Optimizing Quantitative Trait Outcome Measures for Clinical Trials of Chronic Progressive Disease
M. Colin Ard, U. California San Diego  *Steve Edland, University of California, San Diego 

Keywords: Outcome measures, composite scores, Item response theory, Alzheimer's disease

Statistical methods can be remarkably effective at improving the performance of quantitative trait outcome measures used in clinical trials of chronic progressive disease. We illustrate by way of example two applications of statistical methods to optimize performance of outcome measures. The first example uses item response theory to rescaled a commonly applied outcome measure used in Alzheimer's disease treatment trials and reduce sample size requirements by 15% or more. The second example uses generalized least squares methods to combine multiple candidate outcome measures into an optimally efficient single "composite" score. These examples illustrate the potential value that can be added by considered application of statistical methods to instrument development.