Six patients with venous thromboembolism were treated with heparin, administered intravenously by a constant infusion pump. The initial daily dose of heparin was adjusted to keep the activated partial thromboplastin time, sampled at 0800, between 1.5 and 2.5 times the control level. Once that level was obtained, this dose was kept constant. Anticoagulation was thereafter measured, every four hours for 48 hours, by activated partial thromboplastin time, thrombin time, and coagulation factor Xa inhibition assay. The results of all three coagulation tests showed a circadian variation in the six patients. Maximum values were achieved at night and minimum values in the morning. These circadian variations were reproduced for two consecutive days. Differences between night and morning values reached almost 50% for activated partial thromboplastin time, 60% for thrombin time, and 40% for factor Xa inhibition assay. This circadian variation resulted from two rhythms, a circadian rhythm lasting 24 hours and an ultradian rhythm lasting 12 hours, which were detected by cosinor analysis for each coagulation test (p less than 0.01). A circadian rhythm was detected individually in most of the patients for each coagulation test (p less than 0.05). All patients had a nocturnal peak in activated partial thromboplastin time on both days. In four patients this peak exceeded the upper desired limit of activated partial thromboplastin time. These rhythms should be taken into account when evaluating the dosage of heparin to be administered.