2 Clinical trials in drug development: Some statistical issues

Abstract

This chapter discusses several statistical issues associated with clinical trials in drug development. As in any scientific experiment, there exist many factors other than the treatments that influence the clinical responses, which are commonly referred to as “endpoints.” Some of these factors cause confounding effects with the treatments and others behave as extraneous sources of variation. The factors of the first group cause a bias in the estimation of the treatment difference; often it is not easy to understand the nature of the bias. The factors of the second group can be divided into two subgroups: patients themselves and the environmental conditions during the course of a trial. A great deal of statistical expertise is required in both theory and computation for designing and analyzing today's clinical trials. Several computer packages such as SAS, BMDP and S-PLUS are available for statistical analysis and are widely used by biostatisticians and data processing people. The industry, contract research organizations (CROs), academia and government employ a large number of biostatisticians whose primary responsibilities are to consult with clinical researchers on statistical designs, prepare analysis plans, analyze data, and interpret results. Biostatisticians in the industry have to work in a somewhat restricted environment because of regulatory requirements.