The comorbidity among the substance use disorders is complex. First, cases from one study are much more likely than controls to have another comorbid substance dependence. For example, when nicotine use disorder is analyzed in the alcohol dataset, 48% of the non-nicotine dependent individuals have alcohol dependence and are included in the control group, and therefore the observed association is tempered. In order to evaluate the observed association between the schizophrenia polygenetic risk score and any substance use disorder diagnosis, we performed a secondary analysis in which we extracted individuals without tobacco use disorder from the combined dataset. This decreased the sample size in the mega-analysis from 3,488 to 1,657 participants. We repeated the testing of the association between polygenic risk score for schizophrenia (pT<0.5) and any substance use disorder, adjusted for study, age, sex, and principal components. Although the association remained statistically significant (p=0.0015), the adjusted odds ratio for the standardized score dropped from 1.55 (95% CI 1.42–1.69) to 1.25 (95% CI 1.08–1.44) and the proportion of variance explained dropped from 3.7% to 0.7%. Therefore, we cannot rule out
