Re: Interpreting the negative association of Cohen's data and lung cancer by county

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

On Thu, 30 Jan 1997, John Goldsmith wrote:

> Dear Radsafers:
>      I have enjoyed the many comments, many of which are quite resource-
> ful and provocative.  However, there is a (possible) very simple 
> explanation.  Lung cancer incidence has long been known to be higher in 
> urban than rural areas; the assumption that this was a reflection of
> general urban pollution was widespread, but it is not true.  The primary
> reason is that the urban-rural gradient in lung cancer is greatest where 
> pollution gradients are low, and in the few locations in which actual
> levels of pollution can be included, there is no significant contribution.
> Thus, we know that population density is positively associated with lung
> cancer in virtually every place it has been looked for.  Its mechanism
> is not known.  For understanding Cohen's gradients, let us first 
> recognize that lung cancer goes up with population density. His data
> clearly show that radon in negatively associated with population density.
> Once one accepts that radon is heavy and its source is in the ground, then
> it must follow that people in apartments have on the average lower radon 
> exposures than do residents in the single family home.
>      I have tried to get the data Bernie says is in an "anonymous" file
> to illustrate the paradox and to suggest some approaches to reducing its 
> impact, but in vain.  This argument is not directed to any hypothesis of
> linearity or dealing with a threshold.  It is directed to acceptance of 
> the repeatedly shown positive association  of lung cancer to population
> density.  Given the known source and dispersion of radon and its confirmation
> by measurements in U.S. counties, the relationship could not be other 
> than a negative one, unless the contribution of radon was  much greater
> than in miners, which is implausable.
>       John Goldsmith, Professor of Epidemiology, Ben Gurion University
> of the Negev, Beer Sheva, Israel.  gjohn@BGUMAIL.BGU.AC.IL
>  

----I don't understand John's problem in downloading our data and I have
offered to work with him on the specifics. Many others have successfully
downloaded it, and if anyone else is having problems, please let me know.
	An alternative is to let me know what calculations you want done
and trust me to do them. I am more than happy to work with anyone either
way.
	With regard to the rural-urban problem, this was treated in great
detail in Section L of my paper. I pointed it out to John and have had no
response from him about how he thinks it falls short. The "apartment
problem" as a bias on radon measurements is treated in the second and
third paragraphs of the second column on page 160 of my paper. I also
pointed this out to John and have had no response.
	I really think John should read these sections of my paper and say
what he thinks is wrong with them rather than just ignoring them in making
comments like the above.I have never ignored any of his several messages
concerning my work.
	I can't go thru the arguments in my paper on the above subjects
here, but perhaps the casual reader may be impressed by the following
simple observations: The strong negative correlations between radon and
lung cancer are observed and statistically indisputable if we consider
only the most urban counties; or if we consider only the most rural
counties; the same is true if we take as measures of urbanicity total
population, population density, % of income from manufacturing,% of income
from farming,average acrage of farms, or any of several other measures. It
is also true if we consider only counties with the largest percentage of
people living in apartments, or if we consider only counties with the
smallest percentage of people living in apartments, etc.
	If John (or anyone else) has a model to propose that will allow me
to do a specific calculation, they should please let me know -- it may
even earn them $2500. The model I use in Section L of my paper would be an
example, but you may have other models or objections to my model. I will
do the calculations including exploring the effects of varying input
parameters, and if you like you can check my calculations. In any case, I
would be appreciative if you could read my paper first and say why you
think that it falls short of covering the issue you raise. John has never
done this with my treatments of the urban-rural problem or with the
apartment problem.
	
			---Bernie Cohen