83 Markers of neurodegeneration in subjective cognitive decline visual assessment) as biomarker for A. When both amyloid PET and CSF Aβ were available, the PET result was used. We used CSF p-tau (abnormal >52 pg/mL) as biomarker for T. We compared five different N biomarkers: CSF t-tau (abnormal >375 pg/mL), MTA score (abnormal ≥ 1), HV, serum NfL and serum GFAP. We used a cut-off value of ≥ 1 for MTA score instead of age-dependent cut-off values, to be consistent with thresholds for the other biomarkers, which are also age-independent (19). For HV, NfL and GFAP, no established cut-off values were available. Because of varying rates of N+ in literature (12, 20), we pragmatically took the 75th and 90th percentile for NfL and GFAP, and the 10th and 25th percentile for HV, which provides the reader with a range of possible effects sizes. Hence, for HV, NfL and GFAP, we chose two dichotomous definitions per biomarker. The following describe the procedures used to obtain these measures. A lumbar puncture was performed between the L3/L4, L4/L5 or L5/S1 intervertebral space to obtain CSF, which was subsequently collected in polypropylene tubes (21). Levels of Aβ1-42, tau phosphorylated threonine 181 (p-tau) and total tau were measured using sandwich ELISA’s (Innotest Aβ 1-42, Innotest PhosphoTAU-181p and Innotest hTAUAg) (22). CSF Aβ levels were corrected for the drift that occurred over the years (23). For 79 individuals, amyloid PET was performed using the tracers [18F]florbetapir (n=13), [18F]florbetaben (n=48), [18F]flutemetamol (n=7) or [11C]-PIB (Pittsburgh compound-B, n=11). An intravenous cannula was used to administer the tracers. The following systems were used to acquire the PET scans: Gemini TF PET-CT, Ingenuity TF PETCT, and Ingenuity PET/MRI (Philips Healthcare, Best, The Netherlands). For [18F] florbetaben (24) and [18F]flutemetamol (25) imaging, a static scanning protocol was used, for [18F]florbetapir (17) and [11C]PIB imaging (26), a dynamic scanning protocol. A trained nuclear medicine physician visually rated all scans as ‘positive’ or ‘negative’, according to the radiotracer specific product guidelines. Structural MRI 3D T1-weighted images (n=366 (89%)) were acquired as part of routine patient care from nine different systems. The acquisition parameters are described in eText 1. An experienced neuroradiologist reviewed all scans. T1-weighted images were used for visual rating of medial temporal lobe atrophy (MTA; range 0-4). Scores for the left and right sides were averaged (27). Hippocampal volume was estimated using FMRIB Software Library (FSL) FIRST (v5), as described previously (28). The FIRST algorithm first registers the 3D T1-weighted images to the Montreal Neurological Institute 152 template. Next, it uses a subcortical mask for segmentation based on shape models and voxel intensities to obtain hippocampal volumes. Hippocampal volumes were normalized for head size using the V-scaling factor from SIENAX 4