Disentangling the causal effects of education and participation bias on Alzheimer’s disease using Mendelian Randomization

Abstract

AbstractBackgroundHigher education has been associated with reduced risk of Alzheimer’s Disease (AD) based on previous epidemiological research. However, the relationship between education and AD could be influenced by participation bias whereby people with higher education may be more likely to be enrolled in studies. In this study, we use two‐sample Mendelian randomization (MR) to investigate the potential effects of education and participation bias on AD and further investigate their relationship.MethodGenome‐wide association summary (GWAS) statistics were obtained from GWAS’s of educational attainment (n = 1,100,000), participation in an optional mental health questionnaire (MHQ) of the UK Biobank (n = 451,036) and AD (n = 94,437). Independent genome‐wide significant (r2 = 0.001, 10Mb window, p<5×10−8) single nucleotide polymorphisms (SNPs) were extracted from the exposure summary statistics and harmonized with the outcome SNPs. Inverse variance weighted (IVW) meta‐analysis was the primary MR method and sensitivity analyses included MR Egger, Weighted Median and Weighted mode. Cochran’s Q statistic was used to test for heterogeneity, MR Egger intercept for pleiotropy, and radial MR to identify outliers.ResultGenetically predicted higher educational attainment was causally associated with a reduced risk of AD (IVW OR [95% CI]: 0.691 [0.548, 0.832], p = 3.29e‐07). Similarly, participation in the MHQ was also associated with a reduction in AD risk (IVW OR [95% CI]: 0.987 [0.975, 0.998], p = 0.031]. Our analysis showed no evidence of heterogeneity or pleiotropy and radial MR analysis did not identify any significant outliers, indicating that our univariate MR analyses were robust.ConclusionUnivariant MR analyses indicated that education and participation were both associated with a reduced risk of AD. However, further investigation using multivariable MR is necessary to fully evaluate their independent effects on AD through multiple clumping and proxy variant identification steps. This will allow us to see if the protective effect of education on AD is mediated by participation bias.