OBJECTIVE To prospectively compare the diagnostic accuracy of duplex ultrasonography, captopril renography, computed tomography angiography (CTA), and 3D Gd magnetic resonance angiography (MRA) in diagnosing hemodynamically significant renal artery stenosis (RAS). METHODS The standard of reference was measurement of transstenotic pressure gradient. Fifty-eight hypertensive patients with suspicion of RAS were evaluated, when possible, by all five techniques. Sensitivity and specificity to detect RAS were compared for each technique on both a patient and kidney basis. Discrepancies were evaluated separately and classified as borderline, method dependent, or operator dependent. RESULTS The prevalence of RAS was 77%. The sensitivity/specificity of ultrasonography, captopril renography, CTA, and MRA in detecting kidneys with RAS was 73/71%, 52/63%, 94/62%, and 93/91%, respectively. Ultrasonography had a significantly lower sensitivity than CTA and MRA (P<0.001) but higher than captopril renography (P = 0.013). Borderline RAS was the main cause for discrepancies. CONCLUSIONS MRA and CTA were significantly better than duplex ultrasonography and captopril renography in detecting hemodynamically significant RAS. The ultrasonography criteria for RAS based on the evaluation of renal peak systolic velocity and renal/aortic ratio are questionable. Captopril renography cannot be recommended for assessing RAS.