Substance Use Transmission and Outcomes: Using Genetically Informative Research Designs for Causal Inference with Observational Data

Abstract

One of the most difficult, yet arguably the most important aspect of research is the issue of causal inference using observational data. For phenotypes like substance use, in which it is impractical or unethical to conduct randomized controlled trials, understanding the causal mechanisms that influence substance use behavior as well as the outcomes caused by these behaviors remains difficult. The current work explores how genetically related samples can be exploited to better understand the causal effects of environmental factors on adult outcomes related to early substance use. In Study 1, polygenic risk scores for alcohol and tobacco use are used to identify a genetic nurture effect of parental smoking initiation on offspring alcohol and tobacco use in a large parent-offspring sample. The effect of parental genotype on offspring use is mediated by parental socioeconomic status (SES), suggesting that rearing SES, or the resources higher SES provide, may causally influence substance use in adolescence. Study 2 is a methodological exploration of co-twin control (CTC) designs, in which an exposure- outcome effect is decomposed into a within-twin pair and between-twin pair effect. A limitation of the CTC design is that it cannot implicitly control for environmental factors that are not perfectly shared within a twin pair, the presence of which may bias CTC findings. We use analytical derivations and simulations to show that while inclusion of a covariate as a proxy measure of a non-shared environmental confounder will always reduce bias, results from CTC studies will continue to be biased away from the null to at least some extent in most practical situations. Interpretation and suggestions for use of CTC, and more generally between-within, models are discussed. Finally, in Study 3 we use a large sample of twins to investigate the adult socioeconomic outcomes related to adolescent substance use. Using the co-twin control (CTC) design we find that within monozygotic (MZ) twin pairs, who share all genetic and common environmental factors, the twin who consumes more tobacco and alcohol in adolescence has lower educational attainment and occupational status in adulthood compared to their lesser using co-twin, consistent with a causal effect of early substance use on later socioeconomic outcomes. We focus on interpretation of these results in the context of findings from Study 2.