Longitudinal associations of serum biomarkers with early cognitive, amyloid and grey matter changes

Blood-based biomarkers have been extensively evaluated for their diagnostic potential in Alzheimer's disease. However, their relative prognostic and monitoring capabilities for cognitive decline, amyloid-beta (A beta) accumulation and grey matter loss in cognitively unimpaired elderly require further investigation over extended time periods. This prospective cohort study in cognitively unimpaired elderly [n = 185, mean age (range) = 69 (53-84) years, 48% female] examined the prognostic and monitoring capabilities of glial fibrillary acidic protein (GFAP), neurofilament light (NfL), A beta(1-42)/A beta(1-40) and phosphorylated tau (pTau)181 through their quantification in serum. All participants underwent baseline A beta-PET, MRI and blood sampling as well as 2-yearly cognitive testing. A subset additionally underwent A beta-PET (n = 109), MRI (n = 106) and blood sampling (n = 110) during follow-up [median time interval (range) = 6.1 (1.3-11.0) years]. Matching plasma measurements were available for A beta(1-42)/A beta(1-40) and pTau181 (both n = 140). Linear mixed-effects models showed that high serum GFAP and NfL predicted future cognitive decline in memory (beta(GFAPxTime) = -0.021, P-FDR = 0.007 and beta(NfLxTime) = -0.031, P-FDR = 0.002) and language (beta(GFAPxTime) = -0.021, P-FDR = 0.002 and beta(NfLxTime) = -0.018, P-FDR = 0.03) domains. Low serum A beta(1-42)/A beta(1-40) equally but independently predicted memory decline (beta(A beta 1-42/A beta 1-40xTime) = -0.024, P-FDR = 0.02). Whole-brain voxelwise analyses revealed that low A beta(1-42)/A beta(1-40) predicted A beta accumulation within the precuneus and frontal regions, high GFAP and NfL predicted grey matter loss within hippocampal regions and low A beta(1-42)/A beta(1-40) predicted grey matter loss in lateral temporal regions. Serum GFAP, NfL and pTau181 increased over time, while A beta(1-42)/A beta(1-40) decreased only in A beta-PET-negative elderly. NfL increases associated with declining memory (beta(NfLchangexTime) = -0.030, P-FDR = 0.006) and language (beta(NfLchangexTime) = -0.021, P-FDR = 0.02) function and serum A beta(1-42)/A beta(1-40) decreases associated with declining language function (beta(A beta 1-42/A beta 1-40xTime) = -0.020, P-FDR = 0.04). GFAP increases associated with A beta accumulation within the precuneus and NfL increases associated with grey matter loss. Baseline and longitudinal serum pTau181 only associated with A beta accumulation in restricted occipital regions. In head-to-head comparisons, serum outperformed plasma A beta(1-42)/A beta(1-40) (Delta AUC = 0.10, P-DeLong,P- FDR = 0.04), while both plasma and serum pTau181 demonstrated poor performance to detect asymptomatic A beta-PET positivity (AUC = 0.55 and 0.63, respectively). However, when measured with a more phospho-specific assay, plasma pTau181 detected A beta-positivity with high performance (AUC = 0.82, P-DeLong,P- FDR < 0.007). In conclusion, serum GFAP, NfL and A beta(1-42)/A beta(1-40) are valuable prognostic and/or monitoring tools in asymptomatic stages providing complementary information in a time- and pathology-dependent manner.