Is Meat Carcinogenic?
Meat is considered both extremely nutrient-dense and carcinogenic - how can that be, and what’s behind it?
A frequently cited publication
In 2015 the IARC report “Carcinogenicity of consumption of red and processed meat” (1) appeared, an article and a meta-study that labelled processed meat “carcinogenic” and red meat “probably carcinogenic”.
Let’s take a closer look at the study and its results - one of the committee members, David Klurfeld, said this about how it came about:
This was the most frustrating professional experience of my life
It’s a meta-study that, in its own words,
The Working Group assessed more than 800 epidemiological studies that investigated the association of cancer with consumption of red meat or processed meat in many countries, from several continents, with diverse ethnicities and diets
An impressive number of studies taken into consideration here. But if we follow David Klurfeld’s statements, we learn:
…looked at 800 studies, but they used only 18…
If you have about two hours to spare, you can watch an interview with David Klurfeld in which he offers insights into how the study came about that go far beyond what I summarise here. The statement about the 18 studies actually used can be found there at around 18:30.
He has also published on the topic (2) .
Here are the (scientific) reasons why I can’t take the report seriously.
Epidemiological studies
Every study here is epidemiological - and that design is, by construction, blind to cause and effect unless the correlation is glaringly strong. As a rule: in epidemiological studies, a difference with a factor below 2 is considered weak and susceptible to confounding (3) .
That would mean the two groups under consideration - for example, people who eat meat versus those who don’t - would have to differ by at least a factor of 2 in risk for the study to be regarded as one that warrants a closer look. A factor of 2 also means a 100% increase in risk.
Directly related to this factor, it’s worth looking at a second factor too, and that’s what we’ll do now, before I come back to it.
(From Dr Klurfeld we learn that this threshold of credibility apparently isn’t taken seriously in nutritional science - even though, in this field too, statistics follow the same rules as everywhere else. More on that below.)
Relative vs absolute risk
The statement in the IARC report is
Eating 50 g of processed meat per day increases the risk of colorectal cancer by 18%
In studies of this kind there are two ways to consider a risk: as an absolute or as a relative risk.
IARC is dealing with the relative risk here: the risk increases by 18%, not to 18%. Let’s look at absolute numbers (it took me a while to find any - many sources limit themselves to the factors that raise a risk, without naming the baseline, the absolute risk).
The absolute risk of developing colorectal cancer at some point in life is about 5% (4) . On statistical average, then, roughly 5 out of 100 people will develop colorectal cancer.
If we raise this risk by 18%, we end up at 5.9% absolute risk. An increase of 0.9 percentage points, of course, doesn’t make nearly as impressive a headline as a relative risk of 18%.
Combining this with the statement above: in epidemiological studies, science normally expects factors greater than 2 before a result is even worth looking at in the first place (expressed as relative risk, that would be an increase of 100%), and so we arrive at the conclusion that the statement is, to begin with, not a scientifically sound result (outside of nutritional science).
But what is the reason for demanding such a factor?
One problem: healthy user bias
The IARC data are based on observational studies from the 1980s, 1990s and 2000s. In exactly those decades the official slogan was: “Meat and saturated fats are unhealthy, avoid them!”
Now let’s ask ourselves: Who, in that period, ate a lot of cheap, processed meat?
Simple. People who generally didn’t care about their health. The data show: the group with the highest meat consumption smoked statistically far more, drank more alcohol, barely exercised, was overweight and ate hardly any vegetables or fibre.
Counter-check: Who ate little meat?
Clearly: health-conscious people (who followed the guidelines of the day). They exercised, didn’t smoke, drank less and ate salad.
From the three points so far it becomes clear that epidemiological studies can under some circumstances be good for forming a hypothesis - but as a rule not for demonstrating cause and effect. Especially not when the deviation is very small.
In other words: the study may have measured less “meat makes you sick” than “people who paired instant soup with cigarettes in the ’90s don’t live long”.
Another problem: collection and aggregation
The epidemiological studies common in nutritional science are often so-called “food-frequency questionnaires”. These made up the bulk of the studies examined.
People are asked: “How many grams of liver sausage did you eat per week on average over the last 12 months?” or “How often did you consume bacon in the last six months?”. Nobody knows the correct answer to that question, but of course that doesn’t stop anyone from deriving global recommendations from it…
And of course you don’t only ask about individual ingredients, but about categories or dishes. The problem: the analysis. To be able to evaluate the data, they’re happily lumped together: if, for example, you ask about the frequency of eating hamburgers, these are often simply counted as “meat” even though meat makes up only 10% of a hamburger menu.
Hamburger fun facts, the details
Hamburger: Hypervegetarian?
Is one of the most iconic meat dishes actually a meat dish? An amusing little realization
Read more →So much for the aggregation. But the collection isn’t better: After all, it’s people who answer these questionnaires. Nobody knows exactly what they ate over the past year - and it’s gladly “estimated” - usually the figures don’t match the statistics, nor the total amount required. People tend to fudge themselves healthy by giving the “desired” answers.
In short: building a worldwide recommendation on patchy memories is methodologically reckless - but in nutritional science, apparently just another Tuesday.
What the IARC means by its categories
Now that we’ve seen how seriously the results can be taken, two points remain that are also always gladly cited in connection with the study:
Categories 1 and 2:
The line is always so nicely put: processed meat was placed in the same category as smoking, asbestos and plutonium. Very catchy, but it doesn’t mean what people commonly think. As the WHO explains in its FAQ (5): being sorted into this group explicitly does not mean that the same danger emanates from all the substances in it (as you can actually already tell from the difference between smoking and plutonium). Let’s consider two questions together:
Processed meat was classified as Group 1, carcinogenic to humans. What does this mean?
This category is used when there is sufficient evidence of carcinogenicity in humans. In other words, there is convincing evidence that the agent causes cancer. The evaluation is usually based on epidemiological studies showing the development of cancer in exposed humans.
In the case of processed meat, this classification is based on sufficient evidence from epidemiological studies that eating processed meat causes colorectal cancer.
Processed meat was classified as carcinogenic to humans(Group 1). Tobacco smoking and asbestos are also both classified as carcinogenic to humans(Group 1). Does it mean that consumption of processed meat is as carcinogenic as tobacco smoking and asbestos?
No, processed meat has been classified in the same category as causes of cancer such as tobacco smoking and asbestos (IARC Group 1, carcinogenic to humans), but this does NOT mean that they are all equally dangerous. The IARC classifications describe the strength of the scientific evidence about an agent being a cause of cancer, rather than assessing the level of risk.
Group 1 means “we’re certain”, not “it’s bad”. Bacon and smoking therefore share the confidence level, not the danger level.
And how certain one can be, we’ve already seen above - but to state it explicitly once more: a peculiarity of nutritional science is this. Randomised long-term studies on single nutrients are hard and ethically messy - so nutritional science just waves epidemiology through as ‘good enough’. We’ve already seen the criteria considered sufficient here in the opening paragraphs: a relative risk of 1.18, or 18%, is enough for the IARC to count as strong scientific evidence.
For me, and for scientists outside of nutritional science, it isn’t.
Conspiracy?
Anyone who at the start thought the nearly 2-hour interview with David Klurfeld (reminder: one of the members of the IARC working group) was too long, but who has followed the argument this far shaking their head, still has the option of listening to his detailed remarks or reading his publications on the topic.
He stated publicly (among other places in the video above) that among the 22 voting members of the panel there was a significant number of committed vegetarians/vegans. He criticised the fact that the IARC asks extremely strictly about financial conflicts of interest, but completely ignores personal ideologies or the researchers’ own diet - even though these massively influence bias.
He reported that studies showing no association between meat and cancer (including large, high-quality data) were systematically ignored or downgraded by the majority of the panel. When he or other researchers pointed this out, these objections were simply voted down (800 vs 18 studies in the selection).
Klurfeld writes right in the abstract of his article “Research gaps in evaluating the relationship of meat and health” (2):
The observational studies are heterogeneous and do not fulfill many of the points proposed by AB Hill in 1965 for inferring causality; his most important factor was strength of the association which in dietary studies is usually < 1.5 but is not considered adequate in virtually all other areas of epidemiology outside nutrition
Why do they do this?
I can only speculate, but after hearing David Klurfeld’s arguments, all of these speculations drift into the dangerous vicinity of conspiracy theories.
The IARC may regard such a small factor as “strong scientific evidence” - that’s not enough for me. Even the otherwise rather meat-cautious Deutsche Gesellschaft für Ernährung - German Nutritional Society, a largely government funded NGO, designated as appointed authority on nutritional matters notes that meat is very nutrient-dense.
Simply too many things don’t add up. And insisting on the usual criteria can hardly be called anti-science. Looked at this way, it’s rather nutritional science that ignores established scientific processes - and, faced with legitimate criticism, doesn’t even clear it up.
And last but not least: even the IARC says, in its own report on the Group 2A category into which it has sorted red meat:
(╯°□°）╯︵ ┻━┻…other explanations for the observations (technically termed “chance”, “bias”, or “confounding”) could not be ruled out with reasonable confidence (6) .
