We’ve all read headlines like the following:
“Pfizer rare heart disease drug reduces risk of death by 30 percent in study.”
“Eating processed meat daily increases the risk of colorectal cancer by about 18%.”
“Ditching Meat can help reduce cancer risk by 40% says new World Cancer Research”
Pretty scary huh? Makes you want to pop pills and stop eating meat immediately. Though like most numbers, these numbers can be deceiving. These percents are what are known as “relative risk” percentages also often represented by the abbreviation R/R. These percentages are ratios comparing a subject to a control group . The numbers are largely derived from observational studies also known as epidemiological studies or cohorts. The purpose of such studies is to find associations for further more in depth controlled research. Thus relative risk numbers are “associated” relative risks. The strength of the association is what’s important. 30 or 40% might sound like a lot, but it really isn’t. The associated relative risk with smoking and lung cancer is 2500 to 3000%.
What follows is a more detailed explanation of how relative risk is derived and how it differs from absolute risk. This was something I wrote a few years back, and has been circulated a bit unclaimed on the internet. It is pretty amazing how many doctors and science writers don’t seem to know the difference between absolute and relative risk. Too often relative risk is used to fear monger and or push drugs and food religions.
Absolute Versus Relative Risk- What’s the Difference?
The best way to answer this question is with an example (or two).
Experiment 1: A study is conducted to see which of people is more likely to get colorectal cancer. The study has two groups of people: One of a 100 people that eats red meat and another of 100 people that don’t eat meat. This was the only distinction between the two groups.
Experiment 2: A larger study is conducted to see which of people is more likely to get colorectal cancer. The study has two groups of people: One of a 1000 people that eats red meat and another of 1000 people that don’t eat meat. In this study, none of the participants smoked or drank. Additionally none of the participants were obese and all of the participants exercised at least 3 times per week. This second experiment tries to account for “healthy user bias” that is variables (confounding factors) more associated with healthy lifestyles
Methods both Experiment 1 and Experiment 2: Over the course of 10 years, every two years, both groups respond to food questionnaires recalling what they ate and are monitored for colorectal cancer. So essentially these are not rigorously controlled studies especially since food recall questionnaires are notoriously inaccurate. Thus there are numerous other variables (confounding factors) that may impact the results.
Results Experiment 1: At the end of this 10 year period of time, 4 people out of the 100 meat eaters (4%) had gotten colorectal cancer. In the non-meat eating group, 3 out of 100 people (3%) got colorectal cancer.
Results Experiment 2: At the end of this 10 year period of time, 16 people out of the 1000 meat eaters (1.6%) had gotten colorectal cancer. In the non-meat eating group, 12 out of 1000 people (1.2%) got colorectal cancer.
For Experiment 1, what was the relative risk for getting colorectal cancer?
Relative risk is the relative difference between the two outcomes of the two groups. So 4 people got cancer in the first meat eating group compared to only 3 in the non-meat eating group. So that’s 4/3 or 1.333. Thus the relative risk increase is 33%
For Experiment 2, what was the relative risk for getting colorectal cancer?
Doing that with the numbers for the larger 2nd study, you get 16/12 which also equals 1.33 or a relative risk increase of 33%
For Experiment 1, what was the absolute risk for getting colorectal cancer?
On the other hand the absolute risk is the is the difference in percent risk between the two groups, so that 4% for meat eaters minus 3% for non-meat eaters or an absolute risk of 1%
For Experiment 2, what was the absolute risk for getting colorectal cancer?
Doing that with the numbers for the larger 2nd study, you get 1.6% minus 1.2% or an absolute risk increase of 0.4%
Discussion: Now relative risk is more sensational (33% is scarier than 1% or less), so that’s what gets picked up by the media since it generates more page views. But in reality, the number is somewhat meaningless. What’s actually important is the absolute risk number. In the case of meat studies for various diseases, this absolute risk number is typically very low. Remember too, due to the methodology, not all factors are accounted for even in Experiment 2. The source of the meat, what the meat was consumed with or maybe even other foods consumed (sugar, vegetable oils, etc.) may have been the underlying cause of the cancer rather than meat in general. Such cohorts or observational (epidemiology) studies are therefore constrained and are really only meant to generate associations to further research, since the bottom line is that associations do not equal causation.