Accuracy of T1-weighted chemical-shift MR imaging for the differentiation between benign and malignant adrenal masses was blindly assessed among three radiologists. MR imaging was performed at 1.5 T in 50 patients with 58 adrenal masses, proved benign (n = 38) or malignant (n = 20) based on surgery or growth (malignant) or stable size (benign) for at least 1 year. In-phase spin-echo sequences or in-phase breath-hold fast multiplanar spoiled gradient-recalled echo (FMPSPGR) sequences with a TE of 4.2 msec were compared with opposed-phase breath-hold FMPSPGR sequences with a TR/TE of 35-155/2.2-2.9 and a 90 degrees flip angle for the detection of lipid in adrenal masses. Three radiologists who were blinded to diagnosis and clinical data independently rated the likelihood of a benign adrenal lesion on a five-point scale of confidence. Mean sensitivity, specificity, and positive predictive value for a definite or probable diagnosis of a benign lesion by the three readers were 87%, 92%, and 95%, respectively. At the highest (definite) confidence of a benign lesion, the mean positive predictive value was 99%, with lower sensitivity (54%). Areas under receiver operating characteristic curves for the three radiologists were .98 (95% confidence interval CI = .94-1.00), .96 (CI = .91-1.00), and .95 (CI = .89-1.00). Interobserver variation for the diagnosis of a benign mass was low (kappa = .79). Chemical-shift imaging using breath-hold opposed-phase T1-weighted MR images is a reliable and reproducible technique for the diagnosis of most benign adrenal masses at the highest threshold of confidence.