Re: Radon Health Hazard, Myth or Reality?

[Bernard L Cohen <blc+@pitt.edu>]

On Tue, 4 Nov 1997, John Goldsmith wrote:
>.  Part of the problem is that the analysis depends on
> the lung cancer data for the totality of the county residents and the
> measurements are from a (usually) self-selected subset of households.

	--The validity of my radon data is a very long and complex
question that cannot be explained in a few sentences. I never dreamed of
using self-selected data when I got it, or even for years after confining
my considerations to randomly selected data. The evolution of my thinking
on this point is described in some detail in my papers. 
	But if my study is confined to randomly selected data alone, the
same results are obtained.

> The self-selection favors the more suburban dwelling units, while lung
> cancer is strongly associated with population desnity.

	--My results are the same if I consider only areas of highest
population density, or only areas of lowest population density, or only of
various in between population densities. The same applies to numerous
other socioeconomic variables.

> Professor Cohen has made some of his data available,

	--*All* of my data have been available to any requestor for a very
long time

> and I have
> recently examined a subset obtained by EPA.  Briefly, the data for
> counties in eight states with high radon levels in the EPA subset, show no
> significant associations between log radon and lung cancer for males or 
> for females.  (Lung cancer rates are age adjusted.)

	--Why only eight states, and how were these states chosen? How do
you justify such a choice?  I did analyses based on all of the
EPA data and got the same results as from my combined data set; these
analyses are given in my paper. Why regress against log radon? I was
testing the linear-no threshold theory which makes predictions for lung
cancer vs radon, not vs log radon.
	What did you do to test for confounding factors? I studied over 60
confounding factors in my original paper and have now added about 450
more, but my results were not affected.
	What is the basis for your claim that your study supersedes mine?


>                John Goldsmith, M.D, M.P.H.  Professor of Epidemiology,
>                 gjohn@bgumail.bgu.ac.il  Ben Gurion University,
>                   Beer Sheva Israel
>