91 Markers of neurodegeneration in subjective cognitive decline Dichotomous N biomarkers The proportion of N+ individuals, and hence the distribution of AT(N) categories, strongly depended on the definition of N (Figure 2). Proportions of N+ varied between 10% (NHV10, NNfL90, NGFAP90), and 25% (NHV25, NNfL75, NGFAP75). For Nt-tau and NMTA, proportions of N+ were about 22%. N+ was more common in A- compared to A+ individuals for NMTA or NHV, and more common in A+ compared to A- individuals for NGFAP. For NNfL and Nt-tau, frequencies of N+ were similar between A+ and A-. Figure 2. Distribution of AT(N) profiles according to different definitions of neurodegeneration Distribution of AT(N) profiles for different definitions of neurodegeneration. GFAP 75 = glial fibrillary acidic protein, threshold 75th percentile; GFAP 90 = glial fibrillary acidic protein, threshold 90th percentile; HV 10 = hippocampal volume, threshold 10th percentile; HV 25 = hippocampal volume, threshold 25th percentile; MTA = medial temporal atrophy; NfL 75 = neurofilament light, threshold 75th percentile; NfL 90 = neurofilament light, threshold 90th percentile; t-tau = total tau. Cox proportional hazards analyses using dichotomous N biomarkers to predict clinical progression to MCI or dementia provided overall similar results to analyses with continuous biomarkers, for models 1 and 2 (Table 4). However, only Nt-tau and NHV25 added predictive value to A, and only NHV25 added value beyond A and T. Figure 3 visualizes the combined effect of A and N status for each N on risk of clinical progression in four-level variables (A-N-, A-N+, A+N-, A+N+). Classification of evidence This study provides Class II evidence that HV, NfL and GFAP predicted clinical progression beyond A and T in cognitively unimpaired elderly individuals with SCD. 4