Profiles of cognitive decline in Alzheimer disease

Most recent studies have used only two observations to estimate the rate of cognitive decline in patients with Alzheimer disease (AD); few have data taken from more than a 2-year period; and none report on autopsy-verified cases. Repeated observations over the complete course of the disease are necessary to quantitatively evaluate hypotheses such as the triphasic linear model of Brooks et al. (1993). The goal of this study is to compare the triphasic linear and quadratic models of decline in a group of 12 AD patients confirmed at autopsy with a group of age- and sex-matched normal control subjects. Both groups were taken from the University of Western Ontario Dementia Study, and the Extended Scale for Dementia was used as the outcome measure. The squared multiple correlation as a measure of goodness of fit suggested the superiority of the more parsimonious quadratic model over the triphasic linear model. Quantitative models more accurately reflect the profiles of change in AD and may prove more sensitive in measuring the effects of drugs on these patterns.