Background and Purpose —Five pretreatment variables ( P <0.1 univariate analysis), including serum glucose (>300 mg/dL), predicted symptomatic intracerebral hemorrhage (ICH) in the National Institute of Neurological Disorders and Stroke rtPA trial. We retrospectively studied stroke patients treated <3 hours from onset with intravenous rtPA at 2 institutions to evaluate the role of these variables in predicting ICH. Methods —Baseline characteristics, including 5 prespecified variables (age, baseline glucose, smoking status, National Institutes of Health Stroke Scale [NIHSS] score, and CT changes [>33% middle cerebral artery territory hypodensity]), were reviewed in 138 consecutive patients. Variables were evaluated by logistic regression as predictors of all hemorrhage (including hemorrhagic transformation) and symptomatic hemorrhage on follow-up CT scan. Variables significant at P <0.25 level were included in a multivariate analysis. Diabetes was substituted for glucose in a repeat analysis. Results —Symptomatic hemorrhage rate was 9% (13 of 138). Any hemorrhage rate was 30% (42 of 138). Baseline serum glucose (5.5-mmol/L increments) was the only independent predictor of both symptomatic hemorrhage [OR, 2.26 (CI, 1.05 to 4.83), P =0.03] and all hemorrhage [OR, 2.26 (CI, 1.07 to 4.69), P =0.04]. Serum glucose >11.1 mmol/L was associated with a 25% symptomatic hemorrhage rate. Baseline NIHSS (5-point increments) was an independent predictor of all hemorrhage only [OR, 12.42 (CI, 1.64 to 94.3), P =0.01]. Univariate analysis demonstrated a trend for nonsmoking as a predictor of all hemorrhage [OR, 0.45 (CI, 0.19 to 1.08), P =0.07]. Diabetes was also an independent predictor of ICH when substituted for glucose in repeat analysis. Conclusions —Serum glucose and diabetes were predictors of ICH in rtPA-treated patients. This novel association requires confirmation in a larger cohort.