Genetic Confounding in Observational Studies: Examples from the Canadian Longitudinal Study on Aging

Abstract

Genetic factors influence many complex traits, pointing to potential genetic confounding of observational associations. We investigated the extent of genetic confounding in two settings involving heritable traits: blood lipids and cognitive function, and intraocular pressure (IOP) and retinal vessel traits, using data from the Canadian Longitudinal Study on Aging. We applied the PolygENic Genetic confoUnding INference (PENGUIN) method to estimate associations after accounting for shared genetic factors and compared these with minimally adjusted estimates. Across lipid-cognition associations, PENGUIN adjustment resulted in a median absolute relative 28% change in effect estimates, compared with 13% across IOP-retinal vessel associations. We additionally conducted Mendelian randomization for IOP-retinal vessel relationships, finding evidence that higher IOP is causally associated with lower arteriolar and venular diameter and tortuosity, as well as venular area. Our findings highlight heterogeneity in genetic confounding across trait relationships and support using multiple genetic approaches to strengthen causal inference.