Warfarin treatment is commonly started with a fixed loading dose that might be associated with an increased risk of bleeding. An individual maintenance dose can then be estimated based on a pharmacogenetic algorithm. Starting treatment with the estimated dose implies a longer time to reach the therapeutic range. Our goal was to compare the safety and efficacy of initiating warfarin treatment with a loading dose guided by pharmacogenetics versus a maintenance dose. The primary endpoint was time in the therapeutic range (TTR) in the first 10 days of treatment. Secondary endpoints were time to the first international normalized ratio (INR) in therapeutic range (2.0-3.0) and occurrence of serious adverse events. Consenting cardioembolic stroke patients were genotyped for CYP2C9 (cytochrome P450 2C9 gene) and VKORC1 (vitamin K epoxide reductase complex, subunit 1 gene) polymorphisms and a maintenance warfarin dose was estimated. Patients were randomized into two groups. The loading dose group (LDG) patients received twice the estimated dose in the first 2 days of treatment. The maintenance dose group (MDG) patients received the estimated dose directly from day one. The TTR in the first 10 days was significantly higher in the LDG than in the MDG (50.5% vs. 38.3%, p = 0.003). The time to the first INR in this range was significantly shorter in the LDG (5.24 vs. 7.3 days). There were no significant differences in the INR above this range or serious adverse events. Warfarin loading dose guided by pharmacogenetics after recent cardioembolic stroke improved the efficacy of warfarin initiation without increasing the risk of adverse events.