Forty-eight children and adolescents with severe primary obsessive-compulsive disorder completed a 10-week double-blind crossover trial of clomipramine hydrochloride (mean dose [+/- SD], 150 +/- 53 mg/d) and desipramine hydrochloride (mean dose [+/- SD], 153 +/- 55 mg/d). Clomipramine was clearly superior to desipramine in significantly reducing obsessive-compulsive symptoms. Age at onset, duration and severity of illness, type of symptom, and plasma drug concentrations did not predict clinical response to clomipramine. Sixty-four percent of patients who received clomipramine as their first active treatment showed at least some sign of relapse during desipramine treatment. We further document the specificity of the antiobsessional effect of clomipramine and the need for maintenance treatment.