Comparing neuroimaging and cerebrospinal fluid markers of amyloid, tau, and neurodegenerative biomarkers: implications for understanding the biology of the disease

Abstract

AbstractBackgroundAlzheimer disease (AD) is a complex disorder characterized by changes in amyloid, tau, as well as neurodegeneration. Neuroimaging and cerebrospinal fluid (CSF) assays provide complimentary measures of these pathologies although the Correspondence between such biomarkers is not well elucidated.MethodsData from 313 cognitively unimpaired and 59 impaired individuals aged 45 to 91 was drawn from the Charles F. and Joanne Knight Alzheimer Disease Research Center. Participants had neuroimaging, CSF, and clinical data acquired within one year. Neuroimaging variables included summary measures of amyloid (Centiloids from either PiB or Florebetapir), tau (Flortaucipir), and atrophy (MRI cortical signature). CSF variables were the Aβ42/40 ratio, ptau181, and neurofilament light chain (NfL). Biomarker cutoffs were determined using Gaussian‐Mixture modeling as well as a Youden Index to maximally separate cognitively unimpaired individuals from those with a clinical AD diagnosis. Rank order congruence between biomarkers was determined using Kendall correlations (t).ResultsGaussian Mixture Modeling (dashed lines) and Youden Index (solid lines) approaches produced disparate cutoffs, particularly for neurodegenerative markers (NfL/atrophy) indicating the challenges inherent to binarizing continuous variables (Figure 1). Kendall’s t provide an estimate of pairwise congruence between biomarkers in cognitively unimpaired (CN, CDR=0, upper left diagonal) and impaired individuals (CI, CDR >0, lower right diagonal) (Figure 2). Measures of amyloid had generally high concordance (t ‐0.45 and ‐0.43 for CN and CI respectively ) than those of non‐specific neurodegeneration, which were quite low (t ‐0.16 and ‐0.23). Although ptau181 and tau were moderately related (t 0.21 and 0.23), ptau181 had higher Correspondence with both PET and CSF measures of amyloid. A sequential timing of biomarkers could be clearly seen when plotting z‐scored biomarkers as a function of amyloid PET (Figure 3).ConclusionsNeuroimaging and CSF markers provide complimentary information and help to deeply phenotype the AD biological cascade. While measures of amyloid are highly congruent, non‐specific markers of neurodegeneration have poor agreement., CSF measures of phosphorylated tau likely index a response to amyloidosis rather than being a marker of tau aggregation.