Smoking cessation rates at week 12 were modeled using logistic regression. First, bivariate logistic regression models were fit for each explanatory variable. Then stepwise logistic regression models were used to identify variables prognostic for smoking cessation. Variables were added to the model one at a time in a stepwise method with an inclusion significance level of 0.05. Variables included for selection in the logistic models were minority, treatment arm, age, age started smoking, years smoked, gender, education, marital status, number of cigarettes smoked per day, depression status, FTND, body mass index (BMI), previous quit attempts, previous use of nicotine replacement, and the longest duration of smoking abstinence. To assess whether differences between race groups were dependent on treatment, the race-by-treatment interaction effect was added to the final model determined using the stepwise approach.
