Objective: Wide interindividual variability exists in response to tissue plasminogen activator (t-PA) treatment in the acute phase of ischemic stroke. We aimed to find genetic variations associated with hemorrhagic transformation (HT) and mortality rates after t-PA. We then generated a clinical-genetic model for predicting t-PA response. Methods: Our prospective study used SNPlex to genotype 140 single nucleotide polymorphisms (SNPs) from 97 candidate genes in 3 patient cohorts. The cohorts included 1,172 patients who were treated with t-PA; 20.9% of them developed HT as evaluated by systematic brain computed tomography scan, and 10.6% died. A predictive model was generated by logistic regression (LR). Functional studies included real time quantitative polymerase chain reaction, nephelometry, and Western blot for alpha-2-macroglobulin (A2M) and activated partial thromboplastin time measurement for coagulation factor XII (FXII). Results: Replication analysis revealed that the SNP rs669 (Val1000Ile) in A2M was associated with HT, and rs1801020 (-4C>T) of F12 was associated with in-hospital death. The rs669 SNP withstood Bonferroni correction for HT (p< 3.57E-4). LR-based scores predicted HT occurrence (p = 9.13E-15) and in-hospital mortality (p = 8.7E-9) and were validated in an independent cohort. Val1000Ile modified A2M serum levels at baseline and after t-PA infusion, but not mRNA expression or protein structure; -4C>T affected FXII activity both prior to and after t-PA treatment. Interpretation: Two functional polymorphisms were consistently associated with t-PA safety. Our validated LR-based score predicts t-PA safety in the Spanish population. Copyright © 2012 American Neurological Association.