g., Heatherton et al., 1991; Piasecki et al., 2002; Shiffman et al., 1986). This assumption was based on a great deal of previous research that showed that at least some directly elements of measures of nicotine dependence possessed substantial predictive validity. Thus, we included items that assessed smoking heaviness and strength of urges upon awakening (i.e., withdrawal after overnight deprivation). Using theoretical models of relapse (Piasecki et al., 2002; Shiffman et al., 1986), we then systematically examined other factors or domains that were theoretically linked to relapse but not highly correlated or redundant with physical dependence, for example, external events such as stressors and temptations (Piasecki et al., 2002; Shiffman et al.

, 1986) and environmental factors such as the presence of smokers in an individual’s environment or social network (e.g., Derby, Lasater, Vass, Gonzalez, & Carleton, 1994 [in women only]; Garvey et al., 2000; Mermelstein, Cohen, Lichtenstein, Baer, & Kamarck, 1986; Osler & Prescott, 2001). We determined whether items tapping this social�Cenvironmental domain incrementally improved predictive validity beyond that of nicotine dependence items alone. In addition, we determined whether many other types of measures contributed to relapse prediction (see Appendix A). The research described below addresses the ability of a brief questionnaire tapping physical dependence, environmental factors, and individual difference characteristics to predict short- and long-term relapse.

Methods The goal of this research is the identification of a small number of items that collectively function in an effective way to predict relapse following a quit attempt. This secondary data analysis was based on data derived from three randomized, double-blind placebo-controlled clinical trials of smoking cessation treatments. The total sample (N=1,481) was randomly split into two datasets each containing 703 respondents, after removing individuals with missing data, thus allowing for an initial exploratory search of predictive items using one dataset followed by a cross validation of the resulting scale using the second dataset. Participants Sample one. In this study, 608 adult smokers were randomly assigned to one of three experimental conditions: (a) bupropion SR (150 mg, twice daily) and 4-mg nicotine gum, (b) bupropion SR and placebo gum, or (c) placebo bupropion and placebo gum.

All participants also received three 10-min individual counseling sessions. Pharmacotherapy began 1-week prequit and lasted 8 weeks postquit. To be eligible, participants had to report smoking at least 10 cigarettes/day, have an expired carbon monoxide (CO) reading of greater than 9 parts per million Batimastat (ppm), report being motivated to quit smoking, and meet medical and psychiatric inclusion criteria (e.g., no contraindications for bupropion, such as high blood pressure or alcohol dependence).