Genetic instrumental variable regression: Explaining socioeconomic and health outcomes in nonexperimental data

Abstract

SignificanceWe propose genetic instrumental variable (GIV) regression—a method that controls for pleiotropic effects of genes on two variables. GIV regression is broadly applicable to study outcomes for which polygenic scores from large-scale genome-wide association studies are available. We explore the performance of GIV regression in the presence of pleiotropy across a range of scenarios and find that it yields more accurate estimates than alternative approaches such as ordinary least-squares regression or Mendelian randomization. When GIV regression is combined with proper controls for purely environmental sources of bias (e.g., using control variables and sibling fixed effects), it improves our understanding of the causal relationships between genetically correlated variables.

A Comparison Of Robust Mendelian Randomization Methods Using Summary Data

“…First, the method requires only summary data on estimates (beta-coefficients and standard errors) of genetic variant-exposure and genetic variantoutcome associations. We exclude methods that require individual participant data [10][11][12][13], and those that require data on additional variants not associated with the risk factor [14,15]. This is because the sharing of individual participant data is often impractical, so that many empirical researchers only have access to summary data, and for fairness, to ensure that all methods are using the same information to make inferences.…”

Using genetic path analysis to control for pleiotropy in a Mendelian randomization study

“…Although polygenic scores may be used for a variety of purposes, a lot of emphasis has been placed on using polygenic scores as instrumental variables. However, as noted and addressed by others (Bowden et al 2015;DiPrete, Burik, & Koellinger, 2018;van Kippersluis & Rietveld, 2017), it is not clear that polygenic scores meet the necessary criteria for a sound instrumental variable.…”

“…The present study focuses on the use of polygenic scores to conduct Mendelian randomization studies, with emphasis placed on reviewing whether polygenic scores meet the criteria for a sound instrumental variable. We then present an extension of genetic instrumental variable regression (DiPrete, Burik, & Koellinger, 2018), genetic path analysis, to help overcome a limitation inherent to Mendelian randomization studies of complex phenotypes, specifically the high potential for pleiotropic effects on the exposure and outcome of interest. Using genetic path analysis, we then test whether educational attainment is associated with body mass index (BMI) and smoking initiation in a large sample of adults while estimating both gene-environment correlation and pleiotropy.…”

“…In response to this concern, a number of methods have been developed to help correct for the presence of pleiotropy. For example, statistical techniques have been developed that are more robust to pleiotropic effects violating the exclusion criteria, including Egger regression (Bowden et al, 2015) and summary data-based multiple regression (Zhu et al, 2018), as well as pleiotropyrobust Mendelian randomization (Van Kippersluis & Rietveld, 2017) and genetic instrumental variable regression (DiPrete, Burik, & Koellinger, 2018). The present study intends to contribute to this body of work by integrating two existing methods, genetic instrumental variable regression and path analysis, to estimate and help control for pleiotropy in a Mendelian randomization study using multiple polygenic scores.…”

Can Genetics Reveal the Causes and Consequences of Educational Attainment?

“…The second assumption is likely to hold in many applications because of the random inheritance of alleles from parents at conception, and it can be falsified by examining the association of SNPs of interest with measured confounders. The third assumption is not directly testable, but there are an increasing number of instrumental variable estimators that are robust to violations of this assumption (Bowden et al, 2016;Davies et al, 2018b;DiPrete et al, 2018;Hartwig et al, 2017;Hemani et al, 2018;Windmeijer et al, 2019). Furthermore, it is important to distinguish between vertical pleiotropy ( Fig.…”