Recent studies using in vivo proton magnetic resonance spectroscopy (1H MRS) have suggested that plasma phenylalanine (Phe) may not be a reliable indicator of brain Phe level in subjects with phenylketonuria (PKU). Interindividual variation in cerebral Phe can contribute to the phenotypic variability of the disease. We report the results of the direct assessment of brain Phe by 1H MRS in 10 off-diet PKU patients (aged 15.5-30.5 years), 4 detected and treated early, 6 late. In a single patient, brain Phe was evaluated before and 15 days after diet discontinuation. FLAIR MRI and 1H MRS were performed in the same setting by a 1.5 T clinical MR scanner. MR images were scored according to the extent of the lobar white-matter hyperintensity. Brain 1H MRS Phe signal (resonating at 7.36 ppm) was evaluated as a ratio to the creatine+phosphocreatine signal. Brain Phe was correlated with clinical, biochemical and MRI findings. Results were as follows. (1) An abnormal concentration of brain Phe was detected in all 10 PKU subjects (ranging from 0.030 to 0.074), associated with a wide interindividual variability of concurrent plasma Phe (ranging from 724 to 2800 micromol/L). (2) In late-detected subjects, brain Phe concentration correlated with clinical phenotype better than did plasma Phe. The discrepancy between brain and plasma Phe was relevant from a clinical point of view in two cases: in one, a late-detected patient with normal mental development, a high level of plasma Phe was associated with a relatively low concentration of brain Phe; in the other, a late-detected subject with severe neurological impairment, a very high level of brain Phe was associated with plasma Phe compatible with the diagnosis of mild PKU. (3) White-matter alterations were detected in all patients. FLAIR MRI sequences disclosed an involvement of optic chiasma and tracts in 7 subjects. No correlation was found between white-matter alterations and concurrent brain Phe concentrations. (4) In the only case assessed under different intake of Phe, the relevant increase of brain Phe paralleled the concurrent increase of plasma Phe, showing that 1H MRS can be a useful tool in evaluating the individual vulnerability of PKU patients to different values of plasma Phe.