Early diagnosis of Alzheimer's disease through joint modeling

Abstract

Alzheimer's disease is an irreversible neurodegenerative pathology resulting in a global loss of cognitive functions through the deterioration of brain tissue. Early diagnosis helps ensure appropriate treatment to slow the progression to dementia. One of the major issues is therefore early diagnosis. According to several previous studies (eg. Mouiha & Duchesne (2012), Caroli et al. (2010) and Jack, CR. Et al. (2010)), the progression time of subjects with mild cognitive impairment (MCI) to Alzheimer's disease (AD) is on average associated with the evolution of certain biomarkers, such as decreased hippocampal volume and increased atrophy. Despite the fact that these studies provide information on the progression to dementia based on biomarker trajectories, they have certain limitations, including: (1) They do not allow calculating a probability of individual risk of conversion to Alzheimer's disease; (2) They do not estimate the time at which this probability could be realized during the study period. One reason for these limitations lies in the often separate treatment of longitudinal and / or survival data in clinical or epidemiological studies, without taking into account their association, which can introduce bias into the results. Few or no studies to date have analyzed these data jointly, mainly due to a lack of appropriate statistical methods. Our main innovation is to model the evolution of atrophy and therefore the variation in hippocampal volume as a function of age, sex and certain neuropsychological tests (ADAS, MMSE, MoCA). To do this, we initially propose models combining these two types of data. These models exploit the association between data on conversion time to AD and biomarker evolution (cf. Ibrahima et al. (2010), D. Rizopoulos, 2012, Diggle, P. J., Sousa, I. and Chetwynd, A. G. (2008), Tsiatis A. and Davidian M. (2004)) in order to predict the individual and dynamic probability of conversion risk to dementia, in particular Alzheimer's disease. This probability can be updated as new data are collected for a subject.