e., that participants who did not complete a follow-up assessment were smoking). ITT and responder-only analyses for testing SH versus CI Nilotinib group differences for the three primary outcomes were computed using logistic regression. We also conducted a set of secondary analyses examining study participation in relation to participant sociodemographic characteristics (gender, age, race, and education) and smoking history variables (baseline cigarettes per day [CPD], age when daily smoking started, time to first cigarette after waking, and smoking environment). Study participation was indexed by counseling call completion (no or one counseling calls completed vs. two to four calls) and follow-up call completion (no or one follow-up calls completed vs. two or three calls).

Race was coded as White versus non-White; smoking environment was coded as no smokers present at home or at work versus smokers present at home or work (or both). ��2 tests and t tests were used for group comparisons. Because these were post hoc analyses involving tests on eight different participant characteristics, we used the Benjamini�CHochberg procedure to control the Type I error rate (Benjamini & Hochberg, 1995; Keselman, Cribbie, & Holland, 2002). Lastly, we report results of an exploratory analysis of cessation medication use by study participants. This exploratory analysis was motivated, in part, by conflicting findings about the effects of cessation medication in real-world contexts. While such medications have produced evidence of benefit in some studies, such as those examining medication effects in over-the-counter contexts (Fiore et al.

, 2008; Shiffman and Sweeney, 2008), some survey and longitudinal studies have found little evidence that medication use is associated with greater likelihood of successful cessation (e.g., Messer et al., 2008; Pierce, Cummins, White, Humphrey, & Messer, 2012). Among 278 participants who completed the 1-month follow-up, 72 reported using a cessation medication. In these exploratory analyses, we used ��2 tests to the effect of medication use (yes, no) on the three primary outcomes using the responder-only sample. Because these were post hoc tests on three outcomes, we used a Bonferroni-corrected alpha of .017 to evaluate statistical significance. Also, we GSK-3 used logistic regression to test the interaction of treatment group and medication use. Sample Size and Power to Detect Predicted Effects In the original study protocol, the study sample size was set at 460 to have sufficient power (80%; two-tailed test; �� = 0.05) to detect a statistically significant group difference in predicted 7-day point-prevalence abstinence rates at 6-month postenrollment of 15% in the SH group versus 25% in the CI group.