Longitudinal trajectories of cognitive change in preclinical Alzheimer's disease: a growth mixture modeling analysis

Abstract

Preclinical Alzheimer's disease (AD) refers to a period of time prior to diagnosis during which cognitive deficits among individuals who will go on to receive a diagnosis of AD are present. There is great interest in describing the nature of cognitive change during the preclinical period, in terms of whether persons decline in a linear fashion to diagnosis, or exhibit some stability of functioning, followed by rapid losses in performance. In the current study we apply Growth Mixture Modeling to data from The Kungsholmen Project to evaluate whether decline in Mini Mental State Examination (MMSE) scores during the preclinical period of AD follows a linear or quadratic function. At the end of a 7-year follow-up period, some individuals would be diagnosed with AD (n=71), whereas others would remain free of dementia (n=457). The results indicated that a two-group quadratic model of decline provided the best statistical fit measures, as well as the greatest estimates of sensitivity (67%) and specificity (86%). Differences in MMSE scores were apparent at baseline, but the preclinical AD group began to experience precipitous declines three years prior to diagnosis. Finally, persons who were misclassified as preclinical AD had fewer years of education and poorer MMSE scores at baseline.