Compression ultrasonography was compared with contrast venography in 215 hospitalized patients (218 limbs) with suspected deep venous thrombosis. All scans were performed using a 5-MHz linear-array scanner. The calf veins were not assessed owing to their small caliber. Distribution of deep venous thrombosis was proximal in 113 of 215 patients (number of limbs with deep venous thrombosis being the same as the number of patients) and limited to the calf veins (distal) in 29 patients. Deep venous thrombosis was detected by compression ultrasonography in 101 of 113 patients (sensitivity, 89%) and falsely diagnosed in two of 76 limbs (73 patients) with negative venographic results (specificity, 97%). The method was less sensitive below the knee, where deep venous thrombosis of the distal popliteal vein was not detected in five of 10 patients as compared with seven of 103 patients with thrombus extension above the knee. Pelvic vein deep venous thrombosis (n = 34) was detected by compression ultrasonography in 71% of the patients. The results of this study indicate that venography may be omitted in patients where compression ultrasonography demonstrates proximal deep venous thrombosis. In patients with negative compression ultrasonographic results, however, venographic verification is needed since venography has a considerably higher sensitivity than compression ultrasonography in detecting isolated iliac and calf vein deep venous thrombosis.