Bridging Representation Gaps in Big Data: An Alzheimer's Disease Neuroimaging Initiative (ADNI) Investigation

Abstract

Listen

Translate article

Abstract Objective Large-scale research consortium studies are increasingly used to help aid in early identification of age-related brain changes and neurodegenerative disease risk. Racial/ethnic minorities (REM) are severely underrepresented in publicly available datasets related to neurodegenerative disease. This is problematic given that the US older adult population will be comprised of 40% REM by 2050. As this population grows more diverse, it is crucial that Alzheimer’s disease (AD) risk factors are appropriately characterized and REM are represented in longitudinal research. The purpose of this study was to examine memory, Apolipoprotein E (APOE) status, and hippocampal volume (HV) data among REM participants in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort. Participants and Method We sampled 91 Latinx and Non-Hispanic Black (NHB) ADNI participants, which included 41 controls (CN; 39% Latinx) and 50 individuals with mild cognitive impairment (MCI; 52% Latinx). Measures included Logical Memory (LM) Delay, Rey Auditory Verbal Learning Test (RAVLT), Alzheimer’s Disease Assessment Cognitive 13-item Scale (ADAS13), APOE e4 and APOE e2 status, and HV. Groups did not differ by age, education, or memory performance. None of the Latinx sample were e2 carriers, whereas 32% of NHB were e2 carriers. Results For NHB and Latinx MCI groups, LM Delay scores positively predicted greater HV (NHB: R2 = .17, p = .04; Latinx: R2 = .16, p = .04). Better RAVLT performance predicted greater HV for the Latinx MCI group but not for the NHB MCI group (Latinx: R2 = .13, p = .06). For NHB CN, e4 status and ADAS13 were positively correlated (r = .59, p = .002) but negatively correlated for Latinx CN (r = -.58, p = .002). APOE e2 and RAVLT percent forgetting were negatively correlated for NHB MCI (r = -.49, p = .01). Conclusions There appears to be differential risk associated with heterozygous e4 status in Latinx and NHB. APOE e2 may have been protective in the context of e4 in NHB. The degree to which e2 is protective in Latinx is unknown. Results highlight the need for improved REM representation in large-scale studies to understand genetic risk and protective factors in REM.