Via Tyler Cowen at Marginal Revolution I came across this post earlier in the year from a blog called EffectMeasure on the use of rodent models to predict human cancer risks. It's a broadside against the American Council on Science and Health and a petition they filed against the use of high-dose rodent carcinogenicity tests.

Quote the anonymous "Revere":

The main rhetorical lever ACSH employs is the use of high doses in the animal studies, doses that are much higher than usually faced by humans. But as ACSH knows well (but didn't divulge) there is a technical requirement for using these doses. If one were to use doses in animals predicted to cause cancer at a rate we would consider a public health hazard, we would need tens of thousands of animals to test a single dose, mode of exposure and rodent species or strain. This makes using those doses infeasible. Thus a Maximum Tolerated Dose is used, one that causes no other pathology except possibly cancer and doesn't result in more than a 10% weight loss. The assumption here is that something that causes cancer at high doses in these animals will also do so at low doses. This is biologically reasonable. It is a (surprising) fact, that most chemicals, given in no matter how high a dose, won't cause the very unusual and specific biological effect of turning an animal cell cancerous. Cancer cells are not "damaged" cells in the individual sense but "super cells," capable of out competing normal cells. It is only in the context of the whole organism that there is a problem. It is not surprising, then, that very few chemicals would have be ability to turn a normal cell into a biological super cell of this type. Estimates are that is far less than 10%, perhaps only 1% of all chemicals that have this ability. Thus western industrial civilization doesn't have to come to a screeching halt if we eliminate industrial chemical carcinogens from our environment.

We know of no false negatives with this process. Every chemical we know that causes cancer in humans also does so in rodents (with the possible exception of inorganic trivalent arsenic, which is equivocal). The reverse question, whether everything that causes cancer in animals also is a human carcinogen, is not testable without doing the actual natural experiment: waiting to see if people get cancer on exposure, an experiment ACSH is only too happy to conduct on the American people to make their corporate sponsors happy."

I've left out (as did the MR post) the part where he called the ACSH "right wing whores", which is the kind of thing that doesn't enhance the statistical arguments very much. Dropping the invective, I want to take up Tyler Cowen's question: is there anything to this critique? My answer: there might be. But there might not be. It's certainly not as clear-cut as the author would like to make it, cancer epidemiologist though he is, which would seem to be one of the criticisms he's making against the ACSH petition.

Here are some complicating details:

1. The effects of high doses of compounds can be due to their effects on cell division. At such levels, test substances cause irritation and inflammation that promotes cell proliferation. The more cells are forced to divide, the more opportunities there are for the defects that lead to cancer. These effects do not scale well to lower doses. It's the opinion of Bruce Ames (inventor of the Ames test genotoxicity screen) that this problem has completely confounded the interpretation of high-dose animal data. (His article in Angew. Chem. Int. Ed. 29, 1197, 1990 is a good statement of this argument).

2. The statement that "most chemicals, given in no matter how high a dose, won't cause the very unusual and specific biological effect of turning an animal cell cancerous" is not accurate. As Revere surely knows, there are many mutations and pathways that can turn a cell cancerous (which is why I keep harping on the idea that cancer isn't a single disease). Somewhere between one-third and one-half of all synthetic chemicals tested in cell assays or in high-dose animal assays show up as possible carcinogens, depending on your definitions. Interestingly, basically the same proportion of natural products (isolated from untreated foods and other sources) show up as positives, too.

Now, if you want to talk confirmed human carcinogens, then Revere may have a point. There are only some three or four dozen specific chemicals that are confirmed as causes of human cancer. Here's the list. If you read through it, you'll note that many of the 95 agents on it are radioactives or broad categories such as "alcoholic beverages." (Mention should be made of things like nickel, all compounds of which are under suspicion. Check your pockets, though, for your most likely exposure). Specific compounds known as human carcinogens are quite rare. But doesn't that fact support the ACSH's point more than Revere's?

3. Revere's statement that "Cancer cells are not "damaged" cells in the individual sense but "super cells," capable of out competing normal cells. It is only in the context of the whole organism that there is a problem" is also inaccurate. Cancer cells are indeed damaged, right in their growth-regulation and/or apoptosis pathways. A car whose throttle is damaged will run at a higher RPM than a normal model, but I wouldn't call it a "super car". And cancerous cells are often quite recognizably problematic, whole animal or not. They divide like crazy in petri dishes, the same as they do in an animal.

4. The majority of the cancers seen in rat and mouse models are in the liver (which supports the idea that these tumors occur through general strain on their metabolic systems). Human liver cancer is much more rare. The most common human cancer in many countries is lung, caused to a great degree by smoking (which is also likely to have constant-irritant cell-proliferation component). Of the agents on that ICAR list in point #2, only three or four are chemicals (or mixtures) known to induce human liver cancer specifically. This is a significant mismatch.

5. Revere states that "We know of no false negatives with this process. Every chemical we know that causes cancer in humans also does so in rodents. . ." But how about false positives? There are hundreds of compounds that seem to cause cancer in rodents that (as far as we can tell) do not pose a risk to humans. I say "seem to", because these are almost always high-dose studies. But I can even think of some compounds (the PPAR-alpha ligands) that cause all sorts of trouble (including tumors) in rodent livers at reasonable doses, but don't do so in humans. Rodent tox is necessary, but it sure isn't perfect.

There, that should be enough to complicate things. It doesn't make for as dramatic a story as the evil henchmen poisoning America on behalf of their corporate masters, I have to admit. But we'll have to try to get along without the excitement.