A Generic Latent Variable Approach for Measuring Cognitive Reserve: Phenotype Validation and Genetic Association Results

WAGNER, M; MAIER, W; FRULICH, L; WOLFSGRUBER, S; RAMIREZ, A; KORNHUBER, J; PETERS, O; KLEINEIDAM, L; OKONKWO, OC; WILTFANG, J; DALMASSO, MC

Conferencia; AAIC 2018; 2018

Background: Cognitive reserve (CR) is an important protective factor against Alzheimer?s disease(AD) but its biological underpinnings are still poorly understood. Genetic association studies couldincrease our knowledge in this regard but they require a reliable measurement of CR. A promisingapproach is to define CR as the residuum of cognitive performance regressed on pathology (Reed atal. 2010). However, most previous attempts using this operational definition restricted theassessment of pathology to MRI markers and assessed only a single cognitive domain. We present anovel comprehensive CR operationalization, using exploratory structural equation modeling (ESEM).We also present preliminary associations with genome-wide significant AD-related single nucleotidepolymorphisms (SNPs), based on the assumption that some of these SNPs may in fact be related toCR rather than AD pathology. Methods: We used data from 409 patients with MCI from the GermanDementia Competence Network (DCN) who underwent extensive neuropsychological testing,structural MRI and lumbar puncture. Within a single ESEM, multimodal pathology indicators wereaggregated using exploratory factor analysis, cognitive performance was regressed on these factors,and the resulting residuals were integrated into a latent CR-factor. For validation, we tested whetherthis CR-factor would moderate the effect of abnormal CSF-AD-biomarkers (Aß1-42 positive or taupositive) on cognitive decline in 268 longitudinally assessed patients. Associations between the CRfactorand 38 AD-related SNPs were examined using LASSO-regression in 276 patients. Results:Three factors of pathology representing cortical atrophy, lacunar infarction and AD pathology,respectively, and a CR- factor were derived with an excellent model fit (Figure 1). As hypothesized,the CR- factor significantly attenuated the effect of abnormal CSF-AD-biomarkers on cognitive decline(p=0.02, Figure 2). Finally, a protective SNP in GAB2 was associated with CR using LASSO-regression.Importantly, GAB2 was unrelated to the latent factors of pathology. In contrast, APOE4 was associatedwith AD-pathology, but not with CR. Conclusions: ESEM offers a generic and versatile approach toderive a measure of CR incorporating multimodal pathology assessment and extensive cognitive testdata. GAB2 might be a candidate gene for CR rather than AD pathology, but replications in largersamples with multimodal assessments (e.g. ADNI) are needed.