Response to Mossman’s “Response to Cohen’s Letter” in
Health Physics News, July 2003, page 10
1a. “Cohen does not refute Puskin’s conclusions”
----Puskin’s conclusion is that my results and his observation are explained as due to errors in my values of smoking prevalence, S, for each county; he concludes that my S-value miss a strong negative correlation between S and radon levels, r. I show that even an infinitely strong (I called it “perfect”) negative correlation does not improve on the paradox posed by Puskin’s observation, and does not explain the huge discrepancy between my results and the predictions of linear-no threshold theory (LNT). Why does that not refute Puskin’s conclusion?
1b. “Cohen admits he cannot explain Puskin’s observation”
---As I show in 1a above, Puskin’s explanation fails. My suggested explanation is offered in the fourth paragraph of my letter.
2. “Lack of evidence that radon causes cancer in tissues other than lung”
---I don’t know that anyone has looked for such evidence. But, in any case, I did not claim that radon causes these cancers. I only suggest that processes by which low level radiation may reduce the risk of lung cancer may also apply to those cancers. I give examples of such processes.
3a. “inappropriate to interpret Cohen’s data to mean that radon reduces lung cancer”
---I have never claimed that my data mean that radon reduces lung cancer; that would be an application of “the ecological fallacy”. If LNT fails, I can’t logically interpret my data in quantitative terms. If LNT is assumed to be correct, its predictions are grossly discrepant with my data. Putting these two things together, I conclude that LNT fails.
3b. “why does Cohen talk about protective mechanisms”
---They are a not unreasonable way to explain the data. I cannot prove that they are the explanation, although Puskin’s observation pushes me toward that idea. All I claim to have proven is that LNT fails.
3c. “Krewski … concludes that none of the analyzed studies reported a statistically significant negative association”
---That is very different than saying that they show strong evidence against a negative association. That is what would be required to justify Mossman’s original statement “Cohen … and the case-control studies cannot both be right” (if, as Mossman incorrectly assumes, my data is interpreted to represent a dose-response relationship).
4. “reward to anyone who could explain Cohen’s negative correlation”
---My reward was for a plausible explanation for the discrepancy of my data with the strong positive correlation (more accurately “slope” of lung cancer vs radon) predicted by LNT. That is very different from reducing the negative slope. In any case, I do not consider an infinitely strong (perfect) correlation between smoking and radon to be “plausible”. The fact that this perfect negative correlation would eliminate the negative slope was shown in my 1995 paper in Health Physics. Plausibility considerations are discussed there, and much more in item #7 on my web site.