Untangling the mechanistic links between heart and brain health in older populations: An AI assisted toolkit for assessing dementia risk

Dementia is a leading cause of disability and death globally. Poor heart health has consistently been linked to an increased risk of dementia. Yet not everyone with cardiovascular disease develops dementia and we don't yet have a method to separate people into different risk groups (e.g. high, medium, and low dementia risk groups). Therefore, this project aims to use advanced analytical methods to develop, for the first time, new approaches for identifying those individuals with cardiovascular disease who would be at highest risk of (1) dementia; and (2) dementia-related mortality. Being able to identify individuals at high risk of dementia is important as it could: (1) assist in timely identification; (2) allow for individually tailored treatment to reduce risk, where available; (3) assist with the development of new interventions focused on dementia; and (4) assist patients, their families, and clinicians with future planning.

We will analyse data from two large international resources including the Cohort Studies of Memory in an International Consortium (COSMIC) and the Emerging Risk Factors Collaboration (ERFC). These resources comprise baseline and follow-up data from over one million men and women in more than 30 longitudinal cohort studies conducted in over 30 countries. We will also access data from the UK Biobank. This includes extensive information from approximately 500,000 men and women aged 40-69 years at baseline in the UK. Access to such a large quantity of high-quality data will help to reduce bias and ensure that the findings are valid. Data collected in all studies includes questionnaire data, biological samples, and physical measurements. We will assess the likely complex pattern of associations between different risk variables (e.g. age, sex, ethnicity, health, diet, lifestyle and genetics) and dementia risk, in people with cardiovascular disease (e.g. coronary heart disease). We will answer questions like (1) what are the most important factors driving risk of dementia and dementia-related mortality; and (2) how can we incorporate knowledge across different factors to improve prediction of dementia?

With no cure available, the current strategy for dementia is to try and prevent, or delay, its onset in people at higher risk; and to diagnose as early as possible to allow access to support. The results are essential to informing treatment and management planning according to dementia risk status. They will also help to target existing treatments more efficiently in those at highest risk of dementia.

Dementia is a global public health priority. There is abundant evidence of a strong association between cardiovascular diseases (CVD) and associated risk factors (e.g. hypertension) with dementia risk. Yet, knowledge of the mechanistic pathways linking CVD and brain health is limited. Further, there is a lack of screening instruments to reliably assess risk of dementia and suggest possible interventions for further study.

To date, dementia treatment, prevention and risk reduction trials have shown limited success. This is largely due to inadequate knowledge of the disease pathways underlying dementia. Therefore, using Artificial Intelligence (AI) methods synthesising clinical, lifestyle, and socio-demographic data from the Cohort Studies of Memory in an International Consortium (COSMIC), the Emerging Risk Factors Collaboration (ERFC), and UK Biobank, we aim to develop, in people with CVD, novel models for predicting the risk of (1) incident dementia; and (2) dementia-related mortality.

The project will be completed in two phases. The first (Months 1-5) will involve secondary data analysis. The aim will be to develop and incorporate novel techniques for advanced analytics, including machine learning architectures, to enhance knowledge regarding correlations between dementia and CVD to establish a better patient-centric disease prediction methodology. We will answer questions like: What are the disadvantages and limitations of current approaches to calculate risk; what are the most important factors driving risk for a particular individual or data cohort; and, how can we incorporate knowledge domain to improve predictions? The second (Month 6) involves dissemination including knowledge sharing and design of a full application focused on translating the models into electronic toolkits. The timing of the study is critical as new approaches are needed to accelerate the development and delivery of precision interventions for dementia risk reduction and prevention.