$2500 award

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

	My offer of $1000 awards for help on the following matter did not
prove fruitful, so I am raising the award to $2500 and will try to explain
in more detail what I am seeking .
	In the February 1995 issue of HEALTH PHYSICS (vol.69, pp157-174),
I published a paper entitled Test of the Linear - No Threshold (LNT)
Theory of Radiation Carcinogenesis.....    in which I reported that lung
cancer mortality rates for U.S. Counties, with or without correction for
smoking prevalence, decreases rapidly (about -8% per pCi/L) as average
radon exposure increases. This represents a very large discrepancy (20
standard deviations!!) with the prediction of LNT theory that lung cancer
rates should increase rapidly (about +7% per pCi/L) with increasing
average radon exposure . My problem is in understanding this discrepancy.
	I have examined the effects of over 60 confounding factors, and
have done many other tests, but this work has done little to explain our
discrepancy. I have gone through the literature on Recological studiesS
and can easily show how the results of any other published ecological
study can be erroneous, but I cannot figure out how one can avoid
concluding from our data that LNT theory fails in this low dose region
where it has never been tested.
	What I need very badly is suggestions for not implausible specific
potential explanations for our discrepancy, in at least semi-quantitative
numerical terms, on which I can carry out calculations to determine if
they can resolve it, or can be modified to resolve it. As a possible
example, one might suggest that urban people smoke more frequently and for
unrelated reasons have lower radon exposures than rural people, both of
which are true. What I need is data for each of our 1601 counties on which
to do calculations to see if they resolve our discrepancy. You can make-up
the data, as long as you consider them to be not implausible. Since I need
these made-up data for each of the 1601 counties, it might be most
practical to give me a prescription for deriving these data. For example
you might say that the radon exposure for a rural person is x% higher than
for an urban person and an urban person is y% more likely to smoke than a
rural person. Since I know the average radon level in each county, the
fraction of people in each county who are urban and rural and the fraction
that smoke, I can then determine the predicted lung cancer rate in each
county from BEIR-IV for various values of x and y, and make comparisons
with the data.  
	The only problem with this example is that I reported calculations
based on it in Section L of my paper and it did very little to reduce our
discrepancy. But you might not agree on how I did the calculation and
suggest an alternative method, or you can suggest some alternative
prescription for making up the data, perhaps utilizing random numbers or
anything else you can think of that will allow me to do calculations. Or
you can just present me with tables of numbers that you consider to be not
implausible. 
	I offer a $2500 award to anyone who submits a suggestion that,
after a detailed evaluation, leads to a not-implausible explanation of our
discrepancy. I can give up to three such awards. If the submitter and I do
not agree on plausibility, I would be happy to accept the public judgement
of any prominent radiation health scientist suggested by the submitter
(letUs define prominent as 10 papers in HEALTH PHYSICS or equivalent
journals over the past 10 years). I would hope to publish a paper on this
with the submitter and judge as coauthors, but in any case, the $2500
award will be paid promptly.
	Of course the urban-rural effect discussed above was meant only as
an example; any other ideas would be equally acceptable. Alternative
suggestions for implementing my offer would be most welcome. I really need
help on this problem.
	If anyone would like a copy of our data file, I would be happy to
provide it.

Bernard L. Cohen
Physics Dept.
University of Pittsburgh
Pittsburgh, PA 15260
Tel: (412)624-9245
Fax: (412)624-9163
e-mail: blc+@pitt.edu