Translating from a stochastic model of ageing to clinically observable events in older adults at risk for cognitive impairment.

The world population is rapidly ageing, and with that dementia is increasing. The number of people with dementia is expected to rise from 24 to 81 million by 2040. With this, attention is now being paid to preserving brain health so as to preventing Alzheimer's disease (AD) and other dementias for developing. Apart from rare genetic forms, no reliable, single test now diagnoses AD. A comprehensive evaluation is needed before physicians can make a diagnosis, but often, by then, this neurodegenerative disease is well progressed. In consequence, it is important to identify who is at high risk. Worldwide, people have undertaken many studies on aging and cognition, with thousands of participants followed for many years. Invaluable information is contained in their databases that may hold the key to, as yet undetected ways of understanding who is at higher risk of dementia/AD. Just as business has turned to advanced computerized "data mining" to detect hidden patterns in their data, so too does it make sense to try such an approach in these large datasets. Traditional epidemiological and standards statistical methods have identified many single risk factors for AD such as age, family history and prior stroke, but they have not been used to develop risk profiles. In particular, they lack the power to address problems of heterogeneity that define complex systems such as the human body. For example, while age is very important for predicting AD, not every older person develops AD. Recently at Dalhousie University, where I plan to do this fellowship, a novel statistical framework has been developed for comprehensive analyses of longitudinal data. It is based on a new appreciation that the brain in Alzheimer's disease is not passive; it fights back by compensating. This compensation is dynamic - factors interact and change over time. These facts can be taken into account with dynamic systems modeling. Already, this analytical technique has successfully been applied to analysing general health status during aging, as well as its cognitive dimension. It allows analysis of changes in all directions, including improvement - which has had relatively little attention, but which appears to be essential to understand how AD occurs. In this application, the technique developed for general health and cognition will be directly applied to AD databases, including neuroimaging data. Four things are very important here. First, we can consider all outcomes (improvements, worsening and death) simultaneously, as competing events, and make projections of their likelihood in individuals. Second, we can analyze all outcomes in relation to risks factors (demographic, life styles, genetics, clinical, etc.) Third, we can study the onset of clinically detectable changes in cognition in relation to the risks factors. Fourth, we can estimate the likelihood of improvement or stabilization of cognition in relation to the risks factors. In particular, we will find how age influences the probability of transition between different states of disease. This research will contribute to our understanding of AD research with new knowledge about risk factors and improved research methods for studying complex systems like the ageing brain.