Multivariate outcome research in arthritis. Implications for sample sizes.

Abstract

Planning appropriate sample sizes prior to data collection is critical to the overall success of a research project. However, the investigator often encounters practical limitations regarding the number of subjects available. While the literature discusses the various options of analysis of clinical trial data, there is little discussion as to the sample size implications of the statistical method chosen. Since the analysis technique employed affects the required sample size, the investigator should select the analysis that is most appropriate for the data and has the most efficient sample size requirement. This paper addresses the sample size implications of three commonly used methods for comparing two populations when there are two or more outcome measures (multivariate outcome). The concepts and methods are illustrated using a Pain Management Study that was conducted using people with rheumatoid arthritis as subjects.