Rate of cognitive change in Alzheimer's disease: methodological approaches using random effects models

Subtraction of serial scores, Least Squares Estimators, and Best Linear Unbiased Predictors (BLUPs) were compared for estimating rates of cognitive change for Mini-Mental State Exam (MMSE) and Dementia Rating Scale (DRS) scores for 299 probable Alzheimer's disease patients. The BLUPs provided cleaner group estimates of subjects' intercepts and slopes and are preferred. Regression analysis of the BLUP estimates of rate of change indicated that steeper declines were associated with higher levels of education and older age at onset. These effects were much smaller than those due to estimated initial cognitive test score. Differences in longitudinal metric characteristics of the MMSE and DRS were found, with the DRS yielding more precise change estimates. We discuss modeling these longitudinal data, and discuss use of the estimates of rate of change and intercept as data in their own right.