OBJECTIVE: To identify early predictors of suicidal ideation in young adults, and to determine when specific time-varying determinants become important in predicting later suicidal ideation. METHODS: Data were available for 877 participants in the Nicotine Dependence in Teens study, an ongoing prospective cohort of students aged 12 to 13 years at cohort inception in 1999. Time-invariant covariates included age, sex, mother's education, language, and self-esteem. Time-varying covariates included depression symptoms, family stress, other stress, alcohol use, cigarette use, and team sports. Independent predictors of past-year suicidal ideation at age 20 years were identified in 5 multivariable logistic regression analyses, one for each of grades 7, 8, 9, 10, and 11. RESULTS: Eight per cent of participants (mean age 20.4 years [SD 0.7]; 46% male) reported suicidal ideation in the past year. In grade 7, none of the potential predictor variables were statistically significantly associated with suicidal ideation. In grade 8, participation in sports teams in and (or) outside of school protected against suicidal ideation (OR 0.6; 95% CI 0.4 to 0.8; P = 0.002). Depression symptoms in grades 9, 10, and 11 were independent predictors of suicidal ideation (OR 2.2; 95% CI 1.5 to 3.2, OR 1.6; 95% CI 1.0 to 2.5, and OR 1.9; 95% CI 1.1 to 3.4, respectively). No other variables were statistically significant in the multivariate models. CONCLUSION: Depression symptoms as early as in grade 9 predict suicidal ideation in early adulthood. It is possible that early detection and treatment of depression symptoms are warranted as part of suicide prevention programs.