Prediction of probable Alzheimer's disease in memory-impaired patients

Accurate prediction of who will develop Alzheimer's disease (AD) will be important when appropriate early treatment is available, especially if pharmaceutical agents are developed to arrest the progression of this disease. There have been few longitudinal studies attempting to predict AD in preclinical subjects. The Bronx Aging Study ^[1-3] followed a sample of elderly nondemented subjects and found that their performance on delayed recall measures was an important predictor of dementia. Unfortunately, the predictor tests in these studies were used also as part of their larger diagnostic battery and thus are likely to have influenced final diagnoses. Furthermore, these investigations did not exclude patients who progressed to dementias of etiologies other than AD, and thus their findings are not specific to the prediction of probable AD.

Previously, we found that measures of delayed recall and recognition memory significantly improved the utility of apolipoprotein E (ApoE) in predicting who among a sample of memory-impaired patients would develop AD over a 2-year period. ^[4] We did not examine the role of other neuropsychological predictors in this study. Petersen et al. ^[5] found that ApoE was the best predictor of dementia in patients with mild cognitive impairment when compared with a mental status examination, ^[6] the Dementia Rating Scale ^[7] (DRS) and a measure of free and cued recall. However, as with the Bronx Aging Study, the investigators were interested in predicting dementia of any etiology, and the predictor tests were also used for the purposes of diagnosis.

There have been a number of cross-sectional comparisons of mild AD and normal aging. Some found that measures of delayed recall ^[8,9] and learning ^[10] produce excellent classification accuracies. Storandt et al. ^[11] found that a brief battery of four tests (which included measures of memory, attention, word fluency, and psychomotor function) was very accurate …