How to balance the danger of a risk factor with its benefit.

This article explains different ways of interpreting the results of epidemiological studies especially the measures used to describe the association between an exposure or risk factor and an outcome. Data from cohort or prospective epidemiologic studies are usually presented as rates. 2 measures relative risk and attributable risk are used to compare rates to learn whether a risk factor is associated with a disease. The relative risk is the ratio of the rate of disease or death in individuals exposed to a risk factor divided by the rate in the unexposed and is used to assess whether the association is causal. A harmful exposure produces a relative risk greater than 1.0; relative risks above 4.0 signify strong associations. Attributable risk describes the difference between the rate of disease in individuals exposed to a risk factor and the rates in those not exposed; exposures that increase the risk of disease are associated with attributable risks greater than 0.0. Attributable risk measures the excess hazard associated with a particular exposure and provides a way of balancing the risks with the potential benefits of an exposure while taking into account the frequency of the disease. Heart attacks with a relative risk of 6.4 and an attributable risk of 11.0/100000 woman-years in pill users and hepatocellular adenoma with a relative risk of 26.0 and attributable risk of 3.4/100000 woman years illustrate the differences in the uses and interpretation of the 2 measures of risk. The higher relative risks for both indicate probable causal associations. Because of the rarity of the adenomas a 26-fold increased risk in pill users results in an attributable risk of only 3.4/100000 woman-years suggesting the general principle that an apparently low relative risk for a common disease is more serious than an extraordinarily high relative risk for a rare disease. Results of epidemiologic studies should be interpreted in 2 steps: determining if the association is causal partly on the basis of large relative risk and estimating the importance that the risk factor may have to a patients health using the attributable risk. The attributable risk may be useful in establishing research priorities when it is not known whether an association is causal.