Comparison of Longitudinal Models in Estimating the Rate of Cognitive Decline Among AD Patients

• Activity Number: 571
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Yorghos Tripodis*+
• Companies: Boston University
• Keywords: mixed-effect models ; state-space models ; Alzheimer's Disease

Abstract

A number of statistical models has been used to model rate of decline among AD patients. We assessed different statistical methods in estimating effects of dichotomous variables on the rate of decline among AD patients from population and clinical studies. We generated 1000 perturbations where we randomly selected a subset that have a dichotomous characteristic that supposedly affects the rate of decline and we added it to each selected trajectory.We investigated different types of effects. We then fitted four different mixed effects models to allow for different patterns in the rate of decline: a) linear time and random intercept and slope, b) linear and quadratic time with random intercept and slope, c) linear and quadratic time with random intercept, slope and time squared d) state-space model with random walk plus drift. We calculated the bias and variance of the effect for each model.The choice of a statistical model affects the parameter estimates of the rate of decline and it has also an effect on its standard error. This has repercussions in studies where the rate of decline is the main outcome, as well as clinical studies that investigate disease-modifying effects.