Comparison of florbetapir positron emission tomography scans with cerebrospinal fluid biomarkers in healthy individuals and people with mild cognitive impairment or Alzheimer's disease dementia

Abstract

We assessed the relationships between florbetapir positron emission tomography (FBP-PET) scans and cerebrospinal fluid (CSF) biomarkers in healthy control (HC), clinically diagnosed mild cognitive impairment (MCI), and Alzheimer's disease (AD) dementia subjects. Data were obtained in August 2012 from the Alzheimer's Disease Neuroimaging Initiative (ADNI), ADNI Grand Opportunity (ADNI-GO), and ADNI-2 studies for all HC, MCI, or AD subjects whose clinical and CSF data had been collected ± 90 days of their FBP-PET scan. Variables included demographic, cognitive, CSF beta-amyloid (Aβ), and tau levels and ratios; and FBP-PET standard uptake value ratio (SUVR) values for composite and 6 regions of interest. Pearson correlation coefficients were calculated among clinical, 5 CSF, and 7 FBP-PET variables. Multinomial logit regression modeling with HC subjects as reference group assessed the association between clinical diagnosis and CSF and/or FBP-PET variables. Likelihood ratio test was used to examine whether adding CSF biomarkers to the model which regresses clinical diagnosis on FBP-PET SUVRs significantly improved model fit, and vice versa. Subjects had mean ages of 76.5 (HC, n=187), 72.6 (MCI, n=320), or 75.5 (AD, n=70) years and were mostly male (50.8%, 58.4%, 64.3%) and white (93.0%, 92.8%, 95.7%). The highest correlations (r >0.60) were observed between FBP-PET anterior and posterior cingulate and composite SUVRs with CSF Aβ, tau/Aβ ratio, and phosphorylated tau/Aβ ratio for HC and MCI subjects, but only between FBP-PET anterior cingulate and composite SUVRs with CSF Aβ for AD. In multinomial logit regression modeling, neither CSF biomarkers (p=0.096) nor FBP-PET SUVRs (p=0.675) significantly added information to explain clinical diagnostic group when in the presence of the other biomarker type. Our findings suggest much overlap among CSF biomarkers and FBP-PET SUVRs in discerning HC, MCI, and AD subjects. More of the CSF and FBP-PET variables showed strong correlations in HC and MCI subjects but fewer for AD subjects. In this ADNI cohort, FBP-PET and CSF seem similarly suited to distinguish between the tested participant groups, with no statistically significant gain in information achieved by combining the measures.