Changes in Alzheimer Disease Blood Biomarkers and Associations With Incident All-Cause Dementia

Abstract

Importance: Plasma biomarkers show promise for identifying Alzheimer disease (AD) neuropathology and neurodegeneration, but additional examination among diverse populations and throughout the life course is needed.

Objective: To assess temporal plasma biomarker changes and their association with all-cause dementia, overall and among subgroups of community-dwelling adults.

Design, setting, and participants: In 1525 participants from the US-based Atherosclerosis Risk in Communities (ARIC) study, plasma biomarkers were measured using stored specimens collected in midlife (1993-1995, mean age 58.3 years) and late life (2011-2013, mean age 76.0 years; followed up to 2016-2019, mean age 80.7 years). Midlife risk factors (hypertension, diabetes, lipids, coronary heart disease, cigarette use, and physical activity) were assessed for their associations with change in plasma biomarkers over time. The associations of biomarkers with incident all-cause dementia were evaluated in a subpopulation (n = 1339) who were dementia-free in 2011-2013 and had biomarker measurements in 1993-1995 and 2011-2013.

Exposure: Plasma biomarkers of amyloid-β 42 to amyloid-β 40 (Aβ42:Aβ40) ratio, phosphorylated tau at threonine 181 (p-tau181), neurofilament light (NfL), and glial fibrillary acidic protein (GFAP) were measured using the Quanterix Simoa platform.

Main outcomes and measures: Incident all-cause dementia was ascertained from January 1, 2012, through December 31, 2019, from neuropsychological assessments, semiannual participant or informant contact, and medical record surveillance.

Results: Among 1525 participants (mean age, 58.3 [SD, 5.1] years), 914 participants (59.9%) were women, and 394 participants (25.8%) were Black. A total of 252 participants (16.5%) developed dementia. Decreasing Aβ42:Aβ40 ratio and increasing p-tau181, NfL, and GFAP were observed from midlife to late life, with more rapid biomarker changes among participants carrying the apolipoprotein E epsilon 4 (APOEε4) allele. Midlife hypertension was associated with a 0.15-SD faster NfL increase and a 0.08-SD faster GFAP increase per decade; estimates for midlife diabetes were a 0.11-SD faster for NfL and 0.15-SD faster for GFAP. Only AD-specific biomarkers in midlife demonstrated long-term associations with late-life dementia (hazard ratio per SD lower Aβ42:Aβ40 ratio, 1.11; 95% CI, 1.02-1.21; per SD higher p-tau181, 1.15; 95% CI, 1.06-1.25). All plasma biomarkers in late life had statistically significant associations with late-life dementia, with NfL demonstrating the largest association (1.92; 95% CI, 1.72-2.14).

Conclusions and relevance: Plasma biomarkers of AD neuropathology, neuronal injury, and astrogliosis increase with age and are associated with known dementia risk factors. AD-specific biomarkers' association with dementia starts in midlife whereas late-life measures of AD, neuronal injury, and astrogliosis biomarkers are all associated with dementia.

Figure 1.. Biomarker Rate of Change per Decade From Midlife to Late Life: The Atherosclerosis Risk in Communities Study, 1993-2019

Biomarker changes, standardized to visit 3 (1993-1995), were estimated from linear mixed-effects models. See the Methods section for model adjustments, imputations, and weighting. Intercept indicates differences in SDs at age 60 years, which is 1.4 years after the mean age of 58.6 years at visit 3; slope, the rate of change per decade in SDs; and P values, the difference in intercepts or slopes between subgroups.

Figure 2.. Differences in Biomarker Rate of Change per Decade From Midlife to Late Life: The Atherosclerosis Risk in Communities Study, 1993-2019 (n = 1424)

Change in plasma biomarkers estimated by fitting separate linear mixed-effects models that specified time from visit 3 (1993-1995) included a random intercept and time slope and employed an unstructured variance-covariance matrix. Biomarkers were standardized to visit 3. All risk and protective factors were treated as time-invariant midlife measures. All models were adjusted for time-varying measures of estimated glomerular filtration rate, body mass index, and time-invariant measures of age, sex, race by center, education, and the presence of apolipoprotein ε4 alleles. An interaction was specified between each time-invariant covariate and time. Multiple imputation by chained equations was employed to impute missing covariates. Inverse probability weighting was used to account for selection bias and informative attrition. ^aThe Aβ42:Aβ40 (amyloid-β 42 to amyloid-40) ratio decreases with age; thus, positive values indicate a slower rate of change and negative values, a faster rate. The log₂ values increase with age; thus, positive values indicate a faster rate of change, negative values a slower rate. HDL indicates high-density lipoprotein; p-tau181, phosphorylated tau-181.

Figure 3.. Association of Plasma Biomarkers With Incident Dementia in Late Life: The Atherosclerosis Risk in Communities Study, 2011-2019 (n = 1339)

A dementia diagnosis was determined from adjudicated review of in-person cognitive examinations, telephone interviews, informant interviews, hospitalization records, and death certificates. Hazard ratios (HRs) and 95% CIs for incident dementia were calculated from cause-specific Cox proportional hazards regression models. All models were adjusted for visit 5 measures of age, sex, race center, education, the presence of apolipoprotein ε4 alleles, estimated glomerular filtration rate, body mass index, hypertension, diabetes, total cholesterol, high-density lipoprotein cholesterol, coronary heart disease, cigarette use, and physical activity as time-invariant covariates. Multiple imputation by chained equations was employed to impute missing covariates. Inverse probability weighting was used to account for selection bias. ^aBiomarkers were standardized to visit 3. ^bThe Aβ42:Aβ40 (amyloid-β 42 to amyloid-40) ratio was inverted so that higher values indicate greater risk of incident dementia. ^cBiomarkers were standardized to visit 5. ^dStandardized change in biomarkers measured at visit 3 and visit 5. p-tau181 indicates phosphorylated tau-181 at threonine 181.

Figure 4.. Discriminatory Accuracy of Plasma Biomarkers for Incident Dementia in Late Life: The Atherosclerosis Risk in Communities Study, 2011-2019 (n = 1339)

Biomarkers included the amyloid-β 42 to amyloid-40 (Aβ42:Aβ40) ratio, log₂ phosphorylated tau at threonine 181, log₂ neurofilament light, and log₂ glial fibrillary acidic protein. Receiver operating characteristic curves depict the extent to which plasma biomarkers collectively discriminate incident dementia with or without the apolipoprotein ε4 (APOE ε4) allele or demographics (age, sex, and race by center). A higher area under the curve indicates greater discriminatory accuracy. Differences between panels represent changes caused by using plasma biomarkers from midlife, late life, or change between midlife and late life. Receiver operating characteristic curves were generated from cause-specific, Cox proportional hazards regression models that estimated discriminatory accuracy at the median follow-up time (7.4 years) after visit 5 (2011-2013). Multiple imputation by chained equations was employed to impute missing covariates. Inverse probability weighting was used to account for selection bias.