Year: 2022
Type: article
Study design: Observational Study
Abstract: Background: Alzheimer's disease (AD) is the most common form of dementia and is estimated to affect more than 46 million people around the world.The most studied biomarkers for AD diagnosis are β-amyloid deposits determined by positron emission tomography (PET), glucose consumption in the brain by PET, structural magnetic resonance imaging (MRI), and cerebrospinal fluid biomarkers such as β-amyloid, tau, and phosphorylated tau.However, these biomarkers present severe problems due to their invasive nature and/or high costs.Consequently, the study of biomarkers in peripheral blood is still of great interest.In addition, a new approach to improve AD diagnosis has evolved in recent years using machine learning and other artificial intelligence (AI) systems.These models often use the mentioned blood biomarkers to predict cognitive decline and AD.In previous studies, we proposed a classification model which uses a set of biomarkers composed of 3 proteins that can be measured in serum samples.These proteins are the regulator of calcineurin 1 (RCAN1), clusterin (CLU), and receptor for advanced glycation endproducts (RAGE).Objectives: This work aimed to validate the relative contribution of blood levels of RCAN1, CLU, and RAGE to the AD diagnosis in a new cohort and to assess the relevance of the contribution of a new parameter, the malondialdehyde (MDA), as a biomarker of oxidative stress.Moreover, a secondary goal was to develop an improved predictive model for AD diagnosis with the addition of clinical parameters such as age, sex, and apoE genotype.At the same time, we studied the within-subject evolution of these bloodbased parameters in a 5-year difference in the longitudinal clinical cohorts.Methods: Blood samples were obtained from the longitudinal and retrospective cohort study composed of cognitively healthy (CTL) subjects over 65 years old, some of which developed mild cognitive impairment (MCI).In addition, blood samples were also obtained from a nursing home cohort composed of dementia and AD patients.From each subject, we obtained serum and plasma samples at baseline and 5 years later, along with the clinical data of age, sex, and apoE genotype.Serum levels of CLU, RCAN1, and RAGE were measured by enzyme-linked immunosorbent assay (ELISA).Plasma levels of MDA were determined by high-performance liquid chromatography (HPLC).In a cohort-based study, we compared the levels of CLU, RCAN1, RAGE, and MDA in each longitudinal group as a case-control study.Additionally, we compared these biomolecules between the stable CTL and MCI converter longitudinal groups at baseline, when both groups were clinically diagnosed as CTL.In a study based on clinical diagnosis, we compared the 4 biomolecules between the transversal clinical groups CTL, MCI, and AD.Moreover, we developed a classification model using CLU, RAGE, RCAN1, and MDA blood levels, and age, sex, and apoE genotype as predictors to classify subjects in positive or negative for AD (AD+/AD-) by binomial logistic regression.A total of 147 observations were used to build the model.To validate these models, fold cross-validation was performed.In addition, ROC curves and other performance metrics such as accuracy, sensitivity, and specificity were determined.Results: In the cohort-based study, a decrease in CLU and RCAN1 occurred in a 5-year interval in AD and CTL subjects correspondingly.No significant changes were observed in RAGE or MDA levels in any cohort.However, when CTL and MCI healthy subjects were compared at baseline, lower levels of RAGE and higher levels of MDA were already detected.In the study based on clinical diagnosis, lower levels of MDA and RCAN1 were detected in the AD group compared to CTL and MCI.No significant differences were found in CLU or RAGE levels.Finally, we build a classification model based on the clinical diagnosis with an accuracy of 89.8 %, a sensitivity of 75.0 %, a specificity of 93.3 %, and an area under the curve (AUC) of 93.9 %.Conclusion: The study of cognitive impairment in CTL subjects that later develop MCI did not show any change in blood concentrations of CLU, RAGE, RCAN1 nor MDA in a 5-year difference.Interestingly, a difference in RAGE and MDA was already detected when comparing stable CTL and MCI converter subjects at baseline, when both groups are clinically indistinguishable.Moreover, when compared at a transversal level, no differences were found between CTL and MCI subjects, whereas AD patients were different in terms of RCAN1 and MDA.Finally, we created a classification model using blood-based parameters, along with the age, sex, and apoE genotype of the subject, which classifies a subject in AD+/AD-with an accuracy of 89.8 %.
Authors: Aurora Román-Domínguez, Mar Dromant, Consuelo Borrás
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
Estimated APC paid: $4,290
OpenAlex ID: W4226162748