The Ethics of Underpowered Clinical Trials

To the Editor: Mr Halpern and colleagues¹ conclude that it is unethical to conduct clinical research in humans with a sample size unlikely to reach statistical significance, except in 2 limited circumstances. I disagree for 2 reasons. First, in the past, only methodological transgressions that clearly threatened the welfare of research participants were labeled as unethical; eg, exposing patients to experimental medications without their consent. Halpern et al argue that inadequate power should be ranked with these. This assumes that statistical power is a methodological issue of similar importance. However, in most scales that are used to measure the methodological rigor of randomized controlled trials (RCTs), power is but one of many issues. In fact, other methodological issues have been given at least as much weight as statistical power.² In addition, many statisticians have argued that statistical power is not a prerequisite for validity.³ The validity of an RCT never rests completely on one aspect. For example, a trial that produces a large clinical effect that accords with previous data, shows dose-responsivity, shows convergent validity across measures, and occurs in a generalizable population but has a sample size that resulted in a P value of .10 can be informative and, to me, is not unethical.