Plasma neurofilament light, glial fibrillary acid protein, and phosphorylated tau 181 as biomarkers for neuropsychiatric symptoms and related clinical disease progression

Neuropsychiatric symptoms (NPS) are common in older people, may occur early in the development of dementia disorders, and have been associated with faster cognitive decline. Here, our objectives were to investigate whether plasma levels of neurofilament light chain (NfL), glial fibrillary acid protein (GFAP), and tau phosphorylated at threonine 181 (pTau181) are associated with current NPS and predict future NPS in non-demented older people. Furthermore, we tested whether the presence of NPS combined with plasma biomarkers are useful to predict Alzheimer's disease (AD) pathology and cognitive decline.

Plasma GFAP is associated with NPS while NfL and GFAP are both associated with future NPS and NPS severity. Considering the presence of NPS along with blood-based AD-biomarkers may improve diagnosis and prediction of clinical progression of NPS and inform clinical decision-making in non-demented older people.

One hundred and fifty-one participants with normal cognition (n=76) or mild cognitive impairment (n=75) were examined in a longitudinal brain aging study at the Memory Centers, University Hospital of Lausanne, Switzerland. Plasma levels of NfL, GFAP, and pTau181 along with CSF biomarkers of AD pathology were measured at baseline. NPS were assessed through the Neuropsychiatric Inventory Questionnaire (NPI-Q), along with the cognitive and functional performance at baseline and follow-up (mean: 20 months). Linear regression and ROC analyses were used to address the associations of interest.

Higher GFAP levels were associated with NPS at baseline (β=0.23, p=.008). Higher NfL and GFAP levels were associated with the presence of NPS at follow-up (β=0.29, p=.007 and β=0.28, p=.007, respectively) and with an increase in the NPI-Q severity score over time (β=0.23, p=.035 and β=0.27, p=.011, respectively). Adding NPS and the plasma biomarkers to a reference model improved the prediction of future NPS (AUC 0.73 to 0.84, p=.007) and AD pathology (AUC 0.79 to 0.86, p=.006), but not of cognitive decline (AUC 0.79 to 0.84, p=.068).