Searching for moderators and mediators of pharmacological treatment effects in children and adolescents with anxiety disorders.

Objective: To examine whether age, gender, ethnicity, type of anxiety disorder, severity of illness, comorbidity, intellectual level, family income, or parental education may function as moderators and whether treatment adherence, medication dose, adverse events, or blinded rater's guess of treatment assignment may function as mediators of pharmacological treatment effect in children and adolescents with anxiety disorders.

Method: The database of a recently reported double-blind placebo-controlled trial of fluvoxamine in 128 youths was analyzed. With a mixed-model random-effects regression analysis of the Pediatric Anxiety Rating Scale total score, moderators and mediators were searched by testing for a three-way interaction (strata by treatment by time). A two-way interaction (strata by time) identified predictors of treatment outcome.

Results: No significant moderators of efficacy were identified, except for lower baseline depression scores, based on parent's (but not child's) report, being associated with greater improvement (p < .001). Patients with social phobia (p < .05) and greater severity of illness (p < .001) were less likely to improve, independently of treatment assignment. Blinded rater's guess of treatment assignment acted as a possible mediator (p < .001), but improvement was attributed to fluvoxamine, regardless of actual treatment assignment. Treatment adherence tended to be associated (p = .05) with improvement.

Conclusions: In this exploratory study, patient demographics, illness characteristics, family income, and parental education did not function as moderators of treatment effect. Social phobia and severity of illness predicted less favorable outcome. Attribution analyses indicated that study blindness remained intact. The presence of concomitant depressive symptoms deserves attention in future treatment studies of anxious children.