NIA-AA staging of preclinical Alzheimer disease: discordance and concordance of CSF and imaging biomarkers

Abstract

The National Institute of Aging and Alzheimer's Association (NIA-AA) criteria for Alzheimer disease (AD) treat neuroimaging and cerebrospinal fluid (CSF) markers of AD pathology as if they would be interchangeable. We tested this assumption in 212 cognitively normal participants who have both neuroimaging and CSF measures of β-amyloid (CSF Aβ1-42 and positron emission tomography imaging with Pittsburgh Compound B) and neuronal injury (CSF t-tau and p-tau and structural magnetic resonance imaging) with longitudinal clinical follow-up. Participants were classified in preclinical AD stage 1 (β-amyloidosis) or preclinical AD stage 2+ (β-amyloidosis and neuronal injury) using the NIA-AA criteria, or in the normal or suspected non-Alzheimer disease pathophysiology group (neuronal injury without β-amyloidosis). At baseline, 21% of participants had preclinical AD based on CSF and 28% based on neuroimaging. Between modalities, staging was concordant in only 47% of participants. Disagreement resulted from low concordance between biomarkers of neuronal injury. Still, individuals in stage 2+ using either criterion had an increased risk for clinical decline. This highlights the heterogeneity of the definition of neuronal injury and has important implications for clinical trials using biomarkers for enrollment or as surrogate end point measures.

Keywords: Aging; Alzheimer's disease; Amyloid; Biomarkers; Cerebrospinal fluid; Comorbidities; Diagnosis; MRI; Neuroimaging; Neuronal injury; PET; Prognosis.

Figure 1. Overview of NIA-AA preclinical AD classifications

The “−/+” refers to whether a biomarker is normal or abnormal for that stage. As they were highly correlated and concordant, in the integrated classifications we allowed either CSF Aβ_1–42 or β-amyloid-PET to be abnormal for the β-amyloid measure while we differentiated between CSF and imaging abnormality for the neuronal injury biomarker. Abnormal values for biomarkers were: Aβ_1–42 <459 pg/mL, t-tau >339 pg/mL, p-tau >67 pg/mL, MCBP-PiB >0.2245, HCV <−0.3023. Aβ=amyloid beta; CSF=cerebrospinal fluid; HCV=hippocampal volume; PiB=Pittsburgh compound B; SNAP=suspected non-Alzheimer pathophysiology; t-tau=total tau; p-tau=phosphorylated tau.

Figure 2. Survival plots of preclinical AD for progression to CDR ≥ 0.5 based on different biomarkers

Graphs show the survival probability for progression to CDR ≥ 0.5 for each preclinical AD stage, uncorrected for covariates. The black line represents participants in the normal group; blue, Stage 1; red, Stage 2+; and grey, SNAP. Given the relatively small sample sizes for the subgroups in the combined analyses, we allowed either CSF Aβ_1–42 or β-amyloid-PET to be abnormal for the β-amyloid measure while we differentiated between CSF t-tau or p-tau and HCV for the neuronal injury biomarker. CDR=Clinical Dementia Rating scale, HCV=hippocampal volume, SNAP=Suspected Non-Alzheimer Pathophysiology.

Figure 3. Concordance between β-amyloid biomarkers and neuronal injury biomarkers by outcome

Results are concordance between β-amyloid biomarkers and neuronal injury biomarkers presented by outcome, i.e. CDR=0, CDR=0.5, or CDR=0.5 symptomatic AD. Lines are cut-offs used to define abnormality: Aβ_1–42 <459 pg/mL, t-tau >339 pg/mL, MCBP-PiB >0.2245, HCV <-0.3023. Concordant biomarkers are presented in the upper left and lower right part of each figure. K=Cohen’s kappa for agreement in biomarker classification. Aβ=amyloid beta; AD=Alzheimer’s disease; CDR=clinical dementia rating scale; CSF=cerebrospinal fluid; MCBP=mean cortical binding potential; PiB=Pittsburgh compound B; t-tau=total tau.