Davidson and colleagues [1] report that color Doppler ultrasound has poor diagnostic accuracy for proximal thrombi in high-risk asymptomatic patients receiving medical prophylaxis. The results of this study could affect clinical practice in several different situations. Although clinicians rarely order an ultrasound screening test in patients receiving heparin prophylaxis, they do screen patients at high risk for thrombosis who cannot be given prophylaxis [2] and patients with suspected pulmonary embolism who have an indeterminate lung scan. The low accuracy of ultrasound suggests that venography should be ordered in these situations. However, some reservations about the study methods must be raised.

First, although the ultrasound test was blinded, it was not the primary end point but rather a preliminary tool to determine which leg to image using venography. The radiology technicians, who were aware that they were imaging a study patient, may not have been as careful as they would have been with a nonstudy patient. This bias could explain the higher frequency of both false-positive and false-negative results compared with that of other studies.

Second, each venogram was read by only one radiologist at each study site. Given the substantial interobserver variability in the interpretation of venograms in symptomatic patients [3], an expert panel should have reviewed the venograms in these asymptomatic patients.

Third, were the thrombi noted on the venograms in duplicated popliteal or duplicated superficial femoral veins, areas prone to underdiagnosis by ultrasound? Were they small and perivalvular, with uncertain clinical significance? In the study by Ginsburg and colleagues [4], eight of ten proximal thrombi missed by ultrasound were either small (<4 cm) or were "barely extending into the popliteal vein".

Fourth, the 8% false-positive rate is much higher than in similar studies of asymptomatic orthopedic patients (range, 1% to 3.1%) [4, 5]. In which leg and in which veins were the false-positive thrombi located? Were they near the replaced joint? Were the false-positive and false-negative results distributed equally among all the centers and between patients with hip and knee surgery?

It is possible that most small nonocclusive proximal thrombi are, like 80% of isolated calf thrombi, associated with a low risk for extending or causing symptomatic pulmonary embolization. If so, to assess the utility of ultrasound as a screening test for proximal thrombi would require a large trial using venography to confirm positive findings and objectively documented clinical end points, such as symptomatic venous thrombosis or pulmonary embolism, to evaluate negative ultrasound findings [2].