Deconstructing Milward (part III)
When we first wrote about Milward we said it got causation wrong. The second time we wrote about it we took issue with the opinion for calling paid-for induction the scientific method. Now we address why we deem the opinion radical.
Should an educated guess, a hunch, especially one that's unverified and worse yet unverifiable, be admissible under Rule 702 of the Federal Rules of Evidence? Not if you're a fan of the Enlightenment and of the empiricism running from the likes of Bacon, Locke, Berekely and Hume to the Framers of our Consitution and all the way to Daubert. The idea that someone could be deprived of life, liberty or property on the basis of a mere guess, however impressive the credentials of the guesser, would we think shock the conscience of most Americans. But embracing without saying so the opposite of empiricism, something between Romanticism and and postmodernism, the First Circuit has said it's A-OK. That's pretty radical.
Defenders of the opinion might object to our claim that the opinion of Milward's expert was nothing but an unsupported and unsupportable conjecture. Didn't he offer five pieces of evidence after all? It depends on what you call evidence.
We all know what evidence for a theory looks like in the hard sciences. Albert Einstein makes a prediction about gravity and thereafter other scientists carry out experiments to see if the prediction pans out. Sometimes the answer comes in quickly and sometimes it takes a little longer but ultimately the conjecture stands or falls not on the reputation of the man but on the ability of his theory to withstand all attempts to refute it. Those attempts at refutation that instead confirm a theory's predictions are one form of evidence. In the somewhat softer science of biology including the study of human diseases such evidence tends to be harder to come by largely because biological systems are made up of so many different parts; the existence and complete function of many of which are still largely unknown.
That doesn't, however, mean that theories about biological function can't be tested. Lab animals as proxies for humans can be exposed to suspected carcinogens; their genes can even be manipulated to see if effects observed are the result of some interaction between agent and genome. Randomized controlled trials are the gold standard of drug testing evidence and though you can't experiment on humans by exposing them to known toxins you can certainly compare the outcomes of those exposed in the past to those not exposed to see if there's an association and if so how strong it is. So what sort of evidence did Milward's expert have to offer?
He offered five "bodies of evidence". First, he said that benzene is a known cause of AML. Ok, but plaintiff had APL. Just as the old expert witness canard "cancer is cancer" retreated to "blood cancer is blood cancer", the claim that all forms of AML are different but really the same when it comes to causation is, as the National Academy of Sciences has said, an assertion for which there is no evidence and one which seems odd given the fact that the various forms of AML have unique presentations, treatments and outcomes.
Second, he said that a particular mutation "is often in part" due to a chromosomal translocation in a precursor cell that can lead to all types of AML, that CML (which is almost certainly not caused by benzene) "often in part" has the same mutation and that APL and other AML subtypes with the mutation share "common genetic susceptibility factors", risk factors and incidence (after age 20). Therefore, they all share a common etiology. Hmmm. Well, ask Bruce Ames if all mutagens are carcinogens. And is there any evidence of causation beyond an association that a translocation "often in part" coincides that the mutation? No. Finally, sharing some risk factors some of the time during a certain period of life doesn't get you anywhere because risk factors after all are themselves nothing but reported associations. What we're left with then is as follows: APL shares some common characteristics with other subsets of AML and sometimes they have a common mutation which mutation is "often in part" caused by a translocation which may be caused by benzene. Thus, while "all AMLs have a common etiology" is an hypothesis that can be formed from such bits it's one for which no confirmatory evidence currently exists.
Third, benzene can cause damage to the chromosomes of stem cells in the bone marrow and that "leukemia cases associated with benzene exposure are more likely to contain clonal chromosome aberrations than leukemias arising in the general population." Presumably, though the opinion doesn't say so, benzene is at least sometimes associated with the translocation that "often in part" causes the mutation that supposedly causes APL. Good. At last something testable. If benzene is a potent APL-agen then people exposed to benzene ought to be at a much greater risk of APL.
Fourth, two benzene metabolites are topo II inhibitors and some topo II inhibitors cause so-called secondary APL cases. Fine. Benzene can be broken down into some things that act like other things that can cause APL. Good for hypothesis generation but where's the evidence that the two benzene metabolites survive to reach and act upon stem cells like chemotherapeutic drugs? Indeed Milward's expert admits that he has merely suggested a plausible mechanism.
Finally, there's a small increase in risk of APL (e.g. RR of 1.4) in one epidemiological study and a few others which which do not refute the notion that benzene may have something to do with APL because there isn't a complete absence of APL among those exposed to benzene. Well, here we are at the lick-log and there's nothing here but a proclamation that at least his hypothesis hasn't been refuted. Worst of all, Milward's expert claims that his theory isn't even testable, saying that the rarity of APL "makes it nearly impossible to perform a large enough study" to answer the question of whether or not benzene causes APL.
So that's it. The only test, an epidemiological study of people exposed to benzene, that could possibly refute his theory that benzene causes APL has produced results that are at best equivocal. Oh, and by the way, a test that would actually answer the question is impossible to do. The trial court concluded that without evidence from a test of his theory that verified its prediction (exposure to benzene causes more cases of APL than would otherwise occur) the expert had nothing more than a hypothesis. The First Circuit took a different view.
When we think of the scientific method we think about something very much like this chart. From the green box on down it's all about empiricism, testing, about "bending over backwards to show how you are maybe wrong" and about always publishing the results of any test of one's theory "whichever way it comes out." The First Circuit has however adopted a very different sort of view of the scientific method - one that stops at the yellow box, one that dispenses with the need for proof. One that trades verification for belief.
It's one thing to hold the philosophical view that there is something oppressive about the scientific method. It does, after all, have the well deserved reputation of going about willy-nilly stomping on all manner of cherished beliefs (e.g. the sun revolves around the earth, stress causes peptic ulcers - sorry A.B. Hill, vaccines produce autism, etc.). It's something else entirely though to let an unverified and unverifiable belief, pseudoscience in fact, be the basis upon which a citizen may lawfully be relieved of his hard-earned cash. Yet the view that objective truth does not exist, that fervent belief by a well-credentialed expert originating from nothing more than his subjective weighing of whatever he thinks relevant is enough and that testability, the demarcation between science and pseudoscience, is not a minimum requirement for the belief to be admissible has now made it from comment c, section 28 of the Restatement (Third) of Torts: Liability for Physical Harm to the First Circuit. Now that's radical.
