Abstract
Background. Trials that do not allow rejection of the null hypothesis of no treatment effect may have had an inappropriate design. Self-controlled trials, although routinely used for the study of cardiovascular diseases, are virtually never assessed for correlation between treatment modalities.
Methods. Using three models and a series of published studies as examples, the author studies the influence of correlation levels on the statistical power of self-controlled studies.
Results. Between-subject variation as estimated by SD is largely dependent on the level of correlation, and so is the power of testing. The assessment of paired data as though they are unpaired can be used as a simple test to estimate correlation levels.
Conclusion. It is extremely relevant to assess correlation levels in a paired comparison a priori. With a presumably negative correlation a self-controlled design is likely to lack power.
Get full access to this article
View all access options for this article.