A panel of blood lipids associated with cognitive performance, brain atrophy and diagnosis: A longitudinal study of elders without dementia: Biomarkers (non‐neuroimaging) / plasma/serum/urine biomarkers. (7th December 2020)

• Record Type: Journal Article
• Authors: Ma, Ya‐Hui; Yu, Jin‐Tai; Tan, Lan; Dong, Qiang; Xu, Wei; Tan, Lin; Shen, Xue‐Ning
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 4
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-12-07
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.040681
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Blood metabolites have been suggested as promising biomarkers of Alzheimer's disease. Identifying diagnostic and predictive lipids can provide new spots in worse cognitive decline fields. Method: Total 579 plasma lipid signatures were sampled in 374 participants without dementia at baseline enrolled in the Alzheimer's Disease Neuroimaging Initiative cohort over 10 years. Ten‐fold cross‐validated least absolute shrinkage and selection operator‐logistic regression selected a subset of lipids that best classified individuals with worse slopes of cognitive decline determined by linear mixed models. Multivariable adjusted model examined associations between lipids and neuropsychiatric assessments, CSF biomarkers, brain structural measures and diagnostic categories. Result: In a training and a validation set, a panel of seventeen lipids classified cognitively declined individuals with favorable prediction (training set: area under curve [AUC]=0.768, 95% confidence interval [CI]=0.715‐0.821; validation set: AUC =0.747, 95%CI= 0.627‐0.867) and calibration efficacy (Hosmer‐Lemeshow test: p>0.05). Adding risk score to the predictive model yielded a better AUC of 0.779 (95%CI=0.712‐0.845) than the one with clinical variables alone (p=0.014). The model combining polygenic hazard score, lipid signature and clinical variables yielded the best improvement in prediction. Baseline lipid signatures consisting of all selected lipids independently predicted declined cognition and positively associated with baseline cognitive performance (p=0.007) and cerebrospinal tau protein changes (p‐tau, p=0.025; t‐tau, p=0.010). Longitudinal analyses showed lipid signatures had adverse impacts on cognitive performance and brain atrophy over 10 years, and predicted diverse diagnostic categories (case vs. control, stable mild cognitive impairment [MCI] vs. cognitively normal [CN], progressive MCI vs.CN, A+[positive β‐amyloid deposition] vs. A‐[negative β‐amyloid deposition], A+T+ [positive tau pathology] vs. A‐T‐[negative tau pathology]) and a trend for progressive outcomes. Conclusion: A comprehensive panel of peripheral lipids instead of individual lipid molecule could better diagnose and predict cognitive decline. Further, integrating genetic variations, proteins and other biological information in future may expand the fit of exited detect models.