How to define ‘N’ in the ATN classification in subjective cognitive decline: The SCIENCe project

Abstract

AbstractBackgroundDifferent biomarkers have been suggested to measure neurodegeneration (N) in the ATN classification. We aimed to compare N biomarkers measured with different modalities and assess their association with clinical progression.MethodWe included 401 individuals with subjective cognitive decline (SCD, 61±9y, 42%F, MMSE28±2, follow‐up 3±3y). We used the following biomarkers for N: CSF t‐tau, medial temporal atrophy (MTA) visual rating on MRI, hippocampal volume (HV), serum neurofilament light (NfL) and serum glial fibrillary acidic protein (GFAP). We first compared biomarkers (Pearson correlation) and assessed the association between N biomarkers and progression to mild cognitive impairment or dementia (Cox proportional hazards analysis). We additionally investigated whether the addition of N biomarkers improved a Cox model with CSF abeta and p‐tau as predictors (likelihood ratio test). Last, we compared associations between N biomarkers and MMSE score over time (linear mixed models). Models were corrected for age and sex, and additionally for intracranial volume when HV was used as predictor. P‐tau, t‐tau, NfL and GFAP were log‐transformed, abeta and HV were inverted, and all biomarkers were z‐transformed.ResultN biomarkers were weakly to moderately correlated (range r ‐0.20 – 0.58, Figure 1). Of these, NfL and GFAP, both serum markers, correlated most strongly (r 0.58). T‐tau was strongly correlated with p‐tau, although these biomarkers supposedly represent separate biomarker groups (r 0.89). T‐tau, HV, NfL and GFAP individually predicted clinical progression (range hazard ratios 1.4 – 2.1). Models including abeta and p‐tau improved when HV, NfL or GFAP were added as predictors (t‐tau not entered due to collinearity with p‐tau, Table 1). Last, t‐tau, HV and GFAP were individually associated with MMSE slope (range betas −0.17 – −0.15).ConclusionIn a sample of cognitively normal individuals, correlations between different N biomarkers were low, indicating they do not reflect the same underlying pathology. T‐tau was strongly associated with p‐tau, thereby disqualifying as measure for N. HV, NfL and GFAP predicted clinical progression. Our results do not allow to choose one most suitable biomarker for N, but illustrate room for improved prediction beyond abeta and p‐tau.