OBJECTIVE—To investigate prospectively abnormalities of brain glucose utilisation in relation to major or minor neuropsychiatric symptoms in systemic lupus erythematosus (SLE). METHODS—Positron emission tomography (PET) using F-18-labelled fluorodeoxyglucose was performed in 28 patients with SLE. Patients were classified as having severe neuropsychiatric manifestations (seizures, focal neurological deficits, acute confusional states, mood disorders) (n=12), or mild neuropsychiatric manifestations (headache, reactive depression, cognitive dysfunction, anxiety disorders) (n=11) and five patients without signs of central nervous system (CNS) involvement. Ten clinically and neurologically healthy volunteers served as controls. In 26 patients magnetic resonance imaging (MRI) was performed and autoantibodies against CNS tissue, ribosomal P protein and cardiolipin were measured. In 14 patients follow up PET scans were performed after a mean (SD) period of 11.6 (9.5) months. RESULTS—PET scans showed hypometabolism in at least one brain region in all patients with severe or mild CNS symptoms (100%) as compared with patients without cerebral symptoms (40%) (p<0.0025). Parieto-occipital regions were most commonly affected (96%), followed by parietal regions (32%). In contrast, MRI images were abnormal in only 11 of 22 patients (50%) with neuropsychiatric symptoms and in one of four patients (25%) without symptoms. In 12 of 14 patients examined in follow up PET scans persistence, improvement or worsening of cerebral symptoms were associated with unchanged, decreased or increased brain hypometabolism, respectively. No significant correlation was found between PET or MRI findings and autoantibody profiles. CONCLUSIONS—PET imaging represents a sensitive tool to detect manifest or subclinical CNS involvement in SLE and PET findings correlate well with the clinical course of disease.