OBJECTIVE Several imaging modalities are available for the evaluation of dysfunctional hemodialysis shunts. Color Doppler ultrasonography (CDUS) and digital subtraction angiography (DSA) are most widely used for the detection of access stenoses, and contrast-enhanced magnetic resonance angiography (CE-MRA) of shunts has recently been introduced. To date, no study has compared the value of these three modalities for stenosis detection in dysfunctional shunts. We prospectively compared CDUS and CE-MRA with DSA for the detection of significant (> or = 50%) stenoses in failing dialysis accesses, and we determined whether the interventionalist would benefit from CDUS performed before DSA and endovascular intervention. METHODS CDUS, CE-MRA, and DSA were performed of 49 dysfunctional hemodialysis arteriovenous fistulas and 32 grafts. The vascular tree of the accesses was divided into three to eight segments depending on the access type (arteriovenous fistula or arteriovenous graft) and the length of venous outflow. CDUS was performed and assessed by a vascular technician, whereas CE-MRA and DSA were interpreted by two magnetic resonance radiologists and two interventional radiologists, respectively. All readers were blinded to information from each other and from other studies. DSA was used as reference standard for stenosis detection. RESULTS DSA detected 111 significant (> or = 50%) stenoses in 433 vascular segments. Sensitivity and specificity of CDUS for the detection of significant stenosed vessel segments were 91% (95% CI, 84%-95%) and 97% (95% CI, 94%-98%), respectively. We found a positive predictive value of 91% (95% CI, 84%-95%) and a negative predictive value of 97% (95% CI, 94%-98%). The sensitivity, specificity, positive predictive value, and negative predictive value of MRA were 96% (95% CI, 90%-98%), 98% (95% CI, 96%-99%), 94% (95% CI, 88%-97%), and 98% (95% CI, 96%-99%), respectively. CDUS and CE-MRA depicted respectively three and four significant stenoses in six nondiagnostic DSA segments. The interventionalist would have chosen an alternative cannulation site in 38% of patients if the CDUS results had been available. CONCLUSIONS We suggest that CDUS be used as initial imaging modality of dysfunctional shunts, but complete access should be depicted at DSA and angioplasty to detect all significant stenoses eligible for intervention. CE-MRA should be considered only if DSA is inconclusive.