In a prospective double-blind trial, we treated 194 patients with acute venous thromboembolism with heparin or low molecular weight heparin (LMWH; Fragmin). To evaluate the most important prognostic factors for bleeding, the presenting clinical features of the patients, the patients' anticoagulant responses, and the doses of the drugs were analyzed using univariate and multivariate regression analyses. No significant differences in clinical risk factors associated with bleeding were observed between heparin and LMWH. The univariate analyses ranked the parameters in the following order of importance: World Health Organization (WHO) performance status, history of bleeding tendency, cardiopulmonary resuscitation, recent trauma or surgery, leukocyte counts, platelet counts, duration of symptoms, and body surface area. Patients with WHO grade 4 had an eightfold increase in risk of bleeding as compared with WHO grade 1. Assessment of the individual contribution of each variable using multivariate regression analysis showed that the WHO performance status was the most important independent factor predicting major bleeding. A history of a bleeding tendency, recent trauma or surgery, and body surface area were also independent risk factors. The risk of bleeding was influenced by two factors related to the treatment, the patient's anticoagulant response as measured with the anti-Xa assay and the dose of the drug expressed as U/24 h/m2. An increased risk of bleeding was only observed at mean anti-Xa levels greater than 0.8 U/mL for both drugs. Significantly more major bleedings occurred in patients treated with high doses of the drugs, an observation that was independent of the concomitant anti-Xa levels. It should be considered whether choosing an appropriate initial dose adapted to the patient's body surface area and clinical risk factors can improve the efficacy to safety ratio of heparin treatment.