Compression ultrasonography is regarded as the non-invasive gold-standard to detect deep vein thrombosis (DVT) in patients presenting with symptoms. However, its use as a screening method in symptom-free postoperative patients at high risk of developing DVT remains controversial. In 100 consecutive patients who had undergone craniotomy, we compared the results of bilateral compression ultrasonic measurements of the results of bilateral compression ultrasonic measurements of the entire legs with the outcomes of contrast venography. Proximal DVT was detected in 13 patients, 5 of whom also had an abnormal ultrasonic result (sensitivity 38%, 95% CI 8-69%). Only 5 of the 9 patients with an abnormal ultrasound result for the proximal veins had proximal DVT (positive predictive value, 56%, 18-94%). Calf sonograms were evaluable in 71 of the 91 patients with bilaterally normal ultrasound results for the proximal veins. Of the 16 patients with calf DVT, ultrasound was abnormal in 8 (sensitivity 50%, 25-75%). Overall, ultrasound detected 13 of the 26 patients with proximal or isolated calf DVT (sensitivity 50%, 29-71%). The positive predictive value for the whole leg examination was 41% (24-60%). Most thrombi missed by ultrasound were non-occlusive and smaller than 5 cm. We conclude that compression ultrasound is not useful for screening for DVT in symptom-free postoperative high-risk patients.