Relative Risk
Is an exposure related to a particular outcome? We can measure the incidence of that outcome in the exposed group vs in the unexposed group. In this video we look at the incidence of heart disease in smokers and non-smokers.
Relative Risk and the 2×2 table
The 2×2 table is the most common way of calculating relative risk. Traditionally, you’ll put the exposure/no-exposure down the left side and disease/no-disease across the top.
| DiseaseNo Disease | |
| Exposed | ab |
|---|---|
| Not Exposed | cd |
I would recommend instead of memorizing the letters in the formula, try to understand what it represents:
- a/(a+b) = the risk of getting the disease in the exposed population
- c/(c+d) = the risk of getting the disease in the unexposed population
- a/(a+b)/c/(c+d) = the risk of getting the disease in the exposed population over the risk in the unexposed population
- Relative risk = a/(a+b)/c/(c+d)
Absolute Risk Reduction and Relative Risk Reduction
When comparing risk in two different groups, perhaps between one who took a drug and another who didn’t take the drug (perhaps taking a placebo instead), we want to see how much the risk was reduced. So there are two ways to measure risk reduction:
- Experimental/Exposed Event Rate (EER) is the risk of getting the positive (or negative) event in the experimental (or exposed) population. This is expressed as a percentage.
EER = (# in experimental group with event)/(total # in experimental group) - Control Event Rate (CER) is the risk of getting the positive (or negative) even in the control group (or non-exposed group). This is also expressed as a percentage.
CER = (# in control group with event)/(total # in control group) - Absolute risk reduction (ARR) = the risk in the unexposed population minus the risk in the exposed population
- Relative risk reduction (RRR) = (1 – RR) Remember that RR is the risk in the exposed populated divided by the risk in the unexposed populations
The relative risk reduction often looks like a bigger number than the absolute risk reduction. Be wary of this, a drug company may advertise a high risk reduction! But you know better! You’ll ask if this is an absolute or relative risk reduction.
Number Needed To Treat
No treatment works for everyone, it only works for some people. So how many people do you need to treat in order to find one for whom the treatment works? This is usually expressed over a particular time frame. For example, you may need to treat 6 patients in order to have 1 patient with a beneficial effect.
NNT = 1 / ARR
Test your comprehension
With this risk problem set.
patwari 