Kai Kupferschmidt’s June 16 Email
I’m covering a topic heavy on risk communication once again and would love to get your thoughts for an article in Science.
The IARC yesterday announced that it no longer considers coffee “possibly carcinogenic” but “not classifiable as to its carcinogenicity.” At the same time it announced that drinking very hot beverages is “probably carcinogenic.” They could not say how big a risk that poses, however.
A lot has been written already about the IARC’s classification scheme but I would love to get your opinion as someone in risk communication:
- How useful is it to inform the public solely about the hazard of a substance (as they did yesterday with the very hot beverages)?
- What do you think of the press materials I have attached? [Kai attached the IARC news release and an IARC Q&A about the coffee and very-hot-beverages announcement.]
- Is the IARC doing a bad job of communicating its results? Or is the problem the scope of what they look at in the first place?
The agency also says here that being classified in the same category means that “we can say with the same degree of certainty (or the same strength of the scientific evidence)” that two substances cause cancer. But is that really true? For coffee there now exists a huge amount of data (and that led to its reclassification), so surely coffee can be said to be in Category 3 now with much more certainty than other substances.
Peter M. Sandman’s June 17 Response
Regarding the recent IARC announcements re coffee and very hot beverages, I followed them casually and also read the two attachments you sent me.
Coincidentally, I started writing a long piece on IARC risk communication a few months ago, in response to the announcement that put processed meats in Category 1 (known human carcinogen). I never finished that piece – it’s still on my list of articles I might or might not return to … but I probably won’t.
In the meantime, I am attaching the incomplete draft. If you want to quote from it, feel free to do so. (I’m not sure how you should handle the attribution. Just quote it as if I had said it to you in an interview? Attribute it to an “unpublished draft” that I shared with you? Whatever you think best.)
The draft focused on IARC’s customary confusion of the distinction between how serious a carcinogen something is and how confident the experts are that it’s actually a carcinogen: “how bad” versus “how certain.” I argued that IARC explains this distinction clearly, but in ways and places where IARC has reason to expect most journalists to miss it or ignore it. And I argued that IARC has long been aware of this and considers it a feature, not a bug, of its methodology. IARC gets to say that processed meat is a known carcinogen, leave the impression that processed meat is a serious carcinogen, and thereby discourage people from eating processed meats without having to make claims it can’t support.
Obviously downgrading coffee from “possible” carcinogen to “who the hell knows?” raises somewhat different issues. This time IARC isn’t trying to warn people. It is trying to tell them a very weak prior warning now means even less than it did.
This is a very Jesuit/Talmudic sort of distinction, I think. It reminds me a bit of WHO’s recent brouhaha over the difference between saying women in places with circulating Zika should consider postponing pregnancy and saying such women should consider all their options, including postponing pregnancy. It’s not possible to consider postponing pregnancy without the possibility of getting pregnant coming to mind as well. In much the same way, telling people that it’s possible coffee might cause cancer is only very subtly different from telling people that you don’t know whether or not coffee causes cancer.
IARC seems to be saying that it is looking at the same hard-to-interpret studies it looked at last time, studies showing a weak relationship between coffee and cancer. But now it has more studies, many of which failed to show a relationship at all. So the weak evidence seems even weaker in the context of these new studies that found nothing. It’s not saying coffee isn’t a carcinogen. It’s just saying the evidence that coffee is a carcinogen is even weaker than it used to be.
Where my processed meat analysis is relevant, I think, is that once again IARC wants people to find value in its pronouncement on the quality of the evidence without pronouncing on the size of the effect. With processed meats, IARC wanted people to conclude that the risk is serious, based on its claim that the risk is established. With coffee, IARC presumably wants people to conclude that the risk isn’t serious even if it exists, based on its claim that IARC hasn’t a clue whether or not it exists.
I think the inference makes more sense for coffee than it made for smoked meats.
If you do big studies, you can often prove that an effect is real (statistically significant) even if it is tiny (societally insignificant). In terms of health outcomes, abandoning bacon is no substitute for quitting smoking.
But a conscientious effort that fails to find an effect – that sometimes yields weak evidence of a relationship and sometimes yields no evidence at all of that relationship – really does tell you something about the likely size of the effect if it exists. Small effects are harder to find than big effects – so if there are lots of studies of the carcinogenicity of X and some found weak evidence of a link and others found no evidence at all, you’re on pretty strong ground concluding that either the effect is small or it’s nonexistent.
As always, IARC steadfastly refuses to explain any of this. It was reluctant to explain that declaring processed meat a known carcinogen didn’t mean it’s necessarily a serious one. And it is reluctant to explain that declaring coffee a substance that might or might not be a carcinogen pretty much does mean that if it’s a carcinogen at all, it’s probably not a serious one.
The declaration that very hot beverages are a probable carcinogen puts hot beverages in pretty much the same boat as processed meats (though with less confidence). As with processed meats, IARC wants to say drinking very hot beverages can probably cause esophageal cancer without providing guidance on how often it does so for specific frequencies of drinking very hot beverages, how bad and how common esophageal cancer is, and thus how important or unimportant it is to avoid drinking very hot beverages. The same questions it didn’t answer with regard to processed meats it isn’t answering with regard to hot beverages. (I’m overstating here. Typically, IARC partially answers these sorts of questions in Q&As accompanying its main news release.) Once again, I believe, IARC is counting on reporters and the public to assume that its answer to the “how sure are you” question implies an answer to the “how bad is it” question.
A few bottom lines for me with regard to your specific questions:
- Telling people about the quality of the evidence about carcinogenicity without telling them about the seriousness of the associated cancer risk isn’t very useful to the public. IARC would answer, I suspect, that it is useful to government agencies trying to decide what and how to regulate – which is true, and is IARC’s specified mission. But IARC aggressively seeks media coverage for its assessments, knowing they will be widely misperceived. I have to think this is intentional.
- Many science journalists have charged (sometimes in articles, often in tweets and other internal dialogues) that IARC does awful risk communication. I think it does clever though arguably dishonest risk communication. When IARC says that something is a known or probable carcinogen, it knows people will conclude that it is a serious health risk. It believes, probably correctly, that this misperception motivates people to change their behavior in ways that are probably (even if not provably) good for their health. It knows it can safely include boilerplate explanations of the difference between assessing the quality of the evidence and assessing the size of the risk without significantly diminishing this useful misperception. So it gets to have far more impact than it would have if people took the boilerplate to heart – all without actually saying anything false.
- As to whether IARC “knows more” about coffee than it knows about other Category 3 substances, I’m not sure I see the significance of the question. If you have virtually no studies about the carcinogenicity of X, you have to say you don’t know. If you have a lot of studies about the carcinogenicity of Y, and some find tentative evidence of an effect but many find no such evidence, you also have to say you don’t know. I suppose it makes a kind of sense to point out that you have more evidence showing you don’t know about Y than evidence showing you don’t know about X – but the bottom line in both cases is much the same. One difference that might actually make a difference: More research on X could easily change your decision about what category it belongs in. More research about Y is unlikely to affect the overall picture.
Use any of this you want to, Kai. (But please remember that I haven’t studied closely what IARC actually said about coffee and very hot beverages. If my descriptions of what they’re saying are mistaken, please let me know and give me a chance to rethink.)
If you think it would be helpful for us to talk, we can set up a time to do so over the weekend. But I suspect this email will give you more than you want from me.
Copyright © 2016 by Peter M. Sandman