Promising clinical tools for specific Alzheimer disease diagnosis from plasma pTau217 and ApoE genotype in a cognitive disorder unit.

Alzheimer's disease (AD) diagnosis relies on cerebrospinal fluid (CSF) biomarkers or amyloid PET. Alternatives for AD diagnosis from blood samples are needed to develop a fully-automated early-diagnosis approach, potentially implemented in a cognitive disorder unit. Plasma p-Tau217 was determined in patients diagnosed with AD (n = 134) or non-AD (n = 132), from CSF biomarkers (Aβ42/Aβ40). A logistic regression model was developed. The predictive performance was assessed using a training set (70% of data) and internally validated with a test set (30% of data) and 1000 iterations. A nomogram and a double cut-off strategy were proposed to visualize the model results, and stratify patients (AD, non-AD, uncertain), respectively. The model (plasma p-Tau217, ApoE, age) showed satisfactory performance (AUC 0.94, sensitivity 0.85, specificity 0.89); so, together with the corresponding nomogram, it could be applied in specialized clinical contexts. The model including only plasma p-Tau217 (AUC 0.93, sensitivity 0.72, specificity 0.96) would be a useful approach in less specialized clinics. The corresponding two-cut-off strategy for the first model were as AD probability (< 0.41 non-AD, > 0.57 AD). This study provided clinical tools (nomogram, double cut-off) for identifying Aβ positivity at a cognitive disorder unit, which would lead to reduce the CSF analysis.