Novel fluid biomarkers for brain amyloid and dementia risk in presymptomatic Alzheimer disease

For clinical trials targeting Aβ production or clearance in presymptomatic Alzheimer disease (AD), additional biomarkers beyond CSF Aβ42 and amyloid imaging will be useful. This study sought CSF and plasma proteins that can detect and predict preclinical amyloid, predict subsequent dementia, and provide insight into presymptomatic AD pathophysiology. Paired CSF and plasma samples, collected within 2.5 years of PET-PIB (Positron emission tomography with Pittsburgh compound B) from 118 Knight ADRC participants with normal cognition (Clinical Dementia Rating [CDR] 0), were analyzed using the 190MAP biomarker panel at Myriad/RBM. All participants were APOE-Ɛ3/-Ɛ3 or -Ɛ3/-Ɛ4. Mean cortical PIB binding potential [MCBP] above/below 0.18 (PIB-POS/PIB-NEG) was treated as a categorical variable. Among N=28 PIB-POS at baseline, N=13 developed dementia (CDR>0); N=12 of 67 PIB-NEG participants with multiple scans later became PIB-POS. Receiver operating characteristic (ROC) and C-index analyses were applied with age, gender, education, body mass index (BMI), and APOE-Ɛ4 status as covariates; biomarker combinations (Aβ42-free) were evaluated by logistic regression. BMI and MCBP were negatively correlated (p= 0.03). For detecting PIB-POS, covariate-adjusted ROC analyses identified 7 plasma and 7 CSF proteins with significant areas under the curve (AUC) (each >0.627; max 0.736 & 0.687). CSF tau/Aβ42 yielded AUC=0.923; optimal biomarker panels yielded AUCs of 0.837 (CSF) and 0.814 (plasma). For predicting conversion to PIB-POS, covariate-adjusted ROC analyses yielded significant AUCs for 3 CSF and 9 plasma proteins (each >0.685; max 0.779 and 0.810); optimal panels yielded AUCs of 0.827 and 0.955. For predicting dementia (CDR>0) among PIB-POS, C-index analyses identified 6 CSF and 9 plasma proteins that individually contributed improvements to ‘covariates-only’ models (each >0.778; max 0.825 and 0.873); similar improvements by optimal biomarker panels yielded C-indices of 0.884 and 0.968. Many novel fluid proteins have potential for detecting and predicting presymptomatic brain amyloid, and for predicting subsequent dementia. These candidate biomarkers may have value for clinical prevention trials and provide pathophysiological insights into preclinical AD. Broadly, these results associate presymptomatic amyloid with lower BMI, low insulin, an anti-inflammatory peripheral milieu, and a limited neuroinflammatory response; more robust neuroinflammation at baseline is a harbinger of cognitive decline.