The Proper Size of a Clinical Trial: "Grandma's Strudel" Method

¹ Biometry Branch, National Cancer Institute, National Institutes of Health, Bethesda 14, Maryland.

A well-designed clinical trial will lead to convincing evidence concerning the worth of a medical treatment. To convince, one must not have too few patients. To have failed to convince because of a paucity of patients means that work on man may have been wasted. Any waste of man as an experimental subject is unethical. To use too many patients is also unethical, because it implies that some patients may have been placed on an inferior treatment when the convincing answer could already have been reached.

Statistical procedures have attempted to specify the "Just right" number of patients. However, even the best of these have been criticized recently—mostly because they seem to require too many patients. Some new procedures, other than by "significance levels," are now being developed for designing experiments that could carry conviction. These procedures are preliminary, and they have been given a brief description here. In addition, two graphs and a work sheet have been included to permit estimation of sample sizes needed for a yes-no fixed-size experiment under the most generally accepted current statistical concepts. These graphs and the work sheet should be useful until research leads to better ways of going about clinical trials—of convincing more quickly.

CiteULike    Complore    Connotea    Del.icio.us    Digg    Reddit    Technorati    Twitter    What's this?

This article has been cited by other articles:

				P. L. Twomey Invented Review: Getting Started in Clinical Nutrition Research Nutr Clin Pract, October 1, 1991; 6(5): 175 - 183. [Abstract] [PDF]

				N. I. BERLIN, H. B. ANDERVONT, M. B. SHIMKIN, B. J. KENNEDY, and E. FREI III Breast Cancer: Combined Clinical Staff Conference at the National Institutes of Health Ann Intern Med, August 1, 1965; 63(2): 321 - 341. [Abstract] [PDF]