In 220 consecutive outpatients with clinically suspected deep-vein thrombosis of the leg, we compared contrast venography with real-time B-mode ultrasonography, using the single criterion of vein compressibility with the ultrasound transducer probe. The common femoral and popliteal veins were evaluated for full compressibility (no thrombosis) and noncompressibility (thrombosis). Both veins were fully compressible in 142 of the 143 patients with normal venograms (specificity, 99 percent; 95 percent confidence interval, 97 to 100). All 66 patients with proximal-vein thrombosis had noncompressible femoral veins, popliteal veins, or both (sensitivity, 100 percent; 95 percent confidence interval, 95 to 100). For all patients (including 11 with calf-vein thrombi), sensitivity and specificity were 91 (95 percent confidence interval, 82 to 96) and 99 percent, respectively. The sensitivity for isolated calf-vein thrombosis was only 36 percent. The compression ultrasound test was repeated in a subset of 45 consecutive patients by a second examiner, unaware of the results of the first test, whose results agreed in all patients with those of the first examiner (kappa = 1). We conclude that ultrasonography with the single criterion of vein compressibility is a highly accurate, simple, objective, and reproducible noninvasive method for detecting proximal-vein thrombosis in outpatients with clinically suspected deep-venous thrombosis.