ABSTRACT
For the approval of a new drug, the United States Food and Drug Administration (FDA) requires that substantial evidence of the effectiveness of the drug be provided through the conduct of adequate and wellcontrolled clinical trials. In clinical development, multicenter trials are usually considered adequate and well-controlled clinical trials. A multicenter trial is a single study that is conducted simultaneously at more than one study center according to a common protocol. A multicenter trial is often conducted to expedite the patient recruitment process to accrue suf-Œcient number of patients in order to achieve a desired power within a predetermined time frame. The purpose of a multicenter clinical trial is not only to show that the clinical results are reproducible from center to center, but also to establish generalizability of the clinical results from one patient population to another patient population in different geographic locations (Ho and Chow, 1998). As indicated by Chow and Liu (1998b), a multicenter trial is not equivalent to separate single-site trials. The data collected from different centers are intended to be analyzed as a whole. To pool the data for an overall assessment of the effectiveness and safety of the study drug, however, both the FDA and the International Conference on Harmonization (ICH) guidelines require statistical tests for homogeneity across centers in order to detect possible quantitative or qualitative treatment-by-center interaction. A quantitative interaction indicates that the treatment differences are in the same direction across centers, but the magnitude differs from center to center, while a qualitative interaction reveals that substantial treatment differences occur in different directions in different centers (Gail and Simon, 1985). As pointed out by Gail and Simon (1985), no overall statistical inference regarding the treatment effect can be made if there is a signiŒcant qualitative interaction between treatment and center. In this case, it is suggested that treatment effect be assessed by the study center.
