Adults with major depressive disorder (MDD) demonstrate certain sleep polysomnographic abnormalities, including sleep continuity disturbances, reduced slow-wave sleep, shortened rapid eye movement (REM) latency, and increased REM density. Findings of sleep EEG studies in depressed children and adolescents have yielded conflicting results, possibly because of methodological variations across the studies. Generally, however, studies have demonstrated that depressed children and adolescents exhibit less sleep continuity and non-REM sleep differences in comparison with control subjects than do adults. Thus, results from adult sleep polysomnography studies cannot necessarily be generalized to children and adolescents. Depressed adults who have reduced REM latency during the symptomatic episode appear more likely to have a relapse once treatment is discontinued than those with normal REM latency. No studies of the relationship between sleep polysomnographic variables and clinical course have been reported in depressed children and adolescents. Data for baseline clinical variables and 3 nights of sleep polysomnography were examined in 113 depressed children (< or = 12 yr; n = 51) and adolescents (> or = 13 yr; n = 62) (56 in-patients and 57 outpatients) where data was available on at least 1 yr of naturalistic follow-up. Subjects came from 2 studies of sleep polysomnography in children and adolescents with MDD. Clinical course was assessed using the Kiddie-Longitudinal Interval Follow-Up Evaluation (K-LIFE). This interview was used to define recovery from the index episode of MDD and recurrence, a new episode of meeting full criteria for MDD. Clinically, within 1 yr of initial evaluation 102/113 subjects had recovered from their index episode of depression (minimal or no symptoms for 60 d). Of the 102 subjects who recovered, 36 (35.3%) had a recurrence of MDD. The majority of subjects (55%) who had a recurrence were not on medication at the time of recurrence. Subjects who had a recurrence were more likely to report suicidal thoughts or attempts at baseline compared to those without a recurrence (67 vs. 37%; F = 8.77; p = 0.004). On baseline sleep polysomnography, subjects with a later recurrence had decreased sleep efficiency and delayed sleep onset (sleep latency > 10 min). Probability of recurrence at 12 months was 0.39 compared to 0.15 in subjects with non-delayed sleep onset (p = 0.005). Baseline suicidal ideation and sleep dysregulation on sleep polysomnography predicted recurrence in a large sample of depressed children and adolescents. Depression in children and adolescents is frequently a chronic, recurrent illness. Factors that can predict clinical course are important in increasing our understanding of depression in this age group.