DURATION OF ALZHEIMER’S DISEASE IN THE PRECLINICAL, PRODROMAL AND DEMENTIA STAGE: A MULTI-STATE MODEL ANALYSIS

Abstract

About 25% of the non-demented elderly have evidence of amyloid pathology. The time to and duration of symptomatic disease, dementia and death is not well known. We aim to combine international cohort data to estimate the duration of the preclinical, prodromal and dementia stage according to age, gender and setting. We selected 240 subjects with normal cognition and amyloid pathology (preclinical Alzheimer's Disease (AD)), 626 with MCI and amyloid pathology (prodromal AD) and 2091 subjects with AD-type dementia, of whom 38% had confirmed amyloid pathology, while for 62% amyloid status was unknown. Subjects were selected from the Alzheimer Disease Imaging Initiative (ADNI) and memory clinic based Amsterdam Dementia Cohort, DESCRIPA, and ICTUS studies. Longitudinal data on diagnosis and survival was used to fit a multistate model (MSM) that included four stages: preclinical AD, prodromal AD, AD-type dementia, and death. R-packages msm and ELECT were used to estimate progression rates centered at age 70, hazard ratios per year increase in age, and duration of stages according to age, gender and setting (for preclinical AD population-based versus memory clinic). Figure 1 illustrates the MSM model, including progression rates and hazard ratios per year increase in age. Duration of the preclinical AD stage was on average 8 years and the prodromal AD stage 4.5 years, independent of age. The duration of the dementia stage was shorter with increasing age. (Table 1). Duration of preclinical AD in women was longer than in men, 10.2 and 6.1 years respectively. In addition, in women with prevalent dementia the dementia stage was 3 years longer. The preclinical AD stage in memory clinic subjects (3.3 years) was shorter than population-based subjects (8.5 years, Table 2). Age had an impact on duration of the dementia stage, but not the pre-dementia stages. Disease duration was longer in women than men. The shorter preclinical AD stage in the memory clinic, might be due to a referral bias of subjects close to the prodromal stage. Our MSM model estimates can be used for the planning of future studies and could help to give patients a prognosis. Multi-state model with risk estimates (95% CIs) at age 70, and hazard ratios for a year increase in age (95% CIs), misclassification 1.1% for prodromal AD to dementia.