The establishment of 60% or greater diameter stenosis by Doppler ultrasound is an eligibility requirement of the Asymptomatic Carotid Atherosclerosis Study (ACAS). We used a uniform statistical approach for each of 30 Doppler devices to establish a cutpoint for the peak systolic flow to insure a positive predictive value of 90% in predicting a 60%+ stenosis by angiography. Data were analyzed by device; however, performance relates to the device-sonographer-reader system. For those devices reporting in peak systolic velocity, cutpoints ranged from 151 to 390 cm/s, and for those reporting a peak systolic frequency from 5,400 to 11,250 Hz. Eighteen devices had a sensitivity above 60%, and nine devices had a sensitivity above 80%. However, for six instruments, the relationship between Doppler and angiography was too weak to establish any cutpoint. In addition, for one instrument a value could be established, but the associated sensitivity was only 18%. This remarkable variability in the performance is at odds with the high sensitivity uniformly published in the literature, suggesting (a) that the high reported sensitivity for Doppler may represent an overestimate of average performance, perhaps due to publication bias, (b) the paramount need for documented quality control measures within local laboratories to insure that Doppler examinations are performed reliably, and (c) the need for caution in the generalization of results among laboratories. © 1991, All rights reserved.