The limitations of magnetic resonance angiography in the diagnosis of renal artery stenosis: comparative analysis with conventional arteriography.

PURPOSE: Gadolinium-enhanced magnetic resonance angiography (MRA) is commonly used as a screening modality for the detection of renal artery stenosis. However, evidence supporting its utility in clinical practice is lacking; few rigorous studies have compared MRA with contrast arteriography (CA). After making anecdotal clinical observations that MRA sometimes overestimated the degree of renal artery stenosis, we decided to determine the interobserver variability, sensitivity, specificity, and diagnostic accuracy of MRA compared with CA.

METHODS: From September 1999 to April 2003, we evaluated 68 renal arteries in 34 patients with clinically suspected renal artery stenosis using both MRA and CA. All studies were independently reviewed by four blinded observers. Renal arteries were categorized by MRA as normal, <50%, and >50% stenosis/occlusion. The sensitivity, specificity, and accuracy of MRA detection of renal artery stenosis were compared to CA as the gold standard. Interobserver variability (kappa) was also calculated.

RESULTS: MRA demonstrated 87% sensitivity, 69% specificity, 85% accuracy, 95% negative predictive value, and 51% positive predictive value for the diagnosis of renal artery stenosis. Interobserver agreement was moderate for MRA (kappa = 0.53) and good for CA (kappa = 0.76). In 21 arteries (31%), MRA was falsely positive.

CONCLUSIONS: In patients with a high clinical suspicion of renal artery stenosis, MRA is 87% sensitive in the detection of >50% stenosis. However, MRA is relatively nonspecific compared with CA and results in significant overestimation of renal artery stenosis in nearly one third of patients. To reduce unnecessary CA, clinicians should consider supplemental studies.