The Consortium to Establish a Registry for Alzheimer's Disease (CERAD). Part III. Reliability of a standardized MRI evaluation of Alzheimer's disease.

The Consortium to Establish a Registry for Alzheimer's Disease (CERAD) has developed procedures for standardized imaging and reporting of magnetic resonance (MR) findings in Alzheimer's disease (AD) for use by neuroradiologists in multiple medical centers using a variety of MR equipment and field strengths. After initial pretesting, we revised the protocol, expanded the summary rating scale to seven points, and added more illustrations. Fourteen participating neuroradiologists evaluated 28 MR scans of elderly patients, giving us the basis for judging interrater agreement. We obtained acceptable intraclass correlations (greater than 0.79) for rating the size of the lateral and third ventricles and the temporal horn. Less satisfactory intraclass correlations occurred when rating other areas, including (1) global atrophy of the brain (0.70); (2) dilatation of the sulci of the temporal lobe (0.66); (3) frequency, location, and severity of white matter lesions (0.77); (4) sylvian fissure enlargement (0.70); and (5) cerebral sulcal dilatation (0.64). We also saw considerable variation in the reporting of cortical and lacunar infarcts. Despite careful design of the rating methodology and readings by experienced neuroradiologists, we did not find satisfactory interrater agreement for interpreting MR findings in elderly subjects. These findings may explain the difficulties encountered in applying similar subjective rating techniques that meet with success at one institution to multicenter studies. More objective and reproducible procedures are needed for interpretation of neuroimaging findings of AD in multicenter studies.