Interactions and collider bias in case-only gene-environment data

Abstract BackgroundCase-only design for gene-environment interaction (CODGEI) relies on the rare disease assumption. A negative association due to collider bias appears between gene and environment when this assumption is not respected. Genetic risk estimation can quantify part of the predisposition of an individual to a disease. MethodsWe introduce Disease As Collider (DAC), a new case-only methodology to discover environmental factors using genetic risk estimation: a negative correlation between genetic risk and environment in cases provides a signature of a genuine environmental risk marker. Simulation of disease occurrence in a source population allows to estimate the statistical power of DAC and the influence of collider bias in CODGEI. We illustrate DAC in 831 type 1 diabetes (T1D) patients. Results: The power of DAC increases with sample size, prevalence and accuracy of genetic risk estimation. For a prevalence of 1% and a realistic genetic risk estimation, power of 80% is reached for a sample size under 3000. Collider bias offers an alternative interpretation to the results of CODGEI in a published study on breast cancer. ConclusionDAC could provide a new line of evidence for discovering which environmental factors play a role in complex diseases or confirming results obtained in case-control studies. We discuss the circumstances needed for DAC to participate in the dissection of environmental determinants of disease. We provide guidance on the use of CODGEI regarding the rare disease assumption. Key messagesA complex disease is a collider between genetic determinants and environmental factors; it is a consequence of both.Collider bias can affect the results of the case-only design for gene-environment interactions when the disease is common.Using genetic risk estimation, collider bias can be used to discover or confirm the association between a disease and an environmental factor in a case-only setting.Statistical power of this approach is small when the disease is rare. Power increases with sample size, prevalence and genetic risk prediction accuracy.