SummaryObjective The noncompetitive α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor antagonist perampanel was shown in phase III trials to be an effective and well-tolerated adjunctive treatment for partial-onset seizures. In adolescents, it is necessary to characterize cognitive, neuropsychological, and behavioral side effects of antiepileptic drugs (AEDs). The current analysis focuses on behavioral outcomes, efficacy, and safety of perampanel in adolescents. Methods Adolescents (12–17 years) on a stable regimen of 1–3 AEDs for partial-onset seizures were randomized (2:1 ratio) to receive up to 12 mg/day perampanel or placebo. Alongside efficacy, cognitive, and neuropsychological assessments, behavioral outcomes were measured with the Child Behavior Checklist (CBCL) before and after a 19-week titration and maintenance phase. Results Of the randomized patients, 85 received perampanel and 48 received placebo. Median reduction in seizure frequency from baseline was 58.0% for perampanel and 24.0% for placebo (p = 0.079). More patients had seizure frequency reduced by 50% after perampanel (n = 49 [59.0%]) than placebo (n = 17 [37.0%]; p = 0.0144). Changes in behavior were minimal, and there were no differences between groups on competency (p = 0.619) or problems (p = 0.174). A greater proportion of placebo patients were classified in the CBCL “clinical” range for competency at end of treatment, whereas the number in the perampanel group remained unchanged. The overall safety profile was similar to that reported previously for perampanel; most frequently reported adverse events (AEs) were dizziness (26 patients [30.6% vs. 14.6% placebo]), somnolence (13 patients [15.3% vs. 4.2%]), and headache (nine patients [10.6% vs. 14.6%]). Aggression was reported in seven patients receiving perampanel (8.2% vs. 2.1% placebo); two of these were serious AEs, with neither requiring treatment discontinuation. Significance Adjunctive perampanel is efficacious and well tolerated in adolescents with partial-onset seizures, and appears to have no clinically important impact on behavior measured using the CBCL.