Abstract
<title>Abstract</title> <p> The clinical significance of the DSM-5 "with mixed features" specifier in adolescent major depressive disorder (MDD)—here termed MDD with mixed features (DMF)—remains uncertain, and its longitudinal course has rarely been examined. In a two-site naturalistic cohort of adolescents with early-course MDD enrolled between May 2024 and April 2025, we tested whether DMF status was associated with differential trajectories of pre-registered outcomes (suicide attempts; depression, SDS) and a post-hoc outcome (anxiety, SAS), and conducted, to our knowledge, the first formal causal mediation test of the pharmacological masking hypothesis—the widely invoked but untested idea that anti-manic prescribing conceals emergent bipolarity. Of 448 baseline participants, 108 (75% female; 34.3% DMF) completed all five scheduled visits (median follow-up 18.1 weeks); the study was retrospectively registered (ChiCTR2400088411). We used linear mixed models with random slopes and Benjamini-Hochberg FDR correction for non-pre-registered outcomes. No suicide attempts occurred, and depression trajectories did not differ by DMF status (β = 0.86, p = 0.311). A post-hoc anxiety signal (DMF × time: β = 2.35, raw p = 0.015) was robust to covariate adjustment but did not survive FDR correction (q = 0.18). Mediation analysis did not support pharmacological masking (ACME = –0.030, p = 0.29). The pre-registered outcomes showed no DMF-related differences; the anxiety signal is reported strictly as hypothesis-generating. Adequately powered, pre-registered replication is required before clinical inference. <bold>Trial registration</bold> <italic>: ChiCTR2400088411, retrospectively registered 19 August 2024.</italic> </p>