Here’s we go again. so-called nutritional “scientists” and doctors are out there using relative risk to fear monger against eating red and processed meat. They’re more vociferous than usual since their weak science has been called out for what it is, very weak, in recently published articles and papers in the Annals of Internal Medicine. Here’s an example of such fear mongering from a recent Washington Post article where nutritional epidemiologist Walter Willett from Harvard is referenced:
“Willett says the panel’s conclusions and recommendations do not reflect the study’s findings. Their meta-analyses of large cohorts showed that dietary patterns with a moderate reduction in red and processed meat consumption were associated with lower total mortality by 13 percent. If a drug brought down the number of deaths to that degree, he says, it would be heralded as a success.”
Thirteen percent! THIRTEEN PERCENT !!!! Scary, huh? That is until you realize that this is a relative risk number and is quite meaningless unless you know what the absolute risk is (Noordzij et al, 2017). In a prior blog post, I went over the difference between absolute and relative risk in some detail giving examples of how both are derived. As a quick refresher as to why the relative risk is meaningless without knowing the absolute risk, here’s another simplified hypothetical example:
A study has a control group and a case group. The control group didn’t eat meat, and the case group did. Over the course of twenty years, (7) seven people in the control group died, while (8) eight people in the case group died. They died from a variety of causes. Thus total or all cause mortality.
So what’s the relative risk? That again is the control group divided by the case group or 8 divided by 7 = 1.14 or 14%
What’s the absolute risk? You can’t tell from this example. Why? Because you need to know the size (N) of each group. So let’s use some different group sizes as examples. For simplicity’s sake, let’s use a N of ten (N=10) for each group, and then an N of one hundred (N=100) and finally a N of one thousand (N=1000)
After twenty years of observation, with an N=10, eight out ten people died in the case group and seven out of ten died in the control group. The R/R is still 8/7 or 1.14 (14%). But the absolute risk is 80% – 70% or 10%. Now with an N=100, the R/R is still 8/7 or 14% but the absolute risk is 8% minus 7% or 1%. One more time with an N=1000, the R/R is again 8/7 or 14%, but now the absolute risk is only 0.8% minus 0.7% or 0.1%
So again without seeing the data and knowing the size of the respective control and case groups. and therefore not knowing the absolute risk, the relative risk is completely meaningless.
Let alone with “total” or all cause mortality, people in either group could have died from accidents or suicide or a myriad of other causes. Now even if the cause of death was known and more specific (e;g colorectal cancer), there are a myriad of other variables between people’s diets, food sources and lifestyles that can’t be controlled in observational (epidemiological) studies no matter how these studies are statistically “adjusted” to control for “confounders”. Plus, even the gathering of data for such studies through food frequency questionnaires and or food diaries is very flawed and thus very dubious.
Unfortunately, nowadays people spend a lot of money to attend schools and universities to be taught what to think rather than HOW to think. Anyone capable of actually thinking for his or herself already knew the observational nutritional science was associative and that the associations were very very very weak for red/processed meat and various diseases. Thus such observational science is only intended to generate hypothesis for future research. It should NOT be used to form public or institutional policy. The only exception to this is when the strength of the association is very strong as was the case with smoking and lung cancer where the relative risk increase was around three-thousand (3000) percent and the absolute risk increase was also very high. Thus all peer reviewed nutritional and drug science and media coverage of such science should report BOTH the relative AND absolute risks found in these studies.
So to all the Frank Hu’s and Walter Willett’s out there, stop your fucking fear mongering already. We already realize that your real intent is just to doctor data so everyone will follow your desired food patterns. But sorry, some of us actually understand how the science is done, and thus aren’t persuaded by your food religion.
So no Walter, a 13% relative risk for a drug isn’t a success. It’s just a con to sell more drugs. Always look for the ABSOLUTE RISK !!!!