Prediction of 30-day mortality among patients with thrombolysis-related intracranial hemorrhage.

BACKGROUND: Limited information exists on risk factors for mortality after thrombolysis-related intracranial hemorrhage. We wished to determine the characteristics associated with 30-day mortality after thrombolysis-related intracranial hemorrhage. METHODS AND RESULTS: We performed an observational analysis within a randomized trial of 4 thrombolytic therapies, conducted in 1081 hospitals in 15 countries. Patients presented with ST-segment elevation within 6 hours of symptom onset. Our population was composed of the 268 patients who had primary intracranial hemorrhage after thrombolysis. With univariable and multivariable analyses, we identified clinical and brain imaging characteristics that would predict 30-day mortality among these patients. CT or MRI were available for 240 patients (90%). The 30-day mortality rate was 59.7%. Glasgow Coma Scale score, age, time from thrombolysis to symptoms of intracranial hemorrhage, hydrocephalus, herniation, mass effect, intraventricular extension, and volume and location of intracranial hemorrhage were significant univariable predictors. Multivariable analysis of 170 patients with complete data, 98 of whom died, identified the following independent, significant predictors: Glasgow Coma Scale score (chi2, 19.3; P<0. 001), time from thrombolysis to intracranial hemorrhage (chi2, 15.8; P<0.001), volume of intracranial hemorrhage (chi2, 11.6; P<0.001), and baseline clinical predictors of mortality in the overall GUSTO-I trial (chi2, 10.3; P=0.001). The final model had a C-index of 0.931. CONCLUSIONS: This model provides excellent discrimination between patients who are likely to live and those who are likely to die after thrombolytic-related intracranial hemorrhage; this may aid in making decisions about the appropriate level of care for such patients.

Duke Scholars

Author Christopher Bull Granger Medicine, Cardiology

Author Cynthia Lea Green Biostatistics & Bioinformatics, Division of Biostatistics