Previous studies have shown both increased and decreased regional cerebral glucose metabolism-blood flow (rMRGlu-rCBF) values in diabetes. We sought to elucidate the influence of diabetes on rMRGlu-rCBF in 57 patients with pure cerebral microangiopathy. Sixteen of 57 patients had diabetes requiring therapy (11 NIDDM, 5 IDDM). Using a special head-holder for exact repositioning, rMRGlu (PET) and rCBF (SPET) were imaged and measured in slices, followed by MRI. White matter and cortex were defined within regions of interest taken topographically from MRI (overlay). Diabetic and non-diabetic microangiopathy patients were compared to 19 age-matched controls. The diabetic patients showed significantly lower rMRGlu-rCBF values in all regions than controls, whereas non-diabetic patients did not. There were no significant NIDDM-IDDM differences. rMRGlu-rCBF did not depend on venous blood glucose levels at the time of the PET examination. However, analysis of variance with the factors diabetes, atrophy and morphological severity of microangiopathy showed that lowered rMRGlu-rCBF in the diabetic group was due to concomitant atrophy only (P < 0.005), while neither diabetes nor microangiopathy had any influence on rMRGlu-rCBF (all P > 0.2). These results were confirmed by multivariate factor analysis. It can thus be concluded that a supposed decrease in rMRGlu-rCBF in diabetes mellitus is in fact only an artefact produced by the concomitant atrophy. All previous studies failed to correct for atrophy, and a critical reappraisal is required.