Brain-based predictions of cardiovascular risk factors in midlife populations at risk of dementia

1376 

Abstract Submission 

Bolin Cao^1,2, Qing Qi^1,2, Feng Deng^1,2, Maria-Eleni Dounavi³, Graciela Muniz-Terrera^4,5, Paresh Malhotra^6,7, Ivan Koychev⁸, John O'Brien³, Craig Ritchie^4,9, Brian Lawlor^1,2, Lorina Naci^1,2

¹Trinity College Institute of Neuroscience, School of Psychology, Trinity College Dublin, Dublin, Ireland, ²Global Brain Health Institute, Trinity College Dublin, Dublin, Ireland, ³Department of Psychiatry, School of Clinical Medicine, University of Cambridge, Cambridge, UK, ⁴Edinburgh Dementia Prevention, University of Edinburgh, Edinburgh, UK, ⁵Department of Social Medicine, Ohio University, Ohio, USA, ⁶Department of Brain Science, Imperial College London, London, UK, ⁷UK Dementia Research Institute Care Research and Technology Centre, Imperial College London and the University of Surrey, London, UK, ⁸Department of Psychiatry, Oxford University, Oxford, UK, ⁹Scottish Brain Sciences, Edinburgh, UK

Bolin Cao  
Trinity College Institute of Neuroscience, School of Psychology, Trinity College Dublin|Global Brain Health Institute, Trinity College Dublin
Dublin, Ireland|Dublin, Ireland

Qing Qi  
Trinity College Institute of Neuroscience, School of Psychology, Trinity College Dublin|Global Brain Health Institute, Trinity College Dublin
Dublin, Ireland|Dublin, Ireland

Feng Deng  
Trinity College Institute of Neuroscience, School of Psychology, Trinity College Dublin|Global Brain Health Institute, Trinity College Dublin
Dublin, Ireland|Dublin, Ireland

Maria-Eleni Dounavi  
Department of Psychiatry, School of Clinical Medicine, University of Cambridge
Cambridge, UK

Graciela Muniz-Terrera  
Edinburgh Dementia Prevention, University of Edinburgh|Department of Social Medicine, Ohio University
Edinburgh, UK|Ohio, USA

Paresh Malhotra  
Department of Brain Science, Imperial College London|UK Dementia Research Institute Care Research and Technology Centre, Imperial College London and the University of Surrey
London, UK|London, UK

Ivan Koychev  
Department of Psychiatry, Oxford University
Oxford, UK

John O'Brien  
Department of Psychiatry, School of Clinical Medicine, University of Cambridge
Cambridge, UK

Craig Ritchie  
Edinburgh Dementia Prevention, University of Edinburgh|Scottish Brain Sciences
Edinburgh, UK|Edinburgh, UK

Brian Lawlor  
Trinity College Institute of Neuroscience, School of Psychology, Trinity College Dublin|Global Brain Health Institute, Trinity College Dublin
Dublin, Ireland|Dublin, Ireland

Lorina Naci  
Trinity College Institute of Neuroscience, School of Psychology, Trinity College Dublin|Global Brain Health Institute, Trinity College Dublin
Dublin, Ireland|Dublin, Ireland

Dementia, particularly Alzheimer's disease (AD), is a growing public health challenge. The progression of AD underscores the critical importance of midlife as a period for preventive intervention [1-3]. As dementia is a heterogeneous neurodegenerative condition, single risk factors are inadequate for accurately identifying people who are most likely to develop dementia [4]. Instead, multifactorial risk scores, like the cardiovascular risk factors, aging, and dementia (CAIDE) risk score [5], encompassing cardiovascular (blood pressure, cholesterol, body mass index, and physical inactivity) and non-modifiable risk factors (age, sex, and APOE ε4 genotype), are crucial for dementia risk estimate. However, the underlying functional brain architecture correlating with these risk factors is poorly understood. This study seeks to bridge this gap through network-based statistic (NBS)-Predict [6], a novel predictive model aiming to identify neuroimaging biomarkers for dementia-related risks in midlife, thus aiding in personalized prevention and intervention strategies.

Resting-state fMRI imaging, CAIDE, cardiovascular and non-modifiable risk factors, and a lifetime of experiences questionnaire (LEQ) data were collected in 585 (207/378 female/male) healthy participants aged 40-59 years (mean = 50.9) enrolled in the PREVENT-Dementia study [7-9] who had useable MR data. Following standard preprocessing procedures, the Dosenbach atlas [10] was utilized to extract mean time series and construct functional connectivity (FC) matrices for each participant. We applied the NBS-Predict model to predict the CAIDE, cardiovascular, and non-modifiable risk factors scores based on the whole-brain FC (Figure 1). Specifically, our analysis employed a linear support vector machine characterized by 10-fold cross-validation (repeated 10 times), with feature selection at p<0.05 and 1000 permutations for statistical significance. Additionally, a hierarchical regression model was used to assess the impact of midlife LEQ score on the FC linked with cardiovascular risk factors. The LEQ specific and non-specific score, age, sex, years of education, and mean framewise displacement (FD) were set independent variables.

NBS-Predict models significantly predicted the CAIDE (r=0.214, p<0.001), the cardiovascular (r=0.201, p<0.001), and the non-modifiable (r=0.237, p<0.001) risk factors scores. We observed similar FC patterns related to the CAIDE and cardiovascular risk factor score, with connections between the somatomotor and cingulo-opercular networks and within the somatomotor network, as shown in Figure 2a. By contrast, the FC patterns for the non-modifiable risk score were different from those observed for CAIDE. Furthermore, we found that the non-specific score (physically, socially and intellectually stimulating mid-life activities) was positively associated with the FC of regions impacted by cardiovascular risk factors (β=0.001, p=0.017).

We found considerable overlap in FC patterns associated with CAIDE and cardiovascular risk factors in midlife, which were very different from those associated with non-modifiable risk. This suggests different neurobiological pathways for cardiovascular-based modifiable and other non-modifiable risk factors contributing to dementia risk in midlife. These results bolster the case for personalized approaches to dementia prevention decades before the onset of clinical symptoms. The research highlights the beneficial impact of an active and engaged lifestyle on brain health.

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ²

Aging

Classification and Predictive Modeling ¹

fMRI Connectivity and Network Modeling

Aging

Data analysis

FUNCTIONAL MRI

Machine Learning