Component Scores,or Total Scores over Components?

Abstract

Patients' responses to items on a questionnaire are frequently used to assess treatment effects in clinical trials. In this paper, we will use the Unified Parkinson's Disease Rating Scale to address a common question of whether it is more powerful to compare treatments based on the total scores of part II (questions 5-17 on the questionnaire) and part III (questions 18-31) combined or the individual component scores (either part 11 or part III). We will explore the relationships among the powers associated with the various test statistics and display the relationships using graphs. As an illustration, we will apply our findings to help decide on the primary endpoint for a clinical trial in Parkinson's disease patients.