Plasma Aβ42/40 and cognitive variability are associated with cognitive function in Black Americans: Findings from the AA-FAIM cohort

Abstract

Introduction: It is critical to develop more inclusive Alzheimer's disease (AD) research protocols to ensure that historically excluded groups are included in preclinical research and have access to timely diagnosis and treatment. If validated in racialized groups, plasma AD biomarkers and measures of subtle cognitive dysfunction could provide avenues to expand diversity in preclinical AD research. We sought to evaluate the utility of two easily obtained, low-burden disease markers, plasma amyloid beta (Aβ)42/40, and intra-individual cognitive variability (IICV), to predict concurrent and longitudinal cognitive performance in a sample of Black adults.

Methods: Two hundred fifty-seven Black participants enrolled in the African Americans Fighting Alzheimer's in Midlife (AA-FAIM) study underwent at least one cognitive assessment visit; a subset of n = 235 had plasma samples. Baseline IICV was calculated as the standard deviation across participants' z scores on five cognitive measures: Rey Auditory Verbal Learning Test Delayed Recall, Trail Making Test Parts A and B (Trails A and B), and Boston Naming Test. Using mixed effects regression models, we compared concurrent and longitudinal models to baseline plasma Aβ42/40 or IICV by age interactions. PrecivityAD assays quantified baseline plasma Aβ42/40.

Results: IICV was associated with concurrent/baseline performance on several outcomes but did not modify associations between age and cognitive decline. In contrast, plasma Aβ42/40 was unrelated to baseline cognitive performance, but a pattern emerged in interactions with age in longitudinal models of Trails A and B and Rey Auditory Verbal Learning Test total learning trials. Although not significant after correcting for multiple comparisons, low Aβ42/40 was associated with faster cognitive declines over time.

Discussion: Our results are promising as they extend existing findings to an Black American sample using low-cost, low-burden methods that can be implemented outside of a research center, thus supporting efforts for inclusive AD biomarker research.

FIGURE 1

Model‐predicted log Trails A and B means for varying Aβ42/40 values. Depiction of the model estimated Aβ42/40 by age interaction behavior, showing the model predicted means with 95% CIs for different ages and Aβ42/40 quantiles, while keeping self‐identified sex (female) and years of education (14 years) held constant at their median values at the population level (i.e., unconditional with respect to random effects). Lower percentile represents lower Aβ42/40 ratio. For each test (Trails A and B) two panels display the estimated behavior for model with IICV accounted for and set to the median in the data (main effects and its interaction with age; Panels A and C), versus models in which IICV is not included in any covariate (Panels B and D). Trails A and B performances measured as time to complete. Longer time (higher value) represents worse performance. Figures are annotated to indicate simple age slope (SE) at each level of Aβ42/40 (SE is calculated at the sample mean age of 64 years). A woman with 14 years education whose Aβ42/40 value was equal to the 10th percentile would be expected to take approximately 7 seconds longer to complete Trails A at age 70 then she did at age 60 and 30 seconds longer to complete Trails B. If she had Aβ42/40 values equal to the 90th percentile she would be expected to take 2 more seconds to complete trails A and 2.5 more seconds to complete Trails B by age 70. Aβ, amyloid beta; CIs, confidence intervals; IICV, intra‐individual cognitive variability; SE, standard error; Trails, Trail Making Test, Parts A and B

FIGURE 2

Model‐predicted RAVLT Total of Learning Trials means for varying Aβ42/40 values. Direction of the model estimated by Aβ42/40 by age interaction behavior, showing the model predicted means with 95% CIs for different ages and Aβ42/40 quantiles, while keeping self‐identified sex (female) and years of education (14 years) held constant at their median values, at the population level (i.e., unconditional with respect to random effects). The two panels in the figure display the estimated behavior for models with IICV accounted for and set to the median in the data (main effects and its interaction with age, Panel A), versus the model where IICV is not included in any covariate (panel B). RAVLT performance measured as number correct. Lower number correct (lower value represents worse performance). Figures are annotated to indicate simple age slope (SE) at each level of Aβ42/40 (SE is calculated at the sample mean age of 64 years). A 60‐year‐old woman with 14 years education whose Aβ42/40 was equal to the 10th percentile would be expected to decline by 2 points on RAVLT total by age 70; if the woman's Aβ42/40 was equal to the 90th percentile she would be expected to gain 1 point over 10 years. Aβ, amyloid beta; CIs, confidence intervals; IICV, intra‐individual cognitive variability; RAVLT, Rey Auditory Verbal Learning Test; SE, standard error