ObjectiveWe aimed to (1) assess and compare baseline plasma and CSF neurofilament light (NfL) for cross-sectional and longitudinal associations with neuroimaging or cognition and (2) determine whether change in plasma NfL corresponded with change in these outcomes.MethodsSeventy-nine participants without dementia, median age 76 years, had plasma and CSF NfL, neuropsychological testing, and neuroimaging (MRI, amyloid PET, FDG-PET) at the same study visit, and a repeat visit (15 or 30 months later) with both plasma NfL and neuroimaging. Plasma NfL was measured on the Simoa-HD1 Platform and CSF NfL with an in-house ELISA. Linear mixed effects models were used to examine the associations between baseline plasma or CSF NfL and cognitive and neuroimaging outcomes adjusting for age, sex, and education. The relationship between change in plasma NfL and change in the outcomes was assessed using linear regression.ResultsThere were no cross-sectional associations between CSF or plasma NfL and any neuroimaging or cognitive measure. Longitudinally, higher baseline plasma NfL was associated with worsening in all neuroimaging measures, except amyloid PET, and global cognition. Higher baseline CSF NfL was associated with worsening in cortical thickness and diffusion MRI. The beta estimates for CSF NfL were similar to those for plasma NfL. Change in plasma NfL was associated with change in global cognition, attention, and amyloid PET.ConclusionElevated baseline plasma NfL is a prognostic marker of cognitive decline and neuroimaging measures of neurodegeneration, and has similar effect sizes to baseline CSF NfL. Change in plasma NfL also tracked with short-term cognitive change.