A Systematic Investigation of the Common Genetic Architecture of Substance Use Traits and the Relationship with Mental Health.

INTRODUCTION

Understanding shared and unique genetic vulnerability for substance use traits will give insight into the etiology of their comorbidity with psychiatric and other health outcomes.

METHODS

We applied genomic structural equation modeling to summary statistics from genome-wide association studies (GWASs) to map the genetic architecture of fourteen substance use traits across five substances (tobacco, alcohol, cannabis, cocaine, and opioids) and three stages (initiation, frequency, and dependence). We systematically compared data-driven and theory-based models and conducted sensitivity analyses. We estimated the genetic correlations between the latent factors and psychiatric, physical, socioeconomic, cognitive, and social traits.

RESULTS

A wide range of models could be fit to the data. A data-driven 4-factor model was the most parsimonious, although bifactor models per substance or per stage also fit well. The four factors from the data-driven model were named smoking/other, dependence, alcohol, and cannabis. F1-smoking/other was most strongly related to socioeconomic traits (negatively), whereas F2-dependence and F4-cannabis were mostly related to externalizing and risk behavior (positively). F3-alcohol showed weaker overlap with unfavorable outcomes.

CONCLUSIONS

Genetic risk for substance use traits is partly overlapping, both per substance and per stage of use. Contrasting such categories' liability profiles provides insight into differences in their etiology and associations with poor outcomes.