Epidemiological data on risk factors may be used by clinicians to help diagnose the presence of a disease or assess the probability of its developing in the future. However, the use of risk factor data in this way has not been well evaluated. In this report, relationships are derived between common epidemiological measures of association, such as risk ratios, and common clinically useful measures, such as the operating characteristics of tests. These indicate that even a risk factor with a high relative risk may not be effective when used clinically.