It is difficult to accurately forecast the clinical course of many patients presenting with mild cognitive problems. The utility in prognostic evaluation of various parameters of brain structure and function that can now be noninvasively measured remains to be clearly defined. The present work examined the value of regional cerebral metabolism, assessed with positron emission tomography (PET) and [(18)F]fluoro-2-deoxyglucose, in this context. PET scans of 167 patients (mean Mini-Mental State Examination (MMSE)=24 of 30 possible points) were classified as being positive or negative for evidence of progressive dementia. Results of scans were compared to patients' subsequent clinical course in general and in particular, to their changes in MMSE scores, for up to 10 years following PET. Data were further stratified according to the predictions of referring physicians based upon clinical assessments that had been performed up until the time of PET. Among those patients for whom a progressive dementing course had been predicted by PET criteria (but not those who were predicted by PET criteria to remain stable) a significant decline in general cognitive performance and MMSE scores occurred in the period following PET. Among those patients predicted by clinical criteria to have a progressive dementing illness, 94% of those with positive PET scans did suffer a progressive decline, while only 25% of those with negative scans progressed (relative risk 3.8). Similarly, among those patients who had been predicted by clinical criteria to remain cognitively stable, 74% of those with positive PET scans nevertheless suffered progressive decline, compared with 4% of those with negative PET scans (relative risk 18.4). These data indicate that evaluation of brain metabolism by PET in appropriately selected patients may improve the accuracy of clinical prognostic assessment.