Adjustment for Center Effects in Clinical Trials Data Analysis

• Activity Number: 454
• Type: Contributed
• Date/Time: Thursday, August 7, 2003 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Vijay Chauhan*+
• Companies: NJ Biostatistics, Inc.
• Keywords: small centers ; clinical trials

Abstract

Adjustment for center effects is typically performed by fitting the center as a fixed effect in a two-way model with center and treatment effects to evaluate the efficacy of new drugs in multicenter trials. Center is conventionally defined by investigator, location, or region. Some multicenter studies end up with a large proportion of subjects enrolled in a few centers, resulting in many small centers. Small centers with unbalanced data are subjected to confounding with treatment effects. For instance, small centers with more subjects of the better treatment groups may provide a larger estimate for such centers. Conventional approach to deal with small centers is the pooling of small centers, which is not the most appropriate procedure. This paper examines through simulations the possibility of considering the center as a random effect as an alternative to pooling of small centers. It was observed that a model with center as a random effect gave a higher power of treatment comparisons than a model with center as a fixed effect. The random effect model was more efficient when majority of the centers were small.