“Cancer.” The word itself is one of the most dreaded, and almost no diagnosis is more frightening. So, when people are assured that they can prevent cancer by avoiding red and processed meats, I’d hazard a guess they do. This advice is usually dispensed for all types of cancer, even though only one kind of cancer —colorectal — has ever been classified as carcinogenic by the world’s leading agency in this field. In 2015, the International Agency for Research on Cancer (IARC) of the World Health Organization (WHO) designated processed meats (hot dogs, bacon, ham, sausage, cold cuts) as Group 1 carcinogens, a “certain” cause of cancer, and fresh red meat (beef, pork, lamb) as Group 2A, or “probable” carcinogens.

This news resounded worldwide, with hundreds of menacing headlines, including Bad Day For Bacon: Processed Meats Cause Cancer (NPR), OMG, Bacon causes cancer(NY Post), and Bacon, hot dogs and processed meats cause cancer, WHO says (PBS), and Meat as a Cause of Cancer (NYT editorial). At least one authoritative group, the American Institute for Cancer Research, the leading cancer research organization in the U.S. focused on lifestyle factors, expressed strong support for the IARC decision.

Nina Teicholz

This IARC announcement was an historic event. For the first time ever, a global health organization had declared a major component of all human diets throughout history to be a likely carcinogen. Yet to support this 2015 claim, the IARC released a mere two-page summary of its findings, in Lancet Oncology.  Twenty-two international experts (the “working group”) had met for eight days in Lyon, France, to evaluate a mountain of data–more than 800 observational studies alone. Because the working group produced a mere two-page summary, their analysis could not be independently verified at the time.

More surprisingly, the Lancet’s conclusions leaned heavily on only eight papers, all from epidemiological studies which linked what people ate, or reported they ate, with cancers they developed later in life. Even though it is well-established that the vast majority of this type of nutritional study cannot prove a causal link between food consumption and disease, the agency concluded that the strength of evidence for processed meats causing cancer equaled that of tobacco and asbestos.

The figures from IARC that caught the world’s attention came from its press release accompanying the Lancet article, which asserted that eating just 1.75 ounces of processed meat daily (about the size of a matchbox) increases the risk of colorectal cancer by 18pc. These claims relied on a single paper, a meta-analysis from 2011. And the only significant finding in this study was a link between cancer and fried sausage and pork meat–not beef, mutton, or lamb. That the working group employed weak findings on pork and fried sausage to generalize to all red and processed meats suggests a biased interpretation of the evidence, an important theme to which I’ll return below.

Beyond this, the 18pc claim was highly tenuous, as the Lancet paper reported only tiny associations, known as “relative risks,” between meat and cancer—1.17 for fresh meat and 1.18 for processed meat. The number 1 indicates zero relationship. Increases of 0.17 and 0.18 are minuscule. In the field of epidemiology generally (outside of nutrition), these numbers would be considered far too small to be meaningful.¹

Still, should numbers like 0.17 and 0.18 keep us from eating steak and bacon, if only as a precaution?

It’s doubtful. These increases represent a mere 6pc rise in absolute risk over a baseline risk (for a 50-year-old male) that is only 4-5pc, according to the National Institute of Health’s colorectal cancer risk assessment calculator. Eating meat raises the risk from 4.5pc to 4.68pc. If that sounds like a lot, let’s put it another way: For every 10,000 fifty-year-old men, we can expect 468 of them to get colon cancer. But 450 are expected to get it no matter how much meat they eat. Thus, a mere 18 men can blame it on the meat, according to the IARC itself. The other 9,682 will be fine, regardless.

To add context to this cancer “scare” story, it was not until fully three years later, when the IARC published its complete monograph that any meaningful questions about the scope and quality of the evidence could be addressed. By then, the red-meat-causes-cancer story was firmly settled into the public mind, like layers of sedimentary rock laid down by innumerable news stories, solidifying the perception as ‘settled science’ on the subject. Dozens of “Bacon is Bad for You” stories had been written and policies against red meant enacted. If anyone is still wondering about the influence of this two-page study, according to the metrics system Plum X, as of this writing, the Lancet two-pager has been cited 1,418 times, including 108 citations in policy documents.

Evidence ignored

According to interviews with seven working group members and observers, important evidence was outright excluded from consideration. These barred studies were not minor: they included two randomized, controlled clinical trials (RCTs) that specifically reduced red meat as part of a range of dietary interventions to reduce cancer. In other words, the only studies comprising the most rigorous type of evidence—indeed, the only type of evidence that can reliably show a cause-and-effect relationship between food and disease—were excluded from consideration by the working group.

The first of these trials was the Polyp Prevention Trial, which failed to show any effect on tumor recurrence in people following a low-fat, low-meat, fruits-and-vegetables-based diet, even after eight years of follow-up tests. The second trial was the famous Women’s Health Initiative, which involved almost 50,000 women and cost roughly half-a-billion dollars yet failed to demonstrate that low consumption of red and processed meat had any effect on the risk of colorectal cancer (or indeed, of several other types of cancer) after more than eight years. Both of these trials, funded by the National Institutes of Health (NIH), were “powered”—meaning, they lasted long enough and included enough subjects to obtain meaningful and generalizable data on cancer. When one IARC working group participant suggested including these trials for consideration by the panel, he told me he was rebuffed.

The reason given by Bernard  Stewart, chair of the working group and a professor at the University of New South Wales in Sydney, was that these studies involved dietary changes beyond just meat reduction, suggesting that there were confounding factors that might have significantly affected the study results. That’s true, but it doesn’t change the fact that if meat is reduced and cancer rates do not drop accordingly, that result is significant evidence contradicting the IARC conclusion. Ignoring that contradiction leaves out important data. Even more troubling, as one participant observed, the many epidemiological studies that were included by IARC also involved hundreds of foods other than meat. Yet IARC permitted these studies to be included. It’s therefore fair to say that IARC applied a double standard, giving preference to weak data over strong. As the two-page report stated, “The greatest weight was given to prospective cohort studies done in the general population,” the very opposite of what the basic rules of scientific methodology require.

Three lines of evidence

The studies considered by the working group fell into three categories, none of which included clinical trials in humans, and these were:

1. Mechanistic evidence (possible mechanisms by which meat could cause cancer)
2. Animal data, and
3. Epidemiology.

Animal evidence

The animal-data research subgroup reported that its evidence was “inadequate,” because results showed no clear effect of feeding red or processed meat on tumor development (as members of that subgroup related to me).

Mechanistic evidence

This type of evidence does not show a cause-and-effect relationship but rather evidence for a biological mechanism or pathway that could be at work if a causal relationship is found. The mechanisms proposed were three: the heme iron from red meat, genotoxicity, and oxidative stress. This subgroup decided the evidence was “strong” for red meat and “moderate” for processed meat.

It’s important to note that one member of this subgroup was Denis Corpet, a professor at the University of Toulouse, France, who had been trying since 1998 to show how red and processed meat cause cancer. His first two experiments actually showed the reverse, that bacon-fed rats had lower amounts of cancer-related compounds in their feces compared to controls. The study concludes, “A bacon-based diet appears to protect against carcinogenesis,” and his following study completed in 2000 concludes, “Results suggest that NOC [N-nitroso compounds] from dietary bacon would not enhance colon carcinogenesis in rats.”

At this point, Corpet discovered that rats on calcium-deficient diets who were fed cured meats yielded higher cancer-related compounds. “This explains why our earlier experiments didn’t work,” he told me recently. In later experiments, Corpet tended to use calcium-deficient diets. One in 2013, where rats were fed hot dogs, concluded, “This is the first experimental evidence that a widely consumed processed meat promotes colon carcinogenesis in rats.”² Some working group members told me they tried to insert a caution in the wording for the Lancet paper about the calcium deficiency problem, but Corpet strenuously objected and would not allow it.

Another shortcoming with the mechanistic evidence is that many experiments didn’t isolate the effect of the meat from the fats in which they were cooked. Seed oils are known to create oxidation products when heated,³ which Corpet acknowledged to me: “Oxidated fats are toxic and that findings on cooked meat are due to the fat, yes.” However, the working group did not explore this possibility. Were they blaming meat for what the seed oils possibly did? Overall, the Lancet article said this working group could find “few human data” to support plausible mechanisms on how processed meat might cause cancer.

Epidemiological evidence

The powerhouse that drove the IARC conclusion came from observational, or, as I have noted, epidemiological studies. Consideration of these was left until the final afternoon of the meeting, when the IARC staff presented the working group with “A stack 4 inches thick,” as one participant recalls (and others confirmed), “and they just asked us, does everyone agree?”  Whereas the work of the previous two subgroups had been subjected to extensive review and editing by the larger group, the summary of the epidemiological evidence was not. “We had no chance to do any kind of meaningful review of those papers,” said working-group member David Klurfeld in an interview. At the time of the IARC meeting, he was the National Program Leader for Human Nutrition at the U.S. Department of Agriculture.

The Lancet article acknowledges the limitations of these epidemiological studies, saying that “No clear association was seen in several of the high-quality studies [on fresh meat] and residual confounding from other diet and lifestyle risk is difficult to exclude.” The article also acknowledged that “there is limited evidence in human beings for the carcinogenicity of the consumption of red meat.”  Yet later, without explanation, the paper contradicts itself, saying there is “substantial epidemiological data” to support the conclusions [my emphasis above throughout].

So which is it? “Limited” or “substantial” evidence? In either case, as noted above, the ability of epidemiological evidence to establish cause-and-effect relationships is considered weak. Remember that the associations between red/processed meat and colorectal cancer were inconsistent and tiny—0.17 and 0.18—nothing close to the established 10-to-30-fold greater risk of getting cancer by heavy smokers compared to those who never smoked.^[3]

A more dispassionate view of the evidence

Four years after the IARC’s 2015 decision, the most rigorous, comprehensive reviews of red meat ever conducted were published in the prestigious Annals of Internal Medicine. Unlike most working group members, the authors had never published research on meat and cancer and had no stake in the issue. Instead, they were trained in interpreting scientific evidence using an authoritative review methodology recognized worldwide, the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE). The Annals papers concluded that only “low” to “very low” quality evidence links red or processed meats to any type of cancer or cancer mortality.⁴

Despite this major review, Bernard Stewart recently wrote to me: “To my knowledge, the IARC… [has] not been subject to adverse criticism or made aware of a perceived need for additional information to justify the outcomes made, either at the time or since.” Stewart did not reply to further emails asking him about the contradictory conclusions of the GRADE reviews.

Was the IARC review biased?

Why was the IARC process so problematic? One scientist reckoning with this question was the IARC working group member quoted above, David Klurfeld: “I was an old-timer in nutrition and well-acquainted with its biases against meat in the field, but still, I thought that when leading experts saw the totality of the evidence, they would change their minds,” he told me. “I was naïve.”

Over the course of the IARC meeting, he came to suspect that the outcome had been pre-ordained. This suspicion is credible when we consider that fully 17 (77pc) of the 22 working-group members had published multiple papers endeavoring to show that meat causes cancer, and six had been doing this research for 20-30 years. The papers they authored often concluded that the evidence was inconsistent or inconclusive, yet they kept on.

Their perseverance is either heroic or depressing, depending on your views about the practice of science. At what point does the researcher have an obligation, ethical or otherwise, to concede that a hypothesis is unlikely to be true? For instance, Alicja Wolk from the Karolinska Institute, Sweden has been investigating possible links between meat and various types of cancer since 1993. Her first study “failed to find any association between intake of meat, sausage, cold cuts, liver…[and] the habit of frying, smoking or grilling foods, and risk of gastric cancer.” Her subsequent epidemiological studies failed to find statistically significant connections between meat consumption and renal cancer (in multiple papers), endometrial cancer, breast cancer, cancers of the esophagus and gastric cardia, or colorectal cancer (the evidence in this study was “not conclusive”). Yet the work continued, with Wolk publishing 38 papers on meat and cancer before the IARC meeting.

Other working group members had similar publication histories, seemingly ever-hopeful that another study might yield the desired results. In this light, the IARC meeting could be seen as an opportunity for the majority of the working group to make a decision that would finally vindicate their beliefs. At the Lyon meeting, many of the papers reviewed were the working group’s own papers. As one meeting participant observed to me, “They were not independent reviewers. It was their own data!”

Could we really expect them to be unbiased? The critical flaw seems to lie in how the group is selected from the start. When the IARC searches for leading authorities on a possible carcinogen, who should the agency appoint if not the most-published experts in the field? And because scientific journals have a bias towards putting out “positive” results, successfully published scientists in cancer research will be those who find that X might cause cancer, not those who find X causes nothing. Thus, the very process that forms the IARC’s working groups leads to a pro-cancer bias⁵

Klurfeld suggested to me that IARC staff may also have been prejudiced against meat. He recalls 3 or 4 members out of the 16-person IARC staff mentioning at meals that they were vegetarian,⁶ which, if true, would make the prevalence of vegetarians on the IARC staff as high as 25pc, five times more than the general population rate (4-5pc).⁷

That these personal and professional biases influenced the proceedings seems hard to deny. When a toxicologist on the panel brought up flaws in the mechanistic evidence, for example, he was told “shut up and sit down,” according to several witnesses I interviewed shortly after IARC’s 2015 decision (The member did not respond to emails about this comment). Klurfeld also said that when he tried to introduce language into the Lancet paper softening the certainty of the meat-cancer connection, another panelist strenuously disagreed, forcing him to “give up” on the effort.

IARC says its working groups arrive at conclusions “by consensus” (p. 155), but the group for red/processed meat did not work toward a unanimous consensus and did not reach one. Beyond this, further input from the working group was not invited in the production of the final monograph, an IARC press officer told me.

Does everything cause cancer?

Since its founding in 1965, IARC says it has examined 1,045 natural, biological, and chemical agents and found 52pc carcinogenic to humans. A 2013 study that randomly selected fifty common ingredients a single cookbook discovered that 80pc were reported in studies (not all by IARC) to pose a cancer risk, “even though evidence is weak.”  IARC has found that very hot beverages (but not coffee) causes cancer, as does shift work.

IARC formerly had a category called  Group 4, “probably not carcinogenic to humans.”  Yet the agency eliminated this group in 2019, and now, the best possible outcome for an agent is “not classifiable as to its carcinogenicity to humans.” This designation implies that there’s always a chance, with more evidence, for a food or substance to be found carcinogenic. For IARC, everything is a potential risk.

IARC’s new headquarters, opened last year

Even Bernard Stewart expressed some doubt about the IARC process in a 2016 editorial for the Lancet, entitled, “How do we judge what causes cancer? The meat controversy.” He and a fellow working-group member wrote, “Eating red meat [does] not necessarily lead to cancer,” because cancer is understood as the “complex interplay of numerous agents and biological response.” The concept of “no safe dose”—which can be applied to cigarettes—is “meaningless” for meat, they write. But IARC is not to blame, they add. Rather, those who translate IARC decisions into policy should make a “very clear distinction” between a “no safe dose” carcinogen and other potentially cancer-causing agents. This gap between evidence reviews—reliable or not—and policy implementation is concerning. I’d bet that most policymakers don’t understand the IARC’s distinction, and the IARC certainly doesn’t differentiate in its classifications.

For decades now, researchers have been trying to blame cancer on red and processed meat, and the tenuous evidence has only gotten weaker. While radioactivity, smoking, and industrial chemicals no doubt do cause cancer, red meat—a food eaten by humans since our first days on earth—is not one of them. The IARC process is a demonstration in how, through funding, bias, and power, a failed hypothesis can be kept alive.

Notes:

¹ Relative risks “less than 2.00 are viewed with caution,” noted a 2002 article on the National Cancer Institute website, which has now been scrubbed from the internet (including the Wayback machine and also, frighteningly, from my computer).

² In an interview, Corpet told me that he thought calcium-deficient diets were appropriate, because these better reflected the existing calcium deficiencies in human populations. One wonders why it wouldn’t be preferable to address the calcium deficiencies instead.

³ Crampton, E. W. et al., “Studies to Determine the Nature of the Damage to the Nutritive Value of Some Vegetable Oils from Heat Treatment: IV. Ethyl Esters of Heat Polymerized Linseed, Soybean and Sunflower Seed Oils,” Journal of Nutrition 60, no. 1 (1956): 13–24; John S. Andrews et al., “Toxicity of Air-Oxidized Soybean Oil,” Journal of Nutrition 70, no. 2 (1960): 199–210; and Samuel M. Greenberg and A. C. Frazer, “Some Factors Affecting the Growth and Development of Rats Fed Rancid Fat,” Journal of Nutrition 50, no. 4 (1953): 421–440.

⁴ For more reading on the epidemiology on meat and cancer, see this paper by David Klurfeld.

⁵ IARC receives almost half its funding from private sources, including the World Cancer Research Fund International, the French National Cancer Institute, and Cancer Research UK, groups that also stay relevant by finding new causes of cancer. Thus, the IARC can be seen as providing reinforcement for that agenda.

⁶ Some scholars have argued that dietary preferences are a powerful source of bias, even more influential than intellectual biases. People eat three times a day—far more often than churchgoers attend services—and food is often embedded in a person’s culture, family, ideology, or religion, making dietary choices a deeply integral part of one’s identity.

⁷ One decision by the IARC staff was to attach a TV crew to an invited observer from the meat industry who told me they followed her every move outside the meeting room.

About the author

Nina Teicholz is an investigative journalist and author of The Big Fat Surprise, a 2014 book revealing a lack of rigour in the science underpinning the low fat, low meat, diet that is recommended by most countries. She regularly posts on a Substack blog called ‘Unsettled Science’ and her book is available in stores, on Amazon or on audiobook platforms.

This article was originally published on Substack.