Biostatistics Journal Club
November 1st, 2016
12:00p.m. – 1:00p.m.
COPH 2280
Fei Wan, Ph.D.
Assistant Professor, UAMS Department of Biostatistics
A Comparison of Statistical Methods for Pre-Post Study Designs
Abstract
Many randomized trials use a pre-post design to compare the effects of competing treatments on a continuous outcome measured at baseline and after treatment. Commonly used statistical methods for such design include analysis of variance (ANOVA) on post-treatment measure only, ANOVA on the difference between the pre- and post-treatment measures (“change score”), analysis of covariance (ANCOVA) with post-treatment measure as outcome and pre-treatment measure as covariate, and repeated measure ANOVA. ANCOVA is routinely recommended as the method of choice because of its simplicity and superiority in terms of statistical power. However, it is unclear if ANCOVA can retain this superiority over repeated measure ANOVA in presence of missing data. A simulation study is designed to compare the statistical power and type I error rate of each method for a hypothetical randomized trial with two arms under a range of correlations between baseline and post-treatment measures and different dropout rates.
- Vickers AJ. The use of percentage change from baseline as an outcome in a controlled trial is statistically inefficient: a simulation study. BMC Med Res Methodology 2001; 1:6