Studying the utility of using genetics to predict smoking-related outcomes in a population-based study and a selected cohort.

Abstract

OBJECTIVE\nThe purpose of this study is to examine the predictive utility of polygenic risk scores (PRSs) for smoking behaviors.\n\n\nMETHODS\nUsing summary statistics from the GWAS and Sequencing Consortium of Alcohol and Nicotine use consortium, we generated PRSs of ever smoking, age of smoking initiation, cigarettes smoked per day, and smoking cessation for participants in the population-based Atherosclerosis Risk in Communities (ARIC) study (N=8,638), and the Collaborative Genetic Study of Nicotine Dependence (COGEND) (N=1,935). The outcomes were ever smoking, age of smoking initiation, heaviness of smoking, and smoking cessation.\n\n\nRESULTS\nIn the European ancestry cohorts, each PRS was significantly associated with the corresponding smoking behavior outcome. In the ARIC cohort, the z-score ever smoking PRS predicted ever smoking (odds ratio [OR]: 1.37; 95% confidence interval [CI]: 1.31, 1.43); the z-score age of smoking initiation PRS was associated with earlier age of smoking initiation (OR:0.87: 95% CI: 0.82, 0.92); the z-score cigarettes per day PRS was associated with heavier smoking (OR:1.17: 95% CI: 1.11, 1.25); and the z-score smoking cessation PRS predicted with successful cessation (OR: 1.24: 95% CI: 1.17, 1.32). In the African ancestry cohort, the PRSs did not predict smoking behaviors.\n\n\nCONCLUSION\nSmoking-related PRSs were associated with smoking-related behaviors in European ancestry populations. This improvement in prediction is greatest in the lowest and highest genetic risk categories. The lack of prediction in African ancestry populations highlights the urgent need to increase diversity in research so that scientific advances can be applied to populations other than those of European ancestry.\n\n\nIMPLICATIONS\nThis study shows that including both genetic ancestry and polygenic risk scores in a single model increases the ability to predict smoking behaviors compared to the model including only demographic characteristics. This finding is observed for every smoking-related outcome. Even though adding genetics is more predictive, the demographics alone confer substantial and meaningful predictive power. However, with increasing work in polygenic risk scores, the predictive ability will continue to improve.