Keywords: clinical trials, rare disease, responder analysis, power analysis, simulation
Roughly 30 million Americans are affected by one or more types of nearly 7000 rare diseases. For most rare diseases, it is very challenging to conduct clinical trials with adequate power for detecting treatment effects. Therefore, selecting suitable endpoints by utilizing all available information from the study outcome is essential. Our goal is to allow these rare disease trials to be conducted efficiently, but at the same time not lowering the regulatory standard. While binary endpoints can potentially capture individual patients’ improvement, a well-known disadvantage of the responder analysis based on a binary endpoint is the loss of statistical power and information when compared with those based on continuous endpoints. However, whether binary endpoints indeed require larger sample sizes than continuous endpoints depends on the underlying distribution of the data. We will thoroughly examine the pros and cons when utilizing a binary endpoint as opposed to a continuous endpoint. We will demonstrate that under certain conditions, the binary endpoints will outperform the continuous endpoints and have practical advantages. The impact of outliers and dropouts will also be examined.