Cross‐Sectional Comparison of Structural MRI Markers of Cognitive Impairment in an Ethno‐racially Diverse Cohort

Abstract

Listen

Translate article

AbstractBackgroundStructural MRI can describe neurodegeneration associated with aging and Alzheimer Disease (AD). Brain age gap (BAG) quantifies the difference between chronological age and predicted “brain age” and can be estimated using many published algorithms. Higher BAG indicates accelerated brain aging. Regional estimates of volume and thickness detect structural differences in AD signature regions (e.g., hippocampus, tempo‐parietal regions). This project aimed to identify the structural MRI measure that had the best correspondence with Clinical Dementia Rating (CDR) in an ethno‐racially diverse sample.MethodWe compared structural MRI measures in the Health & Aging Brain Study–Health Disparities (691 non‐Hispanic Black (NHB), 1094 Mexican‐American (MA) and 1085 non‐Hispanic White (NHW) participants). We calculated BAG using DeepBrainNet (BAGDBN) and BrainAgeR (BAGBAR). We derived Freesurfer‐based cortical thickness (meta‐ROI) and hippocampal volume (normalized for intracranial volume). We applied ANOVA followed by post‐hoc Tukey tests to assess differences in the interaction of self‐identified race/ethnicity and CDR, correcting for age, gender, and scanner. We performed ordinal regression to predict CDR using the same covariates, except CDR.ResultAfter correction for multiple comparisons, BAGDBN significantly differed between CDR 0 and CDR 0.5 in MA (pcorrected=0.045) and NHB (pcorrected=0.017) but not NHW (pcorrected=0.110). The other measures significantly differed between CDR 0 and CDR 0.5 for NHW only (Figure 1). For individuals with a 1 Z‐score elevation in BAGDBN (6.1 years), NHW have 1.21 (95%CI: 1.18, 1.26) times the odds of being CDR 0.5, while NHB significantly greater risk (OR = 1.31, 95%CI: 1.27, 1.36). MA have 1.26 times the odds (95%CI: 1.22, 1.31). In contrast, for individuals with a 1 Z‐score decrease in cortical thickness, there is no difference in odds ratio by ethno‐racial group (NHW=1.21, 95%CI: 1.17, 1.27; NHB=1.30, 95%CI: 1.24, 1.37; MA=1.28, 95%CI: 1.24, 1.33) (Figure 2).ConclusionThe neurodegeneration measures with the strongest correspondence to CDR varied by ethno‐racial group. BAGDBN may display greater sensitivity to the subtle changes that occur as individuals transition from CDR = 0 to 0.5 in underrepresented cohorts. Further work is needed to understand what features BAGDBN captures, but this suggests care should be taken when choosing neurodegeneration markers.