Carrasquillo, Minerva M.

Abstract Background Bedside cognitive screening tools intend to reliably and efficiently detect cognitive impairment in research, clinical, and community settings. Black American (BA) and Hispanic/Latino (H/L) adults obtain lower performance on widely used screeners, largely driven by demographics and structural and social determinants of health (SSDoH). This study characterized the performance of the Montreal Cognitive Assessment (MoCA) performance in a cohort of cognitively normal ethnoracially diverse adults while considering cohort‐specific factors that contributed to misclassification of impairment. Method Cognitively normal BA and H/L participants completed a comprehensive neurological assessment, neuropsychological testing, and MRI and amyloid‐ and tau‐PET imaging as part of a longitudinal study of memory and aging at Mayo Clinic in Florida. Demographically adjusted MoCA cutoffs were applied to BA (MoCA ≤ 22) and WH/L (MoCA ≤ 24) subgroups. Hierarchical binary logistic regression assessed whether amyloid burden (Centiloid value) improved prediction of scoring below demographically adjusted MoCA cutoffs beyond demographic/SSDoH factors, including age, education, and Area Deprivation Index (ADI). Result Participants included 105 (47 WH/L, 58 BA/AA) cognitively normal adults (mean age = 64.71 years, SD = 8.91; 62.9% female) with average of 16.07 years of education (SD=2.35) and mild‐to‐moderate neighborhood disadvantage (mean ADI=47.60, SD=23.77). 40% of H/L participants (MoCA (≤ 24) and 17% of BA participants (MoCa ≤22) scored below demographically adjusted MoCA cut‐offs. For the H/L subgroup, the best fitting model included age, education, ADI, and amyloid burden ( R 2 = 0 .43, χ 2 = 10.98, p = 0.03). Age (OR=1.21, 95% CI: 0.986 – 1.487) and amyloid burden (OR = 0.83, 95% CI: 0.67 – 1.03) approached statistical significance as individual predictors ( p = 0.06, p = 0.09). In the BA/AA subgroup, the best fitting model included age, education and ADI only ( R 2 = 0.47, χ 2 = 9.06, p = 0.02), with education approaching significance as an individual predictor (OR=0.46, 95% CI: 0.21 – 1.01, p = 0.05). Conclusion Sociodemographic factors continued to drive low MoCA performance in cognitively normal individuals despite applying demographically adjusted cut‐off scores by subgroup. These factors differentially impacted performance among BA/AA and WH/L participants. Careful consideration of these factors is warranted to mitigate risk of overdiagnosing impairment based on cognitive screening across clinical and research settings.

Abstract Background African Americans (AA) are twice as likely to develop dementia than non‐Hispanic Whites but remain underrepresented in Alzheimer's Disease (AD) research. Recent studies have identified AA specific risk factors for AD. Therefore, targeted biomarkers and/or therapies may be needed to effectively treat AD in AA. In this study, we aimed to determine if genetic variants that show association with AD‐risk, plasma transcript or protein levels in AA, may serve as accurate and accessible AD biomarkers. Method In 189 clinically diagnosed AA AD cases and 183 AA cognitively unimpaired (CU) controls, we performed targeted DNA sequencing of 10 AD‐associated loci. Genetic variants at these loci were tested for association with corresponding plasma transcripts and total tau protein levels previously measured in this cohort using a custom nanoString panel and Simoa assays, respectively. Utilizing phased haplotypes (SHAPEIT4) from target sequencing, we inferred local ancestry (RFMix v2) using five superpopulations from the 1000 Genomes Project as anchors. Subsequently, we tested the association of each ancestry specific allelic dosages with endophenotypes using Tractor. In receiver operating characteristic (ROC) analyses, the most significant e/pQTLs were added sequentially to a base model that included age and sex to identify e/pQTLs that improved the accuracy to correctly classify AD cases and CU controls. Result Of the 5,112 variants that were tested, 70 were nominally associated with AD‐risk and with plasma transcript or protein levels. Seven of these 70 variants also showed nominally significant associations of AD‐risk and transcript levels with either African or East Asian local ancestry. ROC analysis of age, sex, 33 e/pQTLs in APOE , SORL1 , TLR4 , EPHA1 , ABCA7 , CLU , and CR1 , and plasma levels of APP, ABCA7, AKAP9, CD14, CLU, and APOE ‐ɛ4 dosage achieved 88.2% area under the curve to discriminate AD vs. CU, a 30.4% improvement over the base model that only included age and sex. Conclusion Plasma transcript levels and e/pQTLs are promising diagnostic biomarkers for AD that may improve accessibility and reduce costs for a more accurate diagnosis of AD.