The purpose of this study is to determine the role of alcohol, tobacco, and drug use as predictors of survey panel attrition among an occupational cohort of young adults in the U.S. military. Baseline data on substance use and sociodemographic factors were obtained from 2838 men and women through confidential, self-administered questionnaires while they attended Navy basic training or Officer Candidate School in 1998. Longitudinal follow-up using mailed self-administered questionnaires was begun in 2000. Multivariate logistic regression models were developed to estimate the odds of attrition in relation to baseline substance use. Results revealed that tobacco use was a significant predictor of attrition [Odds ratio (OR) = 1.63; 95% Confidence Interval (CI): 1.37, 1.95]. A significant interaction between level of education and drug use indicated that respondents with less than a college education who were also drug users were at elevated risk for attrition (OR = 2.39; 95% CI 1.09, 5.28). Other significant predictors of panel attrition were male gender and younger age. Alcohol use was not significantly associated with attrition. The findings suggest that tobacco users and drug users with less than a college education may be an important source of nonresponse bias in longitudinal surveys of employed young adults.