Absolute and Relative Measures Provide Different Perspectives

Absolute measures of association, such as risk differences and incidence rate differences, like risk difference and incidence rate difference quantify excess risk or rate of disease in the exposed group compared to the unexposed group, and this provides a measure of the actual number of people who might be affected and the potential public health impact of the exposure, providing a focus on two key questions:

• How much impact would prevention have?
• How many people would benefit?

Relative measures of association, such as risk ratio and rate ratio quantify the relative risk or rate of disease in an exposed group compared to an unexposed group and tend to be thought of as representing the strength of the association between an exposure and a disease by addressing questions such as:

• How much more likely are exposed persons to develop the outcome than the unexposed?

To illustrate consider the annual cumulative incidence death from lung cancer and the annual cumulative incidence of death from heart disease in smokers and non-smokers.

Mortality Rate from Lung Cancer

• Smokers: 140/100,000 per year
• Non-smokers: 10/100,000 per year

Risk ratio = 140/10 = 14
This suggests that smokers have 14 times the risk of dying of lung cancer compared to non-smokers over a year.
Risk difference = 140/100,000 – 10/100,000 = 130/100,000 per year
The same data suggests that if we had prevented 100,000 people from smoking, we would have prevented 130 lung cancer deaths per year.

Mortality rate from Heart Disease

• Smokers: 669/100,000 per year
• Non-smokers: 413/100,000 per year

Risk ratio = 669/413 = 1.6
This suggests that smokers have 1.6 times the risk of dying of lung cancer compared to non-smokers over a year.
Risk difference = 669/100,000 – 413/100,000 = 256/100,000 per year
This suggests that if we had prevented 100,000 people from smoking, we would have prevented 256 deaths from heart disease per year.

A bar chart summarizing these measures of association looks like this:

The bar chart emphasizes that the mortality rate from lung cancer is very low, but smoking is a very strong risk factor since it increases the risk 14-fold. In contrast, mortality from heart disease is more common even among non-smokers. The risk ratio for smoking and death from heart disease is only 1.6, or 60% greater in smokers than in non-smokers, but the risk difference is much greater for heart disease mortality than for lung cancer mortality, because heart disease is much more common. As a result, if we were to eliminate smoking, we would prevent more deaths from heart disease than from lung cancer. So, smoking is a stronger risk factor for lung cancer, but smoking causes more deaths from heart disease.

Relative measures of association are commonly used in journal articles presenting research findings on etiology, but absolute measures of association provide a better idea of public health impact, i.e., the number of people affected.