BACKGROUND The value of carotid endarterectomy for stroke prevention depends on reliable identification of patients at higher risk for stroke from their internal carotid artery (ICA) occlusive disease than from surgery. This selection of patients is based on the degree of ICA stenosis. Therefore, preoperative diagnostic measures should strive for a prevalence independent probability for disease of 100%. Aim of this prospective study was to obtain clinically applicable duplex scanning criteria for ICA stenosis > or = 70% with a probability for disease of 100%. METHODS In 124 ICA in 62 patients (79% male) angiography and duplex scanning were performed. Degree of stenosis was classified in 4 categories: I < 50%; II 50%-69%; III 70%-99%; IV 100%. Cohen's kappa statistic was used to estimate agreement between both methods within categories. To improve accuracy post-test likelihood for disease was calculated for each point on the receiver operating characteristics (ROC)-curve for peak systolic (PSV) and end-diastolic velocity (EDV), and cut-off points for velocity criteria were set at a positive likelihood of 100%. RESULTS Diagnostic agreement was good with kappa = 0.77 (95% CI, 0.64-0.90; p < 0.001). For EDV a criterion of > or = 150 cm/sec was associated with a post-test likelihood for disease of 100%. For PSV no appropriate criterion could be detected. CONCLUSIONS A probability of 100% for ICA stenosis > or = 70% can be achieved by mere preoperative duplex scanning. Vascular laboratory specific validation of duplex scanning criteria should consider prevalence independent post-test likelihood for disease to ensure the value of CEA for stroke prevention.