INTRODUCTION

Up to 70% of the 8.8 million dementia cases in India are attributable to modifiable risk factors. These factors may contribute to dementia through accelerated brain aging.

 

OBJECTIVE

Investigate how older Indians’ modifiable risk factor exposures relate to MRI-based brain age gap (BAG). We hypothesized that these exposures increase BAG.

 

METHODS

1. We trained a ridge regression-based brain age prediction model using T1-weighted MRI from 465 cognitively normal (CN) participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study plus 137 CN participants from Wave 2 of the Longitudinal Aging Study in India–Diagnostic Assessment of Dementia (LASI-DAD). We used 158 Freesurfer-based features, including cortical and subcortical volumes, cortical thicknesses and cortical surface areas, with chronological age as the dependent variable. We used 10-fold cross-validation to select the optimal ridge regularization parameter (lambda).

 

2. We computed brain age gap (BAG) as “predicted brain age minus chronological age” in 473 held-out (out-of-sample) LASI-DAD participants. We associated these BAGs with modifiable dementia risk factors using linear regressions, controlling for chronological age and FDR correcting for multiple comparisons.

 

DISCUSSION

Brain age prediction error was lower in ADNI compared with LASI-DAD (Fig 1), which may be due to the LASI-DAD training set being smaller and the cohort being more socioeconomically heterogeneous (Table 1).

We found no associations between modifiable exposures and BAG in LASI-DAD participants (Fig 2). Higher statistical power may be needed to detect these associations in India and similar low- or middle-income country (LMIC) contexts.