The ethics of two- and one-sided hypothesis tests for clinical trials

Abstract

In medical research, hypothesis tests are used to determine if a novel treatment is efficacious in comparison to a control treatment. Traditionally two-sided tests are recommended and one-sided tests are not. In this review the arguments for two- and one-sided tests are presented. It is argued that in many applications one-sided tests are appropriate and that two-sided tests may expose research subjects to unnecessary risk, and are therefore unethical.