Current duplex ultrasound criteria for internal carotid artery (ICA) stenosis (1%–15%, 16%–49%, 50%–69%, 70%–99%) may not be applicable to threshold stenoses used in symptomatic (North American Symptomatic Carotid Endarterectomy Trial [NASCET], Veterans’ Administration [VA]) and asymptomatic (Asymptomatic Carotid Atherosclerosis Study, VA) carotid endarterectomy (CEA) trials. This, along with increasing reports advocating CEA based on duplex results alone, prompted us to identify (1) new velocity criteria consistent with threshold stenoses used by these trials, and (2) velocity criteria with a high positive predictive value (PPV) (> 95%) and accuracy for detecting ⩾ 60% and ⩾ 70% ICA stenoses. This is the first study to propose criteria which can be used for all current CEA trials. The color duplex ultrasound (CDU) and arteriogram results of 462 ICAs were analyzed in blind fashion. Angiographic stenosis was calculated as in NASCET. Three velocity criteria (peak systolic velocity [PSV] of the ICA, end diastolic velocity [EDV] of the ICA, and the ratio of the PSV of the ICA/common carotid artery) were recorded and subjected to receiver operator characteristic curves (ROC) analysis to determine optimum criteria for identifying ICA stenoses of ⩾ 30%, ⩾ 50%, ⩾ 60%, and ⩾= 70%-99%. For ⩾ 30% stenosis (st): PSV ⩾ 120 cm/sec had an overall accuracy (OA) of 87%, sensitivity (sen.) of 93%, specificity (spec.) of 67%, PPV of 90%, and negative predictive value (NPV) of 77%; for ⩾ 50% st: PSV ⩾ 140 cm/sec had an OA of 93%, sen. of 92%, spec, of 95%, PPV of 97%, and NPV of 89%; for ⩾ 60% st: PSV ⩾ 150 cm/sec and an EDV of ⩾ 65 had an OA of 90%, sen. of 82%, spec, of 97%, PPV of 96%, and NPV of 86%; for ⩾ 70%–99% st: PSV ⩾ 150 cm/sec and an EDV of ⩾ 90 had an OA of 92%, sen. of 85%, spec, of 95%, PPV of 91%, and NPV of 92%. An ICA-PSV and EDV of 150, 65, and 150,110 had the best PPV (⩾ 95%) in detecting ⩾ 60% and ⩾ 70% st, respectively. When these new criteria are used, CDU can accurately detect threshold stenoses used by various CEA trials. Selected velocities with a high PPV (> 95%) may be used as the sole preoperative imaging.