Relative risk

Usage in reporting

Relative risk is commonly used to present the results of randomized controlled trials. This can be problematic if the relative risk is presented without the absolute measures, such as absolute risk, or risk difference. In cases where the base rate of the outcome is low, large or small values of relative risk may not translate to significant effects, and the importance of the effects to the public health can be overestimated. Equivalently, in cases where the base rate of the outcome is high, values of the relative risk close to 1 may still result in a significant effect, and their effects can be underestimated. Thus, presentation of both absolute and relative measures is recommended.

Inference

Relative risk can be estimated from a 2×2 contingency table:

The point estimate of the relative risk is

${\displaystyle RR=\frac{IE/(IE+IN)}{CE/(CE+CN)}=\frac{IE(CE+CN)}{CE(IE+IN)}.}$

The sampling distribution of the ${\displaystyle \log (RR)}$ is closer to normal than the distribution of RR, with standard error

${\displaystyle SE(\log (RR))=\sqrt{\frac{IN}{IE(IE+IN)}+\frac{CN}{CE(CE+CN)}}.}$

The ${\displaystyle 1-\alpha }$ confidence interval for the ${\displaystyle \log (RR)}$ is then

${\displaystyle C{I}_{1-\alpha }(\log (RR))=\log (RR)\pm SE(\log (RR))\times z_{\alpha },}$

where ${\displaystyle z_{\alpha }}$ is the standard score for the chosen level of significance. To find the confidence interval around the RR itself, the two bounds of the above confidence interval can be exponentiated.

In regression models, the exposure is typically included as an indicator variable along with other factors that may affect risk. The relative risk is usually reported as calculated for the mean of the sample values of the explanatory variables.

Comparison to the odds ratio

The relative risk is different from the odds ratio, although the odds ratio asymptotically approaches the relative risk for small probabilities of outcomes. If IE is substantially smaller than IN, then IE/(IE + IN) ${\displaystyle \approx }$ IE/IN. Similarly, if CE is much smaller than CN, then CE/(CN + CE) ${\displaystyle \approx }$ CE/CN. Thus, under the rare disease assumption

${\displaystyle RR=\frac{IE(CE+CN)}{CE(IE+IN)}\approx \frac{IE\cdot CN}{CE\cdot IN}=OR.}$

In practice the odds ratio is commonly used for case-control studies, as the relative risk cannot be estimated.

In fact, the odds ratio has much more common use in statistics, since logistic regression, often associated with clinical trials, works with the log of the odds ratio, not relative risk. Because the (natural log of the) odds of a record is estimated as a linear function of the explanatory variables, the estimated odds ratio for 70-year-olds and 60-year-olds associated with the type of treatment would be the same in logistic regression models where the outcome is associated with drug and age, although the relative risk might be significantly different.

Since relative risk is a more intuitive measure of effectiveness, the distinction is important especially in cases of medium to high probabilities. If action A carries a risk of 99.9% and action B a risk of 99.0% then the relative risk is just over 1, while the odds associated with action A are more than 10 times higher than the odds with B.

In statistical modelling, approaches like Poisson regression (for counts of events per unit exposure) have relative risk interpretations: the estimated effect of an explanatory variable is multiplicative on the rate and thus leads to a relative risk. Logistic regression (for binary outcomes, or counts of successes out of a number of trials) must be interpreted in odds-ratio terms: the effect of an explanatory variable is multiplicative on the odds and thus leads to an odds ratio.

Bayesian interpretation

We could assume a disease noted by ${\displaystyle D}$, and no disease noted by ${\displaystyle \mathrm{\neg }D}$, exposure noted by ${\displaystyle E}$, and no exposure noted by ${\displaystyle \mathrm{\neg }E}$. The relative risk can be written as

${\displaystyle RR=\frac{P(D\mid E)}{P(D\mid \mathrm{\neg }E)}=\frac{P(E\mid D)/P(\mathrm{\neg }E\mid D)}{P(E)/P(\mathrm{\neg }E)}.}$

This way the relative risk can be interpreted in Bayesian terms as the posterior ratio of the exposure (i.e. after seeing the disease) normalized by the prior ratio of exposure. If the posterior ratio of exposure is similar to that of the prior, the effect is approximately 1, indicating no association with the disease, since it didn't change beliefs of the exposure. If on the other hand, the posterior ratio of exposure is smaller or higher than that of the prior ratio, then the disease has changed the view of the exposure danger, and the magnitude of this change is the relative risk.

Numerical example