Cell Phones, Cancer and the Ambiguity of Causation

Incredible article from the New York Times:

See: Do Cellphones Cause Brain Cancer?

Our vision of carcinogenesis has become vastly more complex since 1853. We now know that there is no “one cancer.” Breast, lung, prostate and blood cancer share a similarity — the uncontrolled growth of cells — but the specific genes and behaviors of these cancers are far from identical.

Nor is there “one material for cancer” — one archetypal carcinogen. Agents that cause cancer are chemically diverse and cancer-specific. Estrogen can provoke cancer in the breast, but destroys prostate-cancer cells; vinyl chloride is exquisitely carcinogenic to the liver but not to the skin; chlorine and nitrogen mustard are both poison gases, but only one causes leukemia.

Notably, there is also no “one test” for carcinogens. Scientific studies to capture the association between an agent and cancer cast an astonishingly wide net. On one end of that spectrum lie populationwide human trials involving hundreds of thousands of men and women. On the other end are precise laboratory experiments that plumb the molecular depths of cells and genes. The tests range from the telescopic to microscopic, from statistics to biochemistry — from observations of chimney sweeps to bacteria on a petri dish. Often one test must be corroborated by another. Asbestos and tobacco were identified by case-control studies and validated in animal models. Estrogens were implicated by studies on human and animal physiology and then found to be carcinogenic in prospective human trials.

Finding a carcinogen, in short, is not like solving a mathematical equation, with a single formula and solution. It is more like solving an epic detective case, with individual pieces of evidence that, taken together, suggest a common culprit.


  1. jimi
    Apr 26, 2011

    There’s a basic misunderstanding of how science works in that article. “Proving” a negative isn’t really possible in those sorts of designs. You can reject or fail to reject a null hypothesis, but you can’t “prove” one. As such what the author seems to want won’t ever come in (proof that there is no causal relationship). But with the number of failed tests that have sought to establish a link, the consensus would suggest there is no association. Some friends of mine just published a very good article in the American Sociological Review that actually made just this point.

Submit a Comment