Red meat and bowel cancer: what the 17% actually measures

Somebody hands you a number. Seventeen per cent. For every 100g of red meat you eat in a day, so the sentence goes, the risk of bowel cancer rises by 17%.

It is a ratio between two rates. Not seventeen people in a hundred getting cancer. Two groups of people alike in everything else, one of them eating 100g a day more red meat than the other, followed for years, and the higher group ends up with about 17% more cases than the lower one. Neither rate is in the sentence, and neither is the 100g.

100g of cooked red meat is about the rump steak on the NHS portion list, which weighs 102g. Stripped down, the claim is this: eat one more rump steak a day than you do now, and the number of cases goes up by roughly a sixth.

That is a real figure from a real paper. The paper reports something the sentence leaves out, and it does so in the abstract.

Where the number comes from

Chan and colleagues at Imperial College pooled prospective cohort studies published to March 2011: populations followed forward in time, with food recorded before anybody got ill [1]. Highest intake against lowest gave a summary relative risk of 1.22 for colorectal cancer (95% CI 1.11 to 1.34). Per 100g a day of fresh red meat, 1.17 (1.05 to 1.31). Per 50g a day of processed meat, 1.18 (1.10 to 1.28).

IARC read that evidence in 2015, along with the mechanistic and animal work, and put processed meat in Group 1 and red meat in Group 2A [2][3]. Group 1 says how confident the committee was that the agent can cause cancer in humans. It says nothing about size. Alcohol and tobacco sit in the same group as processed meat and are not the same size of problem.

The red meat interval: 5% to 31%. The bottom of it is nearly nothing, which is why the committee gave red meat "limited evidence" and processed meat "sufficient" [3].

The curve bends

Chan's team did not assume one slope for every intake. They ran a nonlinear dose-response analysis. Risk rose roughly in step with intake up to about 140g a day, then the curve approached a plateau. Rectal cancer on its own came out null [1].

That is the finding worth arguing about, and it is one sentence long.

What the bend does to the arithmetic

Multiply 1.17 anyway, the way people do. At 300g a day you get 1.60 (1.16 to 2.25). At 500g, 2.19 (1.28 to 3.86). At 600g, 2.57 (1.34 to 5.05). Six rump steaks a day, and two and a half times the risk.

None of those products are in the paper. The authors tested whether the multiplication holds above 140g, and the curve did not behave that way. So the 600g figure is arithmetic done on a slope the paper had already rejected for that range.

The opposite reading fails too. A plateau means the fitted risk stops climbing. It does not mean the risk comes back down. And the plateau is the least secure part of any dose-response fit, because it lands on the highest intake category in each cohort, where the fewest people sit and the interval is widest. The abstract states the plateau with no confidence interval, because pooling published categories cannot produce a firm one.

What is measured, then, is a rise across the range the cohorts actually covered, which for most of the studies runs from nothing up to about 200g a day, with the shape going flat at the top. Above that the risk is unmeasured. Unmeasured is not safe. It is also not 2.57.

Where a plate lands

The NHS tells anyone over 90g a day of red or processed meat, cooked weight, to come down to 70g [5]. Its portion examples, also cooked: a large grilled steak 163g, a rump steak 102g, a burger 78g. The large steak is 1.8 times the threshold in one sitting.

WCRF works in cooked weight too and gives a different number: no more than three portions a week, 350 to 500g, with little or no processed meat [4]. That is 50 to 71g a day. WCRF sets a target. The NHS offers a step down to a population already above one. Treating the two as the same instruction is most of the confusion in this argument.

Most British adults are nowhere near either. In the National Diet and Nutrition Survey for 2019 to 2023, every age and sex group averaged under 70g a day of red and processed meat together. Men aged 19 to 64 were highest at 66g, and 27% of them were over 90g [6]. Population advice, in other words, aimed at about a quarter of men. Somebody eating 600g a day does not appear in that dataset.

Why one meat is Group 1 and the other Group 2A

Put Chan's two ratios on the same basis. Per 50g, processed meat gives 1.18. Double it to 100g and you get about 1.39. Gram for gram the processed slope is steeper than the 1.17 for fresh red meat, which is part of why the grades differ [1].

The mechanisms offered differ as well: nitrite and nitrate from curing, haem iron, and heterocyclic amines and polycyclic aromatic hydrocarbons from high-heat cooking. Nobody has shown which one is doing the work. If curing were the whole story, uncured bacon would be a different food, and nobody has shown that either. The cohorts cannot separate a charred steak from a slow braise.

Anyone who tells you a single mechanism is settled, in either direction, is ahead of the evidence.

What a cohort cannot see

The low-meat group is not a control group. They ate more fibre and more vegetables, less of everything processed, and were less likely to smoke. Adjustment takes care of what was measured. It cannot take care of what nobody thought to ask in 1995. That is what residual confounding means: a limit, not a debating trick.

Food frequency questionnaires are the other limit. The NDNS puts underreporting of energy intake at around 30% [6]. Misreporting of that kind pulls an association towards no effect, which makes the published figure more likely to understate a real effect than to invent one. That cuts against the dismissal as much as against the alarm.

The one thing here with a measured effect

About 48,200 new bowel cancer cases are diagnosed in the UK each year, 12% of all new cancer diagnoses [7]. The NHS invites everyone aged 50 to 74 and registered with a GP to do a FIT test every two years, with the first kit sent at 50, 52 or 54 depending on when you turned 50 [8]. It finds cancers earlier and removes polyps before they become cancer. The NHS is careful in its own wording that screening does not stop you getting bowel cancer.

If you are over 50 and eat a lot of beef, that kit is the only thing in this article with a measured effect on what happens to you, and it takes four minutes in a bathroom. Do it whether or not you change the meat.

Less processed meat, then, and red meat inside 350 to 500g cooked a week, is what the evidence supports for a population. At 600g a day you are outside the range anyone has measured, in a region where the only published curve bends flat and cannot be trusted to stay there. If you think the extrapolation holds better than I do, the argument to have is about the shape of that curve and where the plateau really sits. The 17% will not settle it.