Parenchymal hypodensity is a proposed risk factor for hemorrhage after recombinant tissue plasminogen activator (TPA) thrombolysis for ischemic stroke. In Buffalo, NY, and Houston, TX, the authors reviewed 70 patients who were treated with intravenous TPA for acute middle cerebral artery (MCA) stroke. Two observers blinded to clinical outcome analyzed initial noncontrast head computed tomography (CT) scans. Basal ganglia CT hypodensity was quantitated in Hounsfield units (HUs). Contralateral-ipsilateral difference in density was calculated using the asymptomatic side as a control. Ictus time to TPA averaged 2.5 hours. Six patients developed symptomatic intraparenchymal hematomas (2 fatal). The hemorrhage group had more severe basal ganglia hypodensity (mean 7.5 +/- 1.4, range 6-10 HU) than the nonhemorrhage group (2.2 +/- 1.4, range 0-9 HU) (P < .0001). The hemorrhage group had hypodensity of > 5 HU; the nonhemorrhage group had hypodensity of < or = 4 HU, except 1 patient with hypodensity of 9 HU. In predicting hemorrhage, the positive predictive value of hypodensity > 5 HU was 86%; the negative predictive value was 100%. Prethrombolysis NIH Stroke Scale (NIHSS) deficit (P = .0007) and blood glucose (P = .005) were also higher in the hemorrhage group. Age, gender, smoking, hypertension, and ictus time to TPA infusion did not differ between the 2 groups. Logistic regression indicated that basal ganglia hypodensity was the best single predictor of hemorrhage. Hypodensity and NIHSS score together predicted all cases of hemorrhage. The authors conclude that basal ganglia hypodensity quantified by CT may be a useful method of risk stratification to select acute MCA stroke patients for thrombolytic therapy.