Detection of tau-PET positivity in clinically diagnosed mild cognitive impairment with multidimensional features.

Abstract

Abstract Background and purpose: The way to evaluate brain tau pathology in vivo is tau positron emission tomography (tau-PET) or cerebrospinal fluid (CSF) analysis. In the clinically diagnosed mild cognitive impairment (MCI), a significant proportion of tau-PET are negative. Interest in less expensive and convenient ways to detect tau pathology in Alzheimer's disease has increased due to the high cost of tau-PET and the invasiveness of lumbar puncture, which typically slows down the cost and enrollment of clinical trials. This study aimed to investigate one simple and effective method in predicting tau-PET status in MCI individuals. Methods: Based on multidimensional data from MCI participants recruited by Alzheimer’s Disease Neuroimaging Initiative, we used stepwise regression to select the unitary or combination of variables that best predicted tau-PET. The sample included 154 individuals which were dichotomized into tau-PET (+) and tau-PET (-) using a cut-off of >1.33. The receiver operating characteristic curve was used to assess the accuracy of single and multiple clinical markers. The relative importance of predictive variables was judged by nomogram. At last, decision curve analysis (DCA) was used to evaluate the clinical diagnostic value of the best model. Results: The combined performance of four variables [β-amyloid42 (Aβ42) , phosphorylated tau (p-tau), total tau (t-tau), β-amyloid42/β-amyloid40 ratio (Aβ42/40)] in cerebrospinal fluid biomarkers demonstrated the best predictive accuracy of tau-PET status [accuracy=84.3%, area under the curve (AUC) = 0.922], followed by neurocognitive measures using the combination of three variables [Alzheimer’s Disease Assessment Scale-Cognitive Subscale 13 (ADAS-Cog13), Mini-Mental State Examination (MMSE), ADNI-Memory summary score (ADNI-MEM)] (accuracy=85.7%, AUC = 0.879). Structural MRI also showed high accuracy in the middle temporal (accuracy=72.6%, AUC = 0.835). In addition, ADAS-Cog13 (AUC = 0.818) and ApoEε4 genotype (AUC=0.757) were the best independent predictors. The combination of clinical markers model (ApoEε4, neurocognitive measures and structural MRI imaging of middle temporal) had the best discriminative power (AUC=0.946). Conclusions: As a noninvasive test, the combination of ApoEε4, neurocognitive measures and structural MRI imaging of middle temporal accurately predicts tau-PET status. The finding may provide a non-invasive, cost-effective and time-saving tool for clinical application in predicting tau pathology among MCI individuals.