Measuring cognitive change in a cohort of patients with Alzheimer's disease

Abstract

Annualized rates of cognitive change in Alzheimer's disease (AD), an important index of disease progression, show marked variability. To determine factors leading to such variability, we computed rates of change in a cohort of patients with AD tested annually with the Mini Mental State Examination (MMSE) and the more detailed Dementia Rating Scale (DRS). Estimates of rates of change (slopes) and intercepts were calculated using least squares and best linear unbiased predictors (BLUPs). Potential predictors of rates of change were examined using multivariate linear regression analysis. We found that the MMSE had more noise than the DRS. For the MMSE, slopes showed a moderate floor effect and a slight ceiling, depending on initial MMSE scores. These effects were less prominent for the DRS, for which slopes increased as intercepts decreased. In analyses of predictors of change, the MMSE was less useful than the DRS. In multiple linear regression models using DRS data, predictors showed statistically stronger effects and explained a greater extent of variation of slopes than did similar models using MMSE data. For example, among patients who died and underwent brain examination at autopsy, neuropathology of Lewy bodies plus AD (Lewy Body variant; LBV) was associated with significantly faster rates of cognitive decline compared to pure AD in analyses that used the DRS, but only trends were identified with the MMSE. The metric properties and longitudinal characteristics of cognitive tests and the statistical methods used to calculate change are key factors in obtaining reliable estimates of change in studying cohorts of patients with AD.