Re: Radon and lung cancer

[John Goldsmith <gjohn@bgumail.bgu.ac.il>]

On Wed, 4 Mar 1998, Bernard L Cohen wrote:

> 
> 
> On Tue, 3 Mar 1998, David Scherer wrote:
> 
> > Dr. Cohen said, in part:
> > 
> > >        -It is a basic principle of "The Scientific Method" that if data
> > >exist that cannot be plausibly explained by the theory, the theory fails.
> > >If you don't agree with this statement, please say so, and I will expound
> > >further.
> > 
> > The Scientific Method does not require that every result be embraced, only
> > those results that have been replicated and confirmed with several
> > techniques.  There are lots of anomalous results in science that are
> > neither refuted nor embraced.
> 
> 	--They don't have to be refuted; they only have to be explainable
> in some way that is not implausible. In most cases, this would be
> experimental error.
> 	In our case, the same experimental errors would have to afflict
> our data, the EPA data, and the data collected by various state agencies.
> Moreover, the experimental errors in counties would have to have an
> extremely strong negative correlation with radon levels--i.e.counties with
> high measured radon levels would have to actually have low radon levels,
> and vice versa, and this would have to apply in each section of the
> country, etc.  
	The data are not experimental data.  They are basically
measurements of radon in dwellings, mostly at the request of the
residents, and the rates of lung cancer for the entire population of the
county.  If the sample of county residents with radon measurements is
biassed,( and it is admitted to be) then it is possible to get radon data
which is biassed with respect to the exposures of the county residents.  
	The original sample is biassed in that residents of more suburban 
(lower density) portions are more likely to get their radon measured.  
	As we have shown elsewhere, there is
a steep gradient of lung cancer with population density.  The "EPA" sample
and "state" samples are probably less biassed in terms of selection of 
the study subjects.  Data by states with high radon do not show the
consistent negative association.  When my time allows, I will submit a
letter to the editor, with a copy to Prof.Cohen.
       I feel that my epidemiological colleagues have been remiss in not
raising the problem of differential sampling of radon households and
cancer households.  The debate over "ecological fallacy" is over another
matter, and in my opinion a less critical one.
       I do strongly concur in the need for more training in epidemiology
of radiation and health problems, and its logical use.
       Because this is not an experiment in physics, much of what follows,
although valid is not germane.
                     

> 
> >I will give one example from physics, since
> > Dr. Cohen has worked in that arena with some distinction.
> > 
> > In the 70's Bill Fairbank (Stanford) detected fractional charges on small
> > superconducting niobium spheres.  To my knowledge, no one has ever
> > explained his results, yet physicists universally hold to charge
> > quantization and absolute confinement of quarks.  I would ask Dr. Cohen
> > whether he disagrees with this statement and whether The Scientific Method
> > has been abridged.
> 
> 	--Fairbank's observation was a single event (the plurals in the
> above are not correct) recorded by an automatic recorder with no one
> present. It was taken very seriously by the scientific community with
> great efforts to reproduce another such event ending in failure. There was
> a very large  body of other evidence that a single quark is not
> observable. Thus the observation was charged to experimental error.
> 	In my study, the result is duplicated with our data, with EPA
> data, and with data collected by state agencies. In the compilation, it is
> duplicated for each region of the country individually,(and for United
> Kingdom). Actually, our study of 1600 counties can be considered as 10
> studies of 160 counties each, with the choice of counties in each of these
> made randomly or by any of hundreds of specifications (e.g. only the most
> urban counties, only the most rural counties, only the richest counties,
> etc, etc.) Always the same result is obtained.

         If there is a strong  association between high population density
and lung cancer and a strong association of radon with low population
density, then stratification of the sorts, Prof. Cohen has done could
still show a spurious negative association of radon levels with lung
cancer.   


> 	Another major difference is that there is no other evidence
> supporting linear-no threshold theory in the dose region of my study. In
> fact there is a substantial body of other evidence suggesting that the
> theory fails, as was found in my study. 
> 	
> 
>   I would also ask for an example (just one) where any
> > accepted physical law that has been changed on the basis on one,
> > unconfirmed study.
> 
> 	--Up to 1952, it was assumed that a neutron striking a nucleus was
> certain to initiate a nuclear reaction; there were excellent reasons for
> believing this. The experimental finding of giant resonances in neutron
> cross sections changed that to "the cloudy crystal ball" model, now better
> known as the 'optical model".
> 	I can give any number of other examples.
> 
>   If The Scientific Method is at stake, show us the history.
> > 
> 	--I know of no example in physics where a theory survives after it
> is shown to be in disagreement with experimental results. If there is no
> experimental data supporting such a theory, it would be abandoned at the
> first hint of disagreement with experimental data.
> 	Let me quote Richard Feynman on this point:
> 
> "...we look for a new law by the following
> process: first we guess at it. Then we compute the consequences of the
> guess to see what would be implied if this law we guessed is right. Then
> we compare the result of the computation with ......observation, to see if
> it works. If it disagrees with experiment [the law] is wrong. In that
> simple statement is the key to science. It does not make any difference
> how beautiful your guess is. It does not make any difference how smart you
> are, who made the guess, or what his name is--- if it disagrees with
> experiment it is wrong. That is all there is to it."
> 
> 	In our case, the law that "was guessed at" is linear-no threshold 
> for individuals in the low dose region,
> and the consequences that were computed was the relationship between lung
> cancer rates in counties and average radon exposures in those counties,
> with suitable corrections for confounding factors. "Our discrepancy" is
> that the computed relationship is very different from the observed
> relationship.
> 
    In my opinion, the scientific effort needs to focus on the data and
its consistent analysis.  I believe that the linear non-threshold question
is, in respect to this matter, of secondary relevance.  Not that it is 
not an important issue, but it is not likely to be resolved on the basis
of data such as Prof. Cohen has so laboriously assembled and so generously
distributed.
                John Goldsmith, M.D., M.P.H.
      Professor of Epidemiology, gjohn@bgumail.bgu.ac.il 

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