Genetic association analyses of cognitive performance across multi-ancestry older adults: Application of Tobit models

BackgroundPrevious studies investigating associations between genetic variants and late-onset Alzheimer's disease (LOAD)-related cognitive functions included primarily European-ancestry individuals and utilized linear models on neuropsychological test scores with ceiling effects.ObjectiveInvestigate associations between LOAD-related single nucleotide polymorphisms (SNPs) and neuropsychological test scores by applying the superior approach, Tobit (versus linear) regression models, to identify population-specific SNPs associated with cognitive performance across multiple genetic ancestry groups.MethodsNational Alzheimer's Coordinating Center (NACC) Uniform Data Set (UDS) and Alzheimer's Disease Genetics Consortium (ADGC) provided phenotype and genotype data on Alzheimer's Disease Research Center (ADRC) participants, respectively. Using the ADGC genotype data, genetic ancestry groups were identified for ADRC participants, including non-Hispanic White (NHW), African American (AA), Hispanic, and Asian. Tobit and linear models were applied to examine genetic associations of 84 LOAD-related SNPs with cognitive performance at the most recent visit, utilizing the NACC UDS data.ResultsGenetic architectures varied across genetic ancestry groups. The Tobit model detected the association of <i>TMEM106B</i>-rs13237518(A) missed by the linear model. <i>APOE-</i>rs429358(C) was negatively associated with global cognitive function across ancestry groups. Subgroup analyses recognized associations among participants with a cognitive status of dementia: <i>ADAMTS1-</i>rs2830489(T) for Asians and <i>SHARPIN</i>-rs34173062(A) for Hispanics.ConclusionsTobit models demonstrated superior model fit for genetic association analyses of global cognition and language test scores exhibiting ceiling effects.