Measures of association are means of estimating the strength of the relationship between observed outcomes and factors that may produce the outcome. Common measures of association are the odds ratio, relative risk, absolute risk reduction, and number needed to treat.

Odds ratios are used in studies in which the incidence (or true risk) of disease cannot be accurately assessed (case series, case-control studies, or some retrospective study designs).

Odds ratios are determined by calculating the ratio of the odds of exposure among the cases compared with the odds of exposure among the controls. For example, if the ratio of the odds of exposure to artificial tanning lamps among cases with melanoma and controls is 3:2, the resulting odds ratio is 1.5.

Relative risk specifies the risk of developing the disease in the exposed group relative to those who are not exposed and can be reported from certain prospective study designs and clinical trials.

Relative risk is the best measure of the association between exposure and disease. However, odds ratios can provide robust estimates of association and can approximate relative risk for outcomes that are rare.

The **absolute risk reduction** (ARR) is the difference in disease rates between the exposed and unexposed groups. It is important for clinicians to understand the difference between absolute and relative risk reductions. For uncommon diseases with high relative risks or common diseases with low or moderate relative risk reductions, the **absolute risk reduction **of an intervention may be quite low.

An intervention may have a 10-fold relative risk reduction, with rates of disease near 0.1 for those with the intervention and 1.0 for patients without it [RR=(1.0%)/(0.1%)=10]. However, the same study findings could be described as producing an **absolute risk reduction** of less than 1% (ARR = 1.0% – 0.1% = 0.99%) for those patients exposed to the intervention.

A clinical use of **absolute risk reduction** is to take its reciprocal, which is known as the number needed to treat (NNT). The NNT for an intervention is the number of people who would need to be exposed to the intervention to produce the desired outcome for one person.

For example, if the ARR is 8% when giving clopidogrel to patients getting stents placed in the setting of symptomatic coronary artery disease with the intention of reducing myocardial infarction (MI), the NNT would be 12 (NNT = 1/ARR = 1/0.08 = 12). It follows that with a NNT of 12, one would need to give clopidogrel to 12 patients to prevent 1 MI.

Although 1 MI was prevented, 11 of the 12 patients received no benefit from the intervention; however, all 12 patients needed to be treated because one could not predict which one would benefit in advance. There is no single number that represents a goodNNT.

The clinical impact of NNT is based on the likelihood of the disease, the cost of the intervention (medications, procedures, harm associated with intervention), and the cost of not doing the intervention (rates and values assigned to patient morbidity and mortality associated with the disease).