Sunday Science Lesson: “carcinogen” is not a useful category

by Carl V Phillips

I was recently asked by a correspondent if we could perhaps make the case that being in the same room as aerosolized cooking oil is more carcinogenic than being exposed to what might be called second-hand vapor.  The context was dealing with the dumb (really really dumb) law in California that requires every venue to post a sign that says, in effect, that being in this facility results in exposure to known carcinogens, and how that whole  California-style “known carcinogens” concept means that e-cigarettes are getting identified as such.

Of course, technically the California law does not require that the sign be posted in every venue, but merely those that have chemical carcinogens floating around in them.  But since this condition is met everywhere that contains, say, furniture or walls, effectively the rule requires posting the sign everywhere.  There is far more that is stupid about this law than I will cover here (e.g., a warning that appears everywhere has zero information content).  I will focus on its very premise: the concept of something dichotomously being either a carcinogen or not.

The only real dichotomy is between exposures that we have evidence that pretty clearly demonstrates that they cause cancer (under particular circumstances, for particular people, in particular doses, and with some possibly very small probability — i.e., the evidence pretty much never suggests that they always create a risk, let alone one that is large enough to matter) and exposures that probably cause cancer (with all those caveats) but the evidence does not exist yet. It is a safe assumption that under the right circumstances, any exposure experienced by a body can cause cancer.

Some of you might notice that this construction sounds disturbingly like the attitude of “public health” people (and sometimes even real public health people) that is epitomized in such statements as “this study failed to demonstrate that the exposure creates any important risks, so more research will have to be done to show that is true.”  But there is a fundamental difference here:  The “public health” people are looking for an excuse to take some action that they have a political/religious desire to take.  The implications of my point are the opposite, that the differences between the many exposures where it is surmised that their is some tiny risk are about the same as all those exposures where no one has chosen to surmise this yet, and in all such cases, almost any proposed actions or restrictions are inappropriate.

Notice that the latter statement applies to only those “known carcinogen” exposures where the guess is that there is some tiny risk of cancer, not exposures that cause important risks.  Just because there is no bright-line dichotomy does not mean that the world does not divide, and some exposures matter, of course.

Most carcinogen exposures — either the “known” ones or the billions of others — cause such a trivial risk that (a) they do not matter and (b) it is only possible to make wild guesses (often dressed up to look like science, but still just wild guesses) about the magnitude of the risk.  If an exposure does not cause a large enough risk, there is no way to detect it, let alone measure it.  It is difficult to imagine ever being able to detect a risk that causes 1  in 100,000 exposed people to get a cancer, even if that cancer were extremely rare and a lot of people were exposed.

You might argue that I should not have even go down to that order of magnitude to make the point, and that the same thing can be said about 1 in 10,000.  In reality, measuring risks in the 1 in 1000 range is beyond practical capabilities.  E.g., while there is no clear evidence that smokeless tobacco causes any risk for oral cancer, it is plausible that there is a relative risk in the order of 1.1 or perhaps even 1.3; these figures would put the risk in that 1 in 1000 range, but despite the exposure and outcome being easy to detect, the outcome being rare, and the great interest by ANTZ in the topic, it is not possible to conclude whether the risk is 1 in 1000 or zero or somewhere in between.  Similarly, any risks from second-hand smoke are down in this barely detectable range.

To the extent that quantification even exists, most of the wild guesses (that get called calculations or extrapolations) about the vast majority of “known carcinogen” exposures put the risks in the order of 1 in a million or 1 in a billion.  To put that in perspective, the risk of dying from driving 1 km is about 1 in a billion.  Yes, you read that right — the lifetime of exposure to some of these “known carcinogenic exposures” that the California-types want you to worry about cause about the same risk of getting a cancer (not even necessarily a bad cancer) as the risk of dying from driving a few blocks.

So with that in mind, there are a couple of key lessons here.

First, the one in the title: Since basically anything is a carcinogen under the right circumstances, merely calling something a carcinogen is uninformative, and quite often is fear-mongering disinformation.  Of course, most of us use the word, but usually we mean something like “carcinogenic, under the relevant circumstances, to a degree that matters or is at least measurable”, and that is how most people interpret it.  Thus, when it is used to describe something that poses a risk that is trivial or speculative, it is terribly misleading.

Second, my last disjunction is a bit misleading because the trivial and the speculative are mostly the same category.  That is, when something is guessed to produce a very small risk, it is never known for sure that this risk exists at all.

Third, given that, it should be clear that it is impossible to compare risks down in that trivial range.  Is the risk from an exposure 1 in a billion, or 1 in a million, or 1 in 10,000 (or zero)?  We will probably never know.  Thus, there is simply no basis for concluding that one exposure that is down in this range is a worse than another.  Of course, all differences among miniscule risks are themselves miniscule, so it does not matter anyway.  Anyone making such comparative claims — unless accompanied by a lot of caveats — is lying.  And, unfortunately, that would include us if we started trying to make such comparisons.

Irrational fear of and despite toward “carcinogens” is the biochemistry equivalent of racism or homophobia.  They are all driven by tendencies toward tribalism (an unconscious desire to find someone to hate) and fear (at the unconscious animal level) that are part of what we inherited from our ancestors.  There is a bit of value in responding to any of these with facts, of course — but the simplest facts (those people are mostly exactly like you; trivial carcinogenic risks are everywhere) probably matter just as much as any details.  There may be a role for the scientific details in, say, court proceedings.  But for changing the general prejudices, even we scientists have to concede that a real solution comes most from in the subconscious release of irrational hatreds and fears that comes from familiarity.

10 responses to “Sunday Science Lesson: “carcinogen” is not a useful category

  1. Us California types are a little tired of hearing how whacked out we are. Although we have gone off the deep end on a few things we are at the forefront of mostly good stuff. Calling for warning signs on everything is obviously going too far. I believe it was a push back on big business types telling us that there is nothing to fear with all the chemicals in our lives. Cancer and other strange illnesses rates have skyrocketed in the last 40 years and we still really don’t know what the hell it is. The anti smoker folks have obviously been empowered by the last 30 or so years of over regulating and it’s a tough sell to change their ugliness. I’m not a scientist but I feel like one after reading this blog ;)

    • Carl V Phillips

      Thanks for that last bit. That is what I am going for. And a bit of apology for the glib caricature, but it kind of an easy way to make the point. (Oh, and I have lived in California about as much as anywhere else during my adult life — and right in the heart of the stereotype, in the East Bay area.)

      The claim you make, though, is very much part of the problem. The rates for most all cancers have gone *down* over the last 40 years (measured properly — based on age so that you are not just measuring the fact that the population lives longer, or in epidemiology-speak, that “competing causes” of mortality have been reduced). For most of those that have famously increased, we can identify viral causes, not chemical ones — so, yes, it was that California-led sexual revolution that caused the increase! Or if you go back further, smoking caused a big increase in some cancers in the 20th century, of course (though not during the time frame you cite, since it had already happened).

      There are some unexplained increases in strange illnesses (severe allergies in particular) that *might* be due to subtle effects of novel chemical cocktail exposures, though there are other compelling explanations. Those who want to blame novel chemical exposures for political reasons make claims, but they have not demonstrated much that is compelling.

  2. Thanks for the clarification and keep up the good fight!

  3. To put that in perspective, the risk of dying from driving 1 km is about 1 in a billion

    got a link for the calculation on that? not to dispute, just an interesting thing to use if the source is readily available.

    • Carl V Phillips

      Yes: :-)

      But really, it is a simple calculation. I just roughed it out in my head for the post.

      • Carl V Phillips

        Ok, given your great comment on the Polosa article, you deserve a non-snarky response :-) (I will try to write something about that series of comments sometime — it is very educational)

        Estimate 30K km/yr of motorized transport for 100 years = 3 million km. Motorized transport has about .33% chance of being someone’s cause of death. Divide to get 1/1B of a statistical death per km. Note that all of these numbers bias upward the number of km you have to drive to get one statistical death (average driving /yr is less than that; few people have 100 years of exposure, probability of death is a bit higher than that). So there is actually more than 1/1B of a death per km, but it is closer to that than 1/100M, so it is conveniently the right order of magnitude.

        • why did I not see that coming…

          anyway! as I thought your first response would be the only one I might get I searched about and found something related on this by David Spiegelhalter.

          they have a 1 in 1M probability of death for 250 miles driven (UK), so this does indeed seem to work out at around 4 in 1Bn (er I think…) per mile.

          bringing this up as the ‘micromort’ they talk about seems like a neat concept in the whole ‘risk communication’ business, something worth adopting in THR or just a fad?

          a quick recalculation of the old ‘lifetime use of low-risk product compared to X months of smoking’ could translate that to ‘micromorts’… but I’m never sure if these things that sound a nice ideas are just too gimmicky once you try to use them.

        • Carl V Phillips

          Yeah, no magic to it. That is the same back-of-the-envelope, just with Imperial units. As I said, my rounding errors biased my number a bit in the direction of underestimating the risk, but 1/1B is still about right.

          I have seen “micromort” elsewhere, and it has been around a very long time. So not a passing fad, but not totally common. It is easy enough to explain that it might as well be adopted. I would hesitate to convert my “compared to 2 months of smoking” into absolute numbers, though, for two reasons. First, it would almost certainly imply much greater precision than we have. Second, while these numbers are useful for risk analysis, they actually diminish the explanatory power when communicating. So (applying your suggestion) we could say that the risk from a lifetime of low-risk product use, or of the average smoker smoking for a couple more months, is a mortality risk of about 3000 micromorts (or, to put it in normal terms, is about 1/300) using the numbers we kick around. But even with just one sig fig, that makes it seem waaaaaaay more precise than it actually is. And it is also very likely to be taken wrong — “hey, that is not zero! ban it!”

          Notice that the way I used these numbers was to convert from quantities (1/1B level cancer risks from environmental chemical exposures) into everyday thinking. Going back the other way is likely to confuse more than elucideate.

  4. Pingback: Sunday Science Lesson: “carcinogen” is not a useful category | vapeforlife

  5. In the phrase ” This product contains X known carcinogens”. The X can refer to known human carcinogens plus known animal carcinogens. Most people will interpret the message as X representing only known human carcinogens. Not sure if the CA message is restricted to known human carcinogens.

    Of course if [ed note: comment ended in the middle of the sentence]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s