Re: Radon and lung cancer

[David Scherer <scherer@uiuc.edu>]

Bernie Cohne said:

> All I ask is that someone offer a not implausible possible
>explanation for my data that is consistent with LNT. Is that asking too
>much? I have offered a sizable monetary reward, and this would settle an
>issue that is bothering a lot of scientists. I am quite sure it would be
>publishable. It seems to me that, in the absence of such an explanation,
>it is reasonable to conclude that the linear-no threshold theory fails in
>the low dose region.

I have already suggested (>1 y ago) that migration could cause the results.
 If someone is exposed to radon in a high radon county but dies of lung ca
in a low radon county, your method attributes the death to low radon
exposure.  Conversely, if someone lives most of their lives in a low radon
county, but dies of not-lun-ca in a high radon county, this also affects
your results in that county.  Your response was that the effect persists
when you look at the highest and the lowest counties alone.  I don't see
how that address the question.  Even if one accepts concentration as a
surrogate for dose, you don't know the correct concentration for anyone in
the study.  You also asked me for something you can calculate.  How can I
give you something to calculate when you have no data on it?  I have
suggested a plausible explanation, but I cannot prove it is the case
because these variable are not in your data set.  This is a problem in the
experimental design, a systematic error, not a problem with numerical
analysis.  The statistical errors in your study are excellent, but you
cannot address the question of systematic errors (so-called confounding
variables).

I will suggest another potential source of error: smoking.  Two comments:
(1) Have you shown that cigarrette sales is a good surrogate for smoking
history?  (2)You say that ecologic studies are only valid for testing LNT
models.  What if smoking risk is non-linear?   Couldn't your results
reflect the inadequacy of the (several) correction methods you have used?

I believe you have said that you have  used a model from an appendix of
BEIR 4.  As a polemical tool, this is very nice: you can hoist BEIR on its'
own petard.  But in a search for truth, it is lacking.  In a case-control
study one can use actual, individual smoking histories and radon exposures
to look for covariences INDUCTIVELY.  Using an _a_priori_ model to correct
the data amounts to a RETRODUCTION and the results are no better than the
correction.

Regards,
Dave Scherer
scherer@uiuc.edu