Blood-based biomarkers of Alzheimer's disease and incident dementia in the community

Abstract

Evidence regarding the clinical validity of blood biomarkers of Alzheimer's disease (AD) in the general population is limited. We estimated the hazard and predictive performance of six AD blood biomarkers for incident all-cause and AD dementia-the ratio of amyloid-β 42 to amyloid-β 40 and levels of tau phosphorylated at T217 (p-tau217), tau phosphorylated at T181 (p-tau181), total tau, neurofilament light chain (NfL), and glial fibrillary acidic protein (GFAP)-in a cohort of 2,148 dementia-free older adults from Sweden, who were followed for up to 16 years. In multi-adjusted Cox regression models, elevated baseline levels of p-tau181, p-tau217, NfL, and GFAP were associated with a significantly increased hazard for all-cause and AD dementia, displaying a non-linear dose-response relationship. Elevated concentrations of p-tau181, p-tau217, NfL, and GFAP demonstrated strong predictive performance (area under the curve ranging from 70.9% to 82.6%) for 10-year all-cause and AD dementia, with negative predictive values exceeding 90% but low positive predictive values (PPVs). Combining p-tau217 with NfL or GFAP further improved prediction, with PPVs reaching 43%. Our findings suggest that these biomarkers have the potential to rule out impending dementia in community settings, but they might need to be combined with other biological or clinical markers to be used as screening tools.

Fig. 1. Blood biomarkers of Alzheimer's disease at baseline.

a, Correlation matrix showing Spearman’s correlations between blood biomarkers of Alzheimer’s disease. b, Baseline biomarker levels by incident all-cause and AD dementia diagnosis. Box plots show the median (center line) and interquartile range (bounds of box) as well as the 2.5th and 97.5th percentiles (whiskers). P values were derived from a two-sided Mann–Whitney test.

Fig. 2. The association between baseline biomarker levels and hazard ratios for all-cause dementia with 95% confidence intervals, in the dementia-free analytical sample.

Hazard ratios (HRs) are derived from Cox regression survival models. Models are adjusted for age; sex; education; disease status, including ischemic heart disease, atrial fibrillation, heart failure, cerebrovascular diseases, chronic kidney diseases, obesity, anaemia, and hypertension; and APOE genotype. Shaded green areas, the distribution of the biomarker levels in the entire population; spikes, incident dementia cases; black line, HR; gray shaded areas, confidence intervals; red line, the reference (HR = 1).

Fig. 3. Blood biomarkers of Alzheimer's disease and prediction of all-cause and Alzheimer's disease dementia.

a–d, Areas under the curves for all-cause (a,c) and Alzheimer’s disease (b,d) dementia, by baseline AD blood biomarker level for single biomarkers and biomarker combinations.

Extended Data Fig. 1. Association between baseline biomarker levels and hazard ratios (HR) for Alzheimer’s disease (AD) dementia with 95% confidence intervals, in the dementia-free analytical sample.

HRs are derived from Cox regression survival models. Models are adjusted for age, sex, education, ischemic heart disease, atrial fibrillation, heart failure, cerebrovascular diseases, chronic kidney diseases, obesity, anaemia, hypertension, APOE genotype. Legend: Green shadow areas represent the distribution of the biomarker levels in the entire population, spikes the incident dementia cases, black line the HR, grey shadow areas confidence intervals, red line the reference (HR = 1).

Extended Data Fig. 2. Flow chart of the study participation.

Created in BioRender. https://BioRender.com/l73u834.

Extended Data Fig. 3. P-tau217 assay precision.

P-Tau 217 levels measured in the commercial control 1, control 2 and pool of sera obtained in the analysis of the SNAC-K cohort (2366 samples divided in 39 runs). Between run CV is indicated for serum pool, control 1 (low) and control 2 (high).