Depression is a prevalent and disabling disorder. Parental depression predicts onset and treatment resistance in offspring, and adolescent onset has been associated with severe and chronic course. Thus, much work has been done to prevent depression, particularly in adolescent offspring of depressed parents. Few prevention trials have evaluated disorder course. Understanding long-term patterns and predictors of prevention response in at-risk adolescents may inform new intervention development. As such, the current study aimed to empirically derive longitudinal patterns of response to prevention and test predictors of response. Adolescents (N = 316) at personal and familial risk for depression were enrolled in a multi-site randomized controlled trial testing a nine-month group-based cognitive-behavioral prevention program (CBP). The Depression Symptom Rating Scale (DSR) was used to establish patterns of prevention response across 6 years of study participation. Candidate predictors included intervention assignment, current parental depression at baseline, adolescent risk (i.e., a history of depressive episode, subsyndromal depressive symptoms, both), adolescent functioning, anxiety symptoms, and hopelessness. Repeated Measures Latent Class Analysis (RMLCA) was used to identify patterns of response across follow-up. Predictors of class membership were then analyzed using Multinomial Logistic Regression. All analyses employed an intent-to-treat design, used an alpha level of .05, and were run using Mplus8. RMLCA model fit indices (i.e., AIC = 3211.24, BIC = 3476.76, BLMRT: p < .001; entropy = 87%) supported a four-class solution: well/late onset (n = 213; 68.5%), recurrent/brief episodes (n = 38; 12.2%), recurrent/persistent episodes (n = 19; 6.1%), early discontinuation (i.e., n = 41; 13.2%). Multinomial logistic regression revealed that increased odds of categorization in the well/late onset class vs. the recurrent/persistent class were related to being randomized to CBP and having better functioning and less hopelessness at baseline. Interestingly, adolescents at highest risk for depression were also more likely categorized in the well/late onset class compared to the recurrent/persistent class. The four distinct trajectories of prevention response identified have implications for booster session timing and prevention program content. Results also support the feasibility of using RMLCA and multinomial logistic regression in future work, progressing the understanding long-term depression prevention effects in at-risk adolescents.