BACKGROUND AND PURPOSE: The Asymptomatic Carotid Atherosclerosis Study (ACAS) Doppler validation study assessed the performance of individual Doppler machines across a spectrum of laboratories. We attempted to establish a threshold specific to individual machines to predict angiographically defined hemodynamic stenosis. The reliability of these Doppler ultrasound criteria was prospectively and independently evaluated among patients screened with ultrasound in the ACAS trial. METHODS: Regression techniques were used to establish the relationship between Doppler velocity and percent stenosis by angiography for 63 specific Doppler machines. This relationship was used to establish a Doppler threshold to provide a 90% positive predictive value (PPV) of a 60% stenosis by angiography. The sensitivity of each Doppler machine to detect a 60% stenosis (at the 90% PPV threshold) was estimated. The efficacy of these Doppler thresholds was then prospectively evaluated by calculating the PPV among ACAS participants eligible by ultrasound. RESULTS: Of the 63 machines, 13 (21%) had an excellent sensitivity (80%+) at 90% PPV. In 32 devices (51%) only a marginal sensitivity (50% to 80%) could be achieved. In 9 devices (14%) the sensitivity was poor (0% to 50%), and in 9 (14%) no threshold could be established. Despite the heterogeneity of Doppler performance, the standardization program worked as designed in the ACAS trial. Of 825 surgical patients, 399 were eligible by Doppler and 395 subsequently underwent angiography. Of these, 32 (8.1%; 95% confidence interval, 5.4% to 10.8%) did not have hemodynamically significant stenosis by arteriography, a proportion nonsignificantly lower than the planned 10% by the PPV. CONCLUSIONS: The performance of Doppler ultrasound was highly variable. This suggests that Doppler performance is likely overstated in the literature, but specific devices may perform satisfactorily to detect individuals with hemodynamically significant stenosis. Because performance differs substantially among devices, local investigators are strongly urged to maintain local standardization series. With such standardization, ultrasound performance is sufficient for admission to clinical trials and as the is sufficient for admission to clinical trials and as the basis for carotid surgery. However, without quality control many ultrasound machines are not adequate to accurately predict the degree of carotid stenosis and should not be the only test to decide whether surgery is warranted.