Re: RADON - Ecologic Studies

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

On Tue, 3 Mar 1998, R. William Field wrote:

> Dr. Cohen,
> 
> I truly understand that you want someone to make a suggestion in
> quantifiable terms that offers an explanation for your findings.  That is
> why we took the time to address your requests in the past.  
> 
> It is my opinion, that people have offered very plausible examples to
> explain how an inverse associations can occur.  I do not think you should
> always expect a scientist to continue following up every rebuttal you
> offer.  Maybe, they thought they made their case with their initial
> presentation and that no follow-up to your rebuttals were necessary.
> People have other commitments other than reviewing your research or
> rebutting your claims.

	--Ultimately, both sides should be presented in the scientific
literature, and let the scientific community decide. This is all that I
expect, or that any scientist has a right to expect.
	Discussions on RADSAAFE may be useful preliminaries to this
process. Nearly all of these discussions have been initiated by others
and I have felt compelled to respond.
> 
> The factors that are causing your inverse association may be
> unidentifiable, do you understand that?  

	--I have addressed this question at great length in my papers.
That is why I have studied 500 potential confounding factors, and
considered what properties an unidentified confounder must have to resolve
our discrepancy.
	Do you understand that unrecognized confounding factors can cause
false results in any case-control study? They typically consider less than
5 potential confounding factors, while I have considered 500. 

> 
> Greenland and Morgenstern have shown that extraneous risk factors
> responsible for ecologic bias may not even be confounders or effect
> modifiers at the individual level.  So in all cases, you may not be able to
> describe the bias in quantifiable terms.  Do you understand this?
> 
	--In my papers, I have responded at great length to the
Greenland-Morgenstern points as applied to my work.
	We could argue endlessly about these very general points,
but the way to settle the question is for someone to suggest a not
implausible specific explanation for our discrepancy in concrete terms. I
will then do the calculations to test whether they are "not implausible".
In the paper you submitted to HEALTH PHYSICS, you did make such a
suggestion and my calculations showed that it could not resolve  or
appreciably diminish our discrepancy. You are always welcome to submit
another suggestion, as is anyone else.


> Further, The problem of identifying the bias is made even more difficult by
> the fact that factors responsible for ecological biases may not even APPEAR
> to be confounders or modifiers at the ecologic level.
> 
	--This is too general for me. How about a specific example?


> There will be no bias in an ecologic study if both the background (rate of
> disease) and the exposure effect do not vary across groups, and there is no
> confounding within groups.  Small departures from these conditions result
> in substantial bias, even reversing the direction of the observed
> observation.  

	--How about a specific example applicable to my work?

> For other radsafers - I highly recommend the paper by Greenland and
> Morgenstern on this topic, (International Journal of Epidemiology
> 18:269-274, 1989).
> 
	--How about the responses to their points, as applied to my work,
in my papers?



> Dr. Cohen - I have 2 direct questions for you.  I hope you can give me a
> direct answer. 
> 
> You apparently have been trying to find the reason you are finding an
> inverse relationship in your data for many years.  You have not been able
> to find a reason. 
> 
> 1) Do you truly believe that if no one else can offer a explanation for
> your findings (that is an explanation that is acceptable to you), then your
> findings have the required validity to scrap the LNTT?   
> 
	--It doesn't have to be acceptable to me. Let their suggestions
and my calculations based on them appear in the scientific literature and
then let the scientific community decide. That is the way science works.

> 2) Is there ANY chance that your inverse ecologic findings can be
> attributed to aggregation or specification bias?

	--These issues have been discussed in my papers. If you are not
satisfied by my treatments, please say why. The short answer to your
question, based on what I know, is NO.
> 
Bernard L. Cohen
Physics Dept.
University of Pittsburgh
Pittsburgh, PA 15260
Tel: (412)624-9245
Fax: (412)624-9163
e-mail: blc+@pitt.edu