We performed the power analysis to demonstrate 1) the sufficient statistical power in testing pleiotropy, 2) the limited power in testing interaction, and 3) the power of contemporary sample sizes in answering these important questions. Understanding the statistical power is especially critical when interpreting the null association results in our study. These negative results need to be interpreted in the context of sample size, frequency and hypothesized effect size of genotype, prevalence and relative risk of environmental factors, and type I error level (Cohen, 1988; Gauderman and Morrison, 2006). We found no evidence for the genetic risk moderation by examined psychiatric disorders, but the null moderation results should be interpreted with caution as we had compromised power to detect an interaction effect size smaller than an odds ratio of 3.6. We can only state that there are not likely to be large (Interaction OR>3.6) moderating influences of psychiatric comorbidity on the genetic risk of developing nicotine dependence. Extremely large samples sizes, in the order of tens of thousands of individuals, will be needed to detect a modest moderation effect if it is present.
