Remember the Ames test ? Test a substance and if it causes mutations then it ought to be a carcinogen, right? After all, isn’t cancer the result of a mutation that causes a cell and its descendants to grow out of control? That was the thinking anyway. Unfortunately (or fortunately) lots of things that cause mutations don’t cause cancer and as a result the Ames test sparked more than a few bogus health scares and plenty of baseless tort cases.
Now, in a flashback to the days of starting with an unproven premise about the origin of cancer and stacking inference upon inference upon inference until mass litigation over an industrial chemical emerges and metastasizes, molecular epidemiology gives us an association between trichloroethylene (TCE) and non-Hodgkin lymphoma (NHL). See: "Occupational Exposure to Trichloroethylene is Associated with a Decline in Lymphocyte Subsets and Soluble CD27 and CD30 Markers".
So what of this association between TCE and NHL? Have workers with long occupational TCE exposures gotten NHL at a high rate? No. Has there been however a small but consistent increase in NHL among TCE workers? No. So the causal claim fails the two big prongs, strength and consistency of the association, of every causal inference test. End of story, right? Not anymore.
Cancer epidemiology had a great run. Uncovering the epidemics of cancer in chimney sweeps, smokers and many others were profoundly important discoveries. Yet a decade ago many were wondering if it was time to call it a day for epidemiology. The strong and consistent causes had already been rooted out while subtler causal associations like that of H. pylori and peptic ulcer disease were being missed and non-existent associations between electricity, peanut butter, etc and cancer were repeatedly being discovered one week only to be refuted the next.
But if the tools work why not apply them to the new and fertile ground of molecular biology? In other words, why not look for strong and consistent associations between molecules as a way to uncover the links in the causal chain from genotype to phenotype? It was a great idea but nature failed to cooperate.
First, the reductionist approach to genetics ran into trouble when it became clear that a gene doesn’t do the equivalent of say executing schematics for an alternator but rather is part of a larger algorithm dedicated to finding a way to convert available mechanical energy into electrical energy. Second, it became apparent that within largely known biochemical pathways there were often so many other factors influencing outcomes that meaningful p-values might often end up having something like "x 10-6" appended to them – in other words you might want to consider rejecting the null hypothesis only if the odds of the results observed being due to chance alone fall below one in a million.
Sadly none of that will stop some from arguing that TCE causes NHL and citing the new paper referenced above. Read it through and here’s what you’ll learn: Eighty workers were exposed to TCE. None got sick. However, they had lowered levels of lymphocytes, some of which lymphocytes have something to do with some of the cells involved in the immune response. Meanwhile some other people with some unstated change in their immune response are sometimes at a somewhat increased risk for NHL. Therefore it’s somewhat more biologically plausible that TCE is possibly a cause of NHL.
I kid you not. Nevertheless you can bet that this paper will be cited in courtrooms and the Federal Register for the proposition that TCE causes cancer just as the Ames test was cited for the proposition that coffee causes cancer.