A1998
Title: On assessment of the pre-specified treatment effect in clinical trials
Authors: Yumi Takagi - Doshisha University (Japan) [presenting]
Sae Murakami - Hiroshima University Hospital (Japan)
Hiroshi Yadohisa - Doshisha University (Japan)
Takashi Omori - Juntendo University (Japan)
Abstract: In clinical trials, sample size is statistically determined during the planning stage based on a pre-specified treatment effect, $\delta$, for the primary endpoint. The effect $\delta$ is a decided value based on the results of previous clinical trials on the target population and therefore represents the value of interest to researchers conducting clinical trials. Nevertheless, researchers do not assess the magnitude of this value after the study data are observed and analyzed. Since $\delta$ is a predefined value at the planning stage, it is necessary to assess whether the obtained data and $\delta$ are in agreement for validity. Few clinical trial results have compared the obtained data with $\delta$, and these comparisons have been limited to qualitative assessments using point estimates and 95\% confidence intervals. Three methods for assessing $\delta$ are examined: the S-value, which interprets a p-value as a measure of information; the Bayes factor, which allows comparison of two hypotheses; and the Relative belief ratio, which uses a ratio of the posterior distribution to the prior distribution. These three methods are assessed and compared by applying them to clinical studies, and the resulting interpretations of $\delta$ are discussed.
