[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: Field's comments on Cohen's Observation
Bill Field has sent to the list (see below), the last of a series of messages he and I exchanged off-list.
To give RADSAFE subscribers some context, I have appended at the end of this message the earlier messages we exchanged.
I have also interpolated a couple of brief responses to his most recent message.
Best regards.
Jim Dukelow
Pacific Northwest National Laboratory
Richland, WA
jim.dukelow@pnl.gov
These comments are mine and have not been reviewed and/or approved by my management or by the U.S. Department of Energy.
-----Original Message-----
From: epirad@mchsi.com [mailto:epirad@mchsi.com]
Sent: Wednesday, June 18, 2003 11:22 AM
To: Dukelow, James S Jr
Cc: BERNARD L COHEN; radsafe@list.vanderbilt.edu
Subject: RE: Field's comments on Cohen's Observation
Dukelow stated that from LNT, it follows as the night the day that
radiologically-induced cancer incidence (and cancer mortality, given
reasonably uniform standards of treatment) is a linear function of person-rem
exposure. As such,an ecologic study is adequate to test the LNT, assuming
dependence and confounding issues can be handled.
I have two brief comments,
First, radiologically induced lung cancer is not a linear function as per
person-rem unless you can describe the population in pretty good detail. For
example, there are many factors that modify the assumed linearity such as age,
smoking, etc.
[JSD comment: I don't believe the "unless you can describe the population in pretty good detail" part of this comment. More to the point, Cohen's ecological data seems to contradict the LNTH (linear no-threshold hypothesis) assumption for human exposure to radon, and I am aware of a lot of animal data (reported in Stannard's Radioactivity and Health) that seems to demonstrate a threshold effect. I am similarly aware of numerous assertions as to the infeasibility of strong confirmation of a linear no-threshold assumption using epidemiological data, due to the expense of collecting and analyzing data from the extremely large sample that would be required.]
Secondly, you state that ASSUMING dependence and confounding issues can be
handled, an ecologic study is adequate to test the LNT"
That is really the heart of the problem isn't it? Dr. Cohen has not persuaded
me or from what I can tell most informed epidemiologist and statisticians that
he has indeed handled effect modifiers and confounders adequately. Dr. Cohen
has not addressed the within county joint distributions of even smoking and
radon let alone all the other socioeconomic factors that are also co-
correlated with smoking. Because Dr. Cohen uses ecologic smoking information,
he can not hope to correct for the problem I mentioned above regardless how
many attempts he makes.
[JSD comment: See below for my reasons for finding Cohen's analysis more persuasive than the anti-Cohen analyses.]
I previously sent Dr. Cohen a possible methodology to address the within
county joint distribution problems, but he indicated to me that he thought the
methodology had little to do with his work. Therefore, that created a real
impasse between us and is why I asked his cooperation in selecting a 3rd party
to weigh in on this issue (especially if he plans to find any potential NCRP
opinion on this topic biased). We are getting no where in these discussions
on the list and in fact the discussions have obviously become irritating to
some (for which I apologize in part).
Regards, Bill Field
==================================
Here are the Field and Dukelow messages that led up to Field's 18 June 2003 message to RADSAFE (see above):
Hi Bill,
I frequently kid, but rarely about issues as serious as this one.
There are a variety of sources of error in any statistical analysis. Cohen's data set is large enough to reduce sampling error is his analysis far below the level in "standard" case-control studies. His large data set allowed him to use stratification of the data to eliminate a number of potential sources of confounding. His approach to dependence issues has been to simply assume significant dependence between various sets of variables and establish that the dependence does not change the conclusions of the study. I have found all of this fairly persuasive and have not found counter-arguments particularly persuasive, although I would have to admit that, with the exception of your ongoing colloquoy with Dr. Cohen, I am a year or so behind in my reading of the contra-Cohen literature.
I talk to our statisticians and epidemiologists, but I have used probability and statistics professionally for several decades and have taught both to undergraduate and graduate students a number of times over the years. I believe I can form my own opinions about the standards of practice of statistics and epidemiology.
Finally, for complicated reasons, I also know a fair amount of biology and for biological reasons consider LNT to be extremely implausible. Thus, I never considered Cohen's results to be particularly surprising.
Best regards.
Jim Dukelow
Pacific Northwest National Laboratories
Richland, WA
jim.dukelow@pnl.gov
These comments are mine and have not been reviewed and/or approved by my management or by the U.S. Department of Energy.
-----Original Message-----
From: epirad@mchsi.com [mailto:epirad@mchsi.com]
Sent: Mon 6/16/2003 12:38 PM
To: Dukelow, James S Jr
Cc:
Subject: RE: Field's comments on Cohen's Observation
Jim,
Are you kidding?
>From LNT, it follows as the night the day that radiologically-induced cancer
incidence (and cancer mortality, given reasonably uniform standards of
treatment) is a linear function of person-rem exposure. As such, an ecologic
study is adequate to test the LNT, assuming dependence and confounding issues
can be handled.
Are you saying you can control confounding with an ecologic study? Ask the
epidemiologists or statisticians down the hall from you if that is the case.
************************************************************************
You are currently subscribed to the Radsafe mailing list. To unsubscribe,
send an e-mail to Majordomo@list.vanderbilt.edu Put the text "unsubscribe
radsafe" (no quote marks) in the body of the e-mail, with no subject line.
You can view the Radsafe archives at http://www.vanderbilt.edu/radsafe/