Comparison of nia-aa preclinical Alzheimer's disease staging with CSF and neuroimaging biomarkers


      Using the recently proposed National Institute of Aging and Alzheimer’s Association (NIA-AA) criterion, cognitively normal individuals can be classified into preclinical stages using cerebrospinal fluid (CSF) and neuroimaging biomarkers: abnormal amyloid markers (stage 1), abnormal amyloid and neuronal injury markers (stage 2), and abnormal amyloid and neuronal injury markers and subtle cognitive changes (stage 3). Prior work has utilized either CSF or neuroimaging biomarkers for operationalization, but no analyses have taken an integrated approach. Here we directly compare CSF and imaging biomarkers, use both approaches to assign individuals to preclinical stages, and examine the longitudinal cognitive outcomes of individuals in each stage.


      We included 212 cognitively normal individuals (mean age 66. 8, females=132), with a clinical dementia rating (CDR) of 0. All participants had clinical data and a measurement of hippocampal volume on MRI, abeta1-42 and tau in CSF and amyloid binding on 11[C]Pittsburgh Compound B (PiB) PET, all within one year. Cutoffs to denote abnormality were determined using ROC analysis comparing cognitively normal individuals to those with very mild dementia (CDR=0.5) and a clinical AD diagnosis. Subtle cognitive impairment was defined using a composite of three neuropsychological tests. Using CSF biomarkers alone, neuroimaging markers alone, and using an integrated approach the cohort was classified into Stages 1-3, Suspected Non-Alzheimer Pathophysiology (SNAP), a normal or an unclassified group (Table 1). Longitudinal progression to CDR>0 was modeled using survival analyses.


      Controlling for appropriate covariates, correlations between CSF Aβ 42 and the mean cortical PIB binding potential (MCBP) were significant (r=-.39, p<.0001), while the relationships between adjusted hippocampal volume and CSF (r=.04, p=.52) and ptau (r=.03, p=.62) were not (Figures 1 & 2). Individuals classified as Stage 2 and 3 were at an elevated risk of later dementia relative to those in Stage 0. (Figures 3-5).


      The NIA-AA preclinical stages successfully stratify dementia risk using both CSF and neuroimaging biomarkers. Measures of amyloid were highly congruent, while measures of neurodegeneration were often incongruent. This suggests that such markers cannot be used interchangeably to represent the same pathology.
      Table 1
      N=212CSF BiomarkersImaging MarkersCSF or Imaging
      Stage 0118 (56%]107 (51%)74 (35%)
      Stage 123 (11%)32 (15%)29 (14%)
      Stage 212 (6%)9 (4%)30 (14%)
      Stage 37 (3%)2 (1%)11 (5%)
      SNAP40 (19%)45 (21%)62 (29%)
      Unclassified12 (6%)17 (8%)6 (3%)
      Figure thumbnail fx1
      Figure thumbnail fx2