Signal detection methods were used to develop an algorithm useful in distinguishing those at risk for late relapse from those likely to maintain abstinence. Four subgroups with 24-month survival (nonrelapse) rates ranging from 79% to 33% were identified. Among participants whose depression symptoms decreased from baseline to the end of treatment, lower levels of nicotine dependence were associated with less relapse at the 24-month follow-up (odds ratio = 2.77; 95% confidence interval: 1.36-5.62). Among participants whose depression symptoms increased from baseline to the end of treatment, greater weight gain was associated with less relapse at follow-up (odds ratio = 2.90; 95% confidence interval: 1.41-5.96). This study suggested that it may become possible to use both baseline and treatment information to "titrate" interventions.