An interaction between haplotype status and age of onset in predicting dependence severity was also obtained via logistic regression analyses (Table 3). A significant interaction effect (P = 0.006) was found, as well as a significant HA vs. HC contrast within the early onset condition (P = 2.0×10−5, OR = 1.82, 95% CI 1.39 to 2.39). HC was associated with reduced likelihood of severe nicotine dependence relative to HA, but only among those smoking daily at age 16 or younger. Gender was associated with age of onset in our cohorts (41% of females began daily smoking before age 17 compared with 49% of males). However, gender did not moderate the relation between dependence and haplotype status within age of onset groups and so was dropped from subsequent analyses. Based on these results for the entire sample, haplotype associations with the dichotomized FTND variable were tested by age of onset condition within each cohort separately (Table 4). There was a significant HA vs. HC contrast in the early onset group for each cohort (P's = 0.02−0.001, OR's = 1.5–3.0), but this
