Smoking is highly co-morbid with several psychiatric conditions, but understanding the causal nature of this relationship is complicated by well-described issues of confounding and reverse causality. Mendelian randomisation uses genetic variants associated with an exposure (e.g., smoking) to examine causal pathways between the exposure and outcomes. Previous genetic instruments for smoking have only captured discrete aspects (e.g., initiation, heaviness of smoking), limiting power and requiring individual level data on smoking status for analyses of heaviness of smoking. To overcome these issues, we are developing a novel genetic instrument for comprehensive smoking exposure, which takes into account duration of smoking, heaviness of smoking, time since cessation, and a simulated half-life constant to capture the exponentially decreasing effect of smoking on health over time. Our instrument includes both smokers and non-smokers, removing the need to stratify on smoking status.
We have begun work on this instrument by conducting a genome-wide association study (GWAS) of our comprehensive smoking measure in the UK Biobank (N=463,003) and identified 124 independent SNPs associated at the genome-wide level of significance. Our two-sample Mendelian randomisation validation analysis confirmed that smoking causes lung cancer and coronary heart disease. To further establish the validity of the instrument we need to check that it predicts smoking in an independent sample. Here we hope to use ALSPAC, checking whether a polygenic risk score for lifetime smoking exposure predicts actual smoking behaviour. Secondly, we need to check that the instrument is not spuriously associated with any traits other than smoking. We can do this by checking for associations with other outcomes in ALSPAC.
If the instrument predicts smoking in ALSPAC and is not associated with other unexpected traits, we hope to go onto use our novel genetic instrument to explore bi-directional effects between smoking and mental health, focusing on schizophrenia and major depressive disorder.