[Skip to Content]
[Skip to Content Landing]
July 6, 1984

When Is the Evidence for 'No Association' Sufficient?

Author Affiliations

University of North Carolina Chapel Hill

JAMA. 1984;252(1):81-82. doi:10.1001/jama.1984.03350010047022

One of the provocative quirks about the scientific method is that we can never prove beyond doubt that there is no association between an exposure factor of interest and a particular disease. Even in the absence of any true association, the general theory of hypothesis testing practically guarantees that a small number of studies will show some "statistically significant" findings if enough studies are carried out. As a result, we can never be 100% sure that the null hypothesis of no association between exposure and disease is true. What is more, few are the scientists who would want to champion research that seeks to establish "no association." Journals are loath to publish negative results; funding agencies are apathetic; and for those who seek broader acclaim, the public press has little enthusiasm for negative studies. Why then is the article on estrogens and breast cancer in The Journal1 worthy of