The 18% that never says 18% of what

Relative risk without the baseline The convenient comparison

Dark navy briefing slide, masthead “Morning Health Briefing · Lyon, 26 October 2015”, headlined “Your breakfast is now a Group 1 carcinogen”. A subhead reads: twenty-two scientists from ten countries, convened by the World Health Organization’s cancer agency, considered more than 800 studies on cancer in humans and reached a verdict; processed meat — bacon, ham, sausages, hot dogs — is carcinogenic to humans, the agency’s top category, the one that already holds tobacco smoking and asbestos, and every 50 g eaten daily raises the risk of bowel cancer by 18%. The main panel is labelled “Group 1 · carcinogenic to humans · sufficient evidence in humans” and holds seven bars of identical length, one per agent: tobacco smoking, asbestos, plutonium, ultraviolet radiation, alcoholic beverages and formaldehyde in blue, processed meat in gold at the bottom. Every bar is annotated “sufficient evidence”, and a caption underneath reads “Every agent above sits at the top of the agency’s scale”. Below it, a second chart headed “Increase in bowel-cancer risk per 50 g of processed meat eaten daily” shows a single gold bar reaching +18% on an axis marked +0%, +5%, +10%, +15%, +20% — an axis of percentage increases, with no case counts anywhere on the slide. Three ticked boxes run along the bottom: “Twenty-two experts, ten countries — a Working Group convened by the WHO’s cancer agency in Lyon, sitting for eight days”, “More than 800 studies — considered on cancer in humans, across more than a dozen types of cancer”, and “The top category — Group 1, sufficient evidence in humans. There is no higher class to be put in.” Down the right, a card carries the figure +18% in large gold type over the words “increase in the risk of bowel cancer for every 50 g portion of processed meat eaten daily. IARC Monographs, Volume 114.” A second card, “Where the agency filed it”, lists processed meat, tobacco smoking, asbestos and plutonium at Group 1 and red meat at Group 2A. A third, “The agency’s ladder”, gives Group 1 carcinogenic to humans, Group 2A probably carcinogenic, Group 2B possibly carcinogenic, Group 3 not classifiable. A gold-bordered card headed “The verdict” reads: sufficient evidence in humans that the consumption of processed meat causes colorectal cancer — twenty-two experts, ten countries, one category, the same one as tobacco. Beneath it, in small grey type, sits the disclosure: “The agency’s classifications describe the strength of the scientific evidence about an agent being a cause of cancer, rather than assessing the level of risk.” A footnote credits IARC Monographs Volume 114 (2015) and the WHO Q&A on the carcinogenicity of red meat and processed meat, and calls the slide a demonstration drawn by us in the style of the coverage that followed. — Showing the misleading chart

The same figure on a cream background, kicked “The same 18% · with the baseline put back” and headlined “18% more of 56 in 1,000”. A subhead explains that the ratio is the agency’s own and nobody disputes it; what it multiplies is a lifetime risk: out of every 1,000 people in the UK about 61 develop bowel cancer at some point in their lives, and Cancer Research UK carried the association through to the two ends of the meat-eating range to get 56 per 1,000 among the people who eat the least processed meat and 66 among those who eat the most — fifty-six times 1.18 being 66, the same 18% arriving with its denominator attached. The main figure is an icon array headed “1,000 people in the UK · lifetime cases of bowel cancer”: a thousand dots in fifty columns and twenty rows, of which the first 56 are dark blue, the next 5 a lighter blue, the next 5 gold, and the remaining 934 pale — so the entire effect occupies the first two rows. Beside it, in gold, “56 → 61 → 66”, over a note that these are cases per 1,000 people over a lifetime for the lowest eaters of processed meat, the UK average and the highest eaters; the whole of the 18% lives in the first two rows of the grid, five cases above the average and ten between the ends of the range, on one axis that runs all the way to 1,000. A legend names each band: 56 in 1,000 expected among the people who eat the least processed meat; +5 to reach 61, the UK average; +5 more, to 66, expected among the people who eat the most — ten more than the lowest eaters; and the remaining 934 who do not develop bowel cancer. Below, a second chart headed “Cancer deaths per year worldwide attributable to…” redraws the Group 1 roster on one shared linear axis running from 0 to 1,000,000: tobacco smoking about 1,000,000, alcohol consumption about 600,000, air pollution more than 200,000, red meat about 50,000 if the reported association proves causal, and processed meat about 34,000 in gold — a bar a thirtieth the length of the first. A note credits the Global Burden of Disease estimates as quoted by WHO. Four cards run down the right. “A ratio is not a risk”: 18% is a multiplier, 56 in 1,000 is the thing it multiplies, and printing the first without the second leaves the reader to supply a baseline from imagination. “Both numbers, one picture”: lowest eaters 56 in 1,000, highest eaters 66 in 1,000, with the arithmetic set out — 56 × 1.18 = 66, and 66 − 56 = 10 in 1,000. “Hazard is not magnitude” quotes WHO: processed meat “has been classified in the same category as causes of cancer such as tobacco smoking and asbestos… but this does NOT mean that they are all equally dangerous”, so the category answers whether something causes cancer, never how much. “The share of the problem”: in 2011 about 3 in every 100 UK cancers were attributed to eating too much red and processed meat, roughly 8,800 cases a year, against 64,500 a year and 19% of all cancers from smoking. A fifth card, “The question that does it”, asks: increase from what? A footnote credits Cancer Research UK’s 26 October 2015 explainer for the lifetime figures, IARC Monographs Volume 114 and the WHO Q&A for the ratio, the classification and the attributable-death estimates, notes that the 18% is per 50 g eaten daily while the 56 and 66 are the two ends of the UK consumption range, and says both drawings are our own. — Showing the misleading chart

Be the first to star this exhibit

Every 50 g of processed meat eaten daily raises the risk of bowel cancer by 18% — the WHO cancer agency’s own figure, and nobody disputes it. What the slide never prints is the number the 18% multiplies. About 61 people in every 1,000 in the UK develop bowel cancer at some point in their lives; carry the association out to the ends of the meat-eating range and you get 56 per 1,000 among the lowest eaters and 66 among the highest. Same ratio, ten cases per thousand, one lifetime.

IARC Monographs Volume 114 (2015) and the WHO Q&A on the carcinogenicity of red meat and processed meat, with lifetime risk figures from Cancer Research UK, “Processed meat and cancer — what you need to know”, 26 October 2015, 2015Added August 15, 2026

01The claim

Your breakfast is now a Group 1 carcinogen. Twenty-two scientists from ten countries, convened by the World Health Organization’s cancer agency, considered more than 800 studies on cancer in humans and reached a verdict: processed meat is carcinogenic to humans, sufficient evidence, the top category on the agency’s ladder — the same one that holds tobacco smoking, asbestos, plutonium, ultraviolet radiation, alcoholic beverages and formaldehyde. Seven bars, all the same length, because all seven agents sit at the top of the scale. And every 50 g of processed meat eaten daily raises the risk of bowel cancer by 18%.

02The trick

Every number on the slide is real and correctly quoted. The 18% is IARC’s own, the classification is IARC’s own, and the working group really did sit for eight days over more than 800 studies. What the slide never prints is the number the 18% multiplies. Look at the axis on the lower chart: it runs from +0% to +20%, and nothing on it is a quantity of anything. An increase of 18% is a ratio, and a ratio is a fact about two numbers you have not been shown — 18% more of a one-in-a-million chance and 18% more of a coin flip both draw the same bar. The reader is left to supply the baseline, and the surrounding company sets the guess: seven identical bars invite you to read the roster as a ranking, when the panel is a list of things the evidence is strong about, not a list of things that are equally dangerous. That is a category being read as a quantity, and the slide’s own fine print says so, in grey, under the last card: the classifications “describe the strength of the scientific evidence about an agent being a cause of cancer, rather than assessing the level of risk.” (This exhibit is our own demonstration, drawn in the style of the coverage that followed the 2015 announcement — no such slide was published, and IARC’s Q&A printed the caveat prominently rather than in grey.)

03The fix

Put the baseline back and draw both numbers on one absolute scale. About 61 people in every 1,000 in the UK develop bowel cancer at some point in their lives. Cancer Research UK carried the association out to the two ends of the meat-eating range and published the arithmetic: about 56 cases per 1,000 among the people who eat the least processed meat, and about 66 among the people who eat the most. Fifty-six times 1.18 is 66 — the identical ratio, now standing on the ground it multiplies. Sixty-six minus fifty-six is ten, which is the number a person can actually act on: ten extra cases per thousand people, across an entire lifetime. An icon array of a thousand dots shows both at once, and the shape of the answer is that the whole effect fits in the first two rows. Then redraw the roster the same way. Using the Global Burden of Disease estimates WHO quotes in the same Q&A, cancer deaths per year worldwide run to about 1,000,000 for tobacco smoking, 600,000 for alcohol, more than 200,000 for air pollution, about 50,000 for red meat if the reported association proves causal — and about 34,000 for processed meat. On one shared axis the identical bars separate by a factor of thirty, which is the difference between asking whether something causes cancer and asking how much. Neither redraw softens the finding; both just answer the question the original leaves open. The habit is small and works on any chart: when you meet a percentage increase, ask what it increased from, and do not accept the chart as an answer if the baseline is not on it.