Little is known about the risk of cigarette smoking relapse after 2 or more years of abstinence. The rates and predictors of late smoking relapse were estimated in 483 men who participated in a prospective study for up to 35 years. Subjects are participants in the VA Normative Aging Study, a prospective observational study of aging in men that began in 1963. Subjects are evaluated approximately every 3 years with physical examinations and questionnaires. Smoking, alcohol use, caffeine consumption, and socioeconomic variables were obtained by questionnaire, and weight and height were measured at clinical examinations every 3 years since 1963. Predictors of smoking relapse were identified using proportional hazards regression models. The rate of smoking relapse in the 2nd-6th years of abstinence fluctuated between 2 and 4% per year, and fell to less than 1% only after 10 years of abstinence. In multivariate regression models, coffee and alcohol consumption, and use of cigars or pipes significantly increased the risk of smoking relapse. A small risk of smoking relapse remains for at least 10 years after smoking cessation. Use of other tobacco products, coffee and alcohol increased the risk of late relapse. These findings may be useful in identifying those at highest risk for late relapse and for motivating former smokers to continue long-term abstinence.