Research Article

Graphical aid for determining power of clinical trials involving two groups.

BMJ 1988; 297 doi: http://dx.doi.org/10.1136/bmj.297.6649.672 (Published 10 September 1988) Cite this as: BMJ 1988;297:672
  1. D. K. Miller,
  2. S. M. Homan
  1. Department of Medicine, Saint Louis University School of Medicine, Missouri 63104.

    Abstract

    Physicians need to evaluate clinical research critically, and determining the power of a study is an essential component of research evaluation. This report presents a graphical aid that permits rapid power determination for clinical trials with two groups. Power curves were developed for dichotomous outcomes by setting two tail alpha at 0.05 and varying the sample size, the control group response rate, and the clinically important difference between control and experimental groups as defined by the user. Use of the graphical aid was demonstrated to a group of 18 medical students, residents, fellows, and faculty in a 15 minute session. Evaluation of the trainees' application of the aid showed a small average bias of -0.0003 and an average variance of 0.006. Ninety percent of power estimates were within 0.05 of the true value determined by formula. This graphical aid is recommended as a rapid and accurate method for determining power in the critical appraisal of clinical research.