What's New
Announcement
Posted on 08.06.02
There's Power in Numbers
By Janet Weiner, MPH
When conducting studies of drugs and devices, researchers rely on "statistical power" to draw conclusions that are scientifically valid. Statistical power is the likelihood that the study will detect a clinically important effect when it truly exists. This likelihood increases as the number of people in the study increases.
In an article published in the July 17 issue of the Journal of the American Medical Association, investigators from the University of Pennsylvania (Penn) CERTs discuss the ethics of conducting trials that don't have enough statistical power to test the treatment, drug, or device in question. Researchers can end up with an "underpowered" trial because of flaws in the study design or because not enough participants are enrolled in a timely fashion.
The Penn CERTs investigators, Drs. Scott Halpern, Jason Karlawish, and Jesse Berlin, note that "underpowered" trials are widespread and that people continue to participate in such research, although they often do not understand its limited clinical value. People often participate in research for altruistic reasons, such as the desire to advance medical science and to help others. The Penn authors argue that researchers must inform study candidates about the limited ability of small trials to produce a public benefit.
If underpowered trials rarely provide solid information, why conduct them at all? Two arguments have been used to support such studies: 1) they can be included in meta-analyses, a special statistical method that pools the results of many separate trials to reach more robust conclusions, and 2) they can suggest the range and scope of effects of treatment.
The Penn CERTs authors refute both of these arguments in their essay, pointing out that retrospective (after-the-fact) meta-analyses have many problems of their own, and that the estimates of treatment effects produced by small trials may be ambiguous or inconclusive.
The authors conclude that underpowered trials are ethical in only two situations:
- Small trials of treatments for rare diseases, when researchers have explicit plans to combine their results with those of similar studies in a prospective meta-analysis, and
- Early trials of drugs or devices, when the results could guide the design of a more definitive (and adequately powered) study.
In either case, the authors stress, researchers must inform study candidates that participation may contribute only indirectly to future health benefits.