Do Antidepressants Cause Suicide in Children? A Bayesian Metaanalysis

• Activity Number: 123
• Type: Topic Contributed
• Date/Time: Monday, August 8, 2005 : 10:30 AM to 12:20 PM
• Sponsor: Section on Health Policy Statistics
• Author(s): Eloise Kaizar*+ and Joel Greenhouse and Howard Seltman
• Companies: Carnegie Mellon University and Carnegie Mellon University and Carnegie Mellon University
• Keywords: meta-analysis ; hierarchical model ; depression ; bayesian methods

Abstract

Determining whether antidepressants cause suicide in children and adolescents has been at the forefront of FDA regulatory priorities. To shed light, the FDA collected data from randomized controlled trials in children that included an antidepressant intervention. Among 4,582 young people involved in 24 studies, there were no completed suicides. However, the overall rate of suicidal behavior and ideation was 1.7%. The FDA performed a metaanalysis and concluded the evidence was sufficiently strong to require a black-box warning on the label of all antidepressants that emphasizes the possible risk of increased suicidality in children and adolescents. This warning is expected to reduce antidepressant prescriptions for children. We extend the FDA analysis using Bayesian multilevel models that allow for variability due to different antidepressants and psychiatric diagnoses for which the efficacy was being assessed. We are particularly interested in how statistical issues such as model specification and sensitivity analysis inform the regulatory decision process and what we can learn from metaanalysis about the scientific questions of interest.