one disorder outside the context of the other. Our secondary analyses using alcohol dependence symptom counts as a covariate suggested that the results observed in this study for CD were not driven solely through association with alcohol dependence. Similarly, there was no overlap in the most highly associated SNPs reported with alcohol dependence (P-value ≤ 1E-06) in the SAGE sample,27 and the most highly associated SNPs with CD, reported in this study. Another limitation is that we did not have sufficient information on the sample to make full CD diagnoses according to DSM-IV; accordingly, we defined case status solely on the basis of the symptom threshold for DSM-IV diagnoses, and without taking into account clustering; this practice has been applied in previous studies.45 Another limitation is that we used retrospective reports of CD. The age range of the sample was quite broad (18–77); accordingly, some participants were reporting on adolescent behavior at a much later time in life. To ensure that the results were not unduly influenced by this, we reran the top hits from Table 2 in the younger half of the sample, as defined by a median split on the age variable (≤ 39 years). The magnitude of
