OBJECTIVE: To evaluate the safety of withholding anticoagulant treatment from patients with clinically suspected deep vein thrombosis but normal findings on compression ultrasonography. DESIGN: Compression ultrasonography was done with a simplified diagnostic procedure limited to the common femoral vein in the groin and the popliteal vein extending down to the trifurcation of the calf veins. Patients with normal ultrasonography findings at presentation were retested 1 week later. MAIN OUTCOME MEASURE: The incidence of venous thromboembolic complications during follow up for 6 months in patients in whom anticoagulant treatment was withheld on the basis of normal results on two ultrasonography tests 1 week apart. SETTING: University research centres in four hospitals. RESULTS: A total of 1702 patients were included in the study. Abnormal results on compression ultrasonography at presentation or at 1 week were found in 400 and 12 patients, respectively, for a prevalence of deep vein thrombosis of 24%. None of the patients were lost to follow up. Venous thromboembolic complications during the week of serial testing occurred in a single patient and in eight patients during 6 months' follow up, resulting in a cumulative rate of venous thromboembolic complications of 0.7% (95% confidence interval 0.3% to 1.2%). The mean number of extra hospital visits and additional tests required per initially referred patient was 0.8. CONCLUSION: It is safe to withhold anticoagulant treatment from patients with clinically suspected deep vein thrombosis who have a normal result on compression ultrasonography at the time of presentation and at 1 week.