ABSTRACT
This chapter considers the problem of pooling information across study sites in a multicenter clinical trial designed to assess the prophylactic effects of imipramine hydrochloride in preventing the recurrence of affective disorders. Assuming exponentially distributed survival times, twoand three-stage Bayesian hierarchical models are used to model the heterogeneity between study sites. Advantages and disadvantages of each approach are discussed. Laplace approximations are used to calculate posterior distributions. Results show clear quantitative, but not qualitative, interaction between centers and treatment effects.
