c‐statistic ranging from 0.59 to 0.88) in the derivation sample and during external validation (conducted for three models). Calibration was acceptable in the three studies where it was assessed. Impact studies to evaluate the effect of adopting ICH risk models on clinical outcomes have not yet been conducted. Conclusions The development and clinical application of ICH risk scores remains a developing field. Extensive external validation and impact studies are needed to strengthen evidence regarding their utility for improving the selection of ischemic stroke victims for thrombolysis while mitigating risks of complications through post‐thrombolysis ICH.