Improving the Diagnostic Accuracy of Plasma Alzheimer’s Disease Biomarkers with Objective and Subjective Cognitive Screening

Corey Bolton, Vanderbilt University Medical Center, Nashville, United States
Timothy Hohman, Vanderbilt University Medical Center, Nashville, United States
Katherine Gifford, Vanderbilt University Medical Center, Nashville, United States
Kaj Blennow, University of Gothenburg, Gothenburg, Sweden
Henrik Zetterberg, University of Gothenburg, Gothenburg, Sweden
Angela Jefferson, Vanderbilt University Medical Center, Nashville, United States


Blood-based biomarkers have shown great promise as tools for identifying underlying Alzheimer’s disease (AD) pathology and are rapidly moving towards clinical implementation. However, it is not yet clear when along the cognitive spectrum (cognitively unimpaired, subjective cognitive decline (SCD), mild cognitive impairment (MCI), dementia) individuals should be tested, and the validity of these biomarkers at varying disease stages is not well understood. This study evaluates the performance of various blood-based biomarkers for predicting amyloid status at different levels of objective and subjective cognitive functioning to determine when and for whom these biomarkers might be most useful.

Participants and Methods:

Vanderbilt Memory and Aging Project participants (n=155, 72±6 years, 37% MCI) completed neuropsychological assessment, self-report questionnaires, and fasting venous blood draw and lumbar puncture. Plasma levels of glial fibrillary acidic protein (GFAP) and phosphorylated tau(p-tau)-231 were quantified on the Quanterix single molecule array platform. Amyloid positivity was defined as cerebrospinal fluid (CSF) β-amyloid 42/40 ratio <0.072. Cognitive performance was dichotomized into low and high performing groups with the following cutoffs for the low group: Montreal Cognitive Assessment (MoCA) total score ≤26, California Verbal Learning Test-II long delay free recall raw score ≤8. Subjective cognitive decline (SCD) was assessed with single items (“How would you rate your memory?” (dichotomized as “no problems” or “some problems”), “Do you have difficulty with your memory?” (yes/no)). Logistic regression models related blood-based biomarkers to amyloid status adjusting for age, sex, and race/ethnicity. Follow-up models assessed the performance of each predictor in separate groups defined by objective and subjective cognitive functioning (from each dichotomized variable above) to examine differences in concordance index (C-index) between groups.


47% of participants were amyloid positive. In fully adjusted models, higher levels of plasma p-tau231 (C-index=0.72; p<0.0001) and GFAP (C-index=0.75; p<0.0001) were associated with increased likelihood of amyloid positivity. Plasma p-tau231 predicted amyloid positivity better in individuals who endorsed “no problems” with memory (C-index=0.85) than those who endorsed “some problems” (C-index=0.66; ∆C-index=0.19, p=0.02) and in individuals who endorsed having no difficulties with their memory (C-index=0.85) better than those who endorsed memory difficulties (C-index=0.62; ∆C-index=0.23, p=0.01). Results were similar when comparing participants with better or worse objective cognitive performance (p-values>0.81). For plasma GFAP, results were similar when comparing those with and without SCD (p-values>0.92). GFAP predicted amyloid positivity better in individuals who had worse delayed recall performance (C-index=0.83) than those with better delayed recall (C-index=0.67; ∆C-index=0.17, p=0.03) and in individuals with impaired MoCA (C-index=0.82) better than those with a normal MoCA performance (C-index=0.63; ∆C-index=0.19, p=0.02).


In this community-based sample of older adults free of dementia, plasma p-tau231 and GFAP each accurately predicted underlying amyloid status, with differing results at different levels of cognitive functioning. Specifically, p-tau231 performed better in individuals with no cognitive complaints, while GFAP performed better in individuals who already demonstrated some objective cognitive difficulties. These results highlight the heterogeneity of blood-based biomarkers for AD and the importance of neuropsychological measures to aid in identifying the right biomarker, for the right patient, at the right time.

Category: Dementia (Alzheimer's Disease)

Keyword 1: dementia - Alzheimer's disease
Keyword 2: cognitive screening
Keyword 3: self-report