Re: Cohen's ecologic studies

[JMUCKERHEIDE@delphi.com]

> On Tue, 24 Feb 1998, R. William Field wrote:
> > 
> > Dr. Cohen,
> > 
> > Do you think your ecologic studies have the power to assess the lung cancer
> > risk/benefit posed by residential radon exposure?
> > 
> > Bill Field
> 
> My study is a test of the linear-no threshold theory (LNT) of radiation
> induced cancer. It shows that LNT fails very badly in the low dose region
> ((below 6 pCi/L), grossly over-estimating the cancer risk in that region.
> It does not determine the risk vs dose to individuals beyond the above
> statement. Interpreting our data directly as risk vs dose suffers from
> "the ecological fallacy", but the above-stated conclusion does not.
>
> Bernard L. Cohen

Hi Bernie,

It is not clear to me what this means; what is the data "suffering" from?

It seems that "the ecological fallacy" correctly reflects a potential weakness 
in results that may be introduced when there is a threshold (NOTE: the
argument that the "ecological fallacy" applies preordains the existance of a
threshold!)  In that case, the premise is that the dose data reflect a mean
that fails to reflect that the effects data only comes from the higher dose
population. It is easy to see how this works for a specific low-dose bin in a
typical simple ecological study. It is not clear how that affects considering
the results of your study when examining the actual data, including error
bars, eg, for counties at 0.5 pCi/l vs those at 1.5 pCi/l, vs 4 pCi/l, and 6
pCi/l. 

Further, while it is easy again to accept that the "ecological fallacy" does
not allow that these results be accepted as the definitive dose-response
relationship, it seems nothing more than 'politically correct' to deny that
the data show actual lower lung cancer in the large populations at 4-6 pCi/l
vs those at 0.5-1.5 pCi/l. How does the actual mathematical treatment of the
effects of the "ecological fallacy" deny that there is clear evidence of lower 
cancer considering the size of these populations and datasets? and considering 
the numerous completely independent analyses of subsets of these populations
such that multiple independent studies demonstrate confirmatory results? (How
many more than 20 independent studies are needed to conclude that the result
is not an anomaly?  :-) 

In addition, Drs. Gary Sandquist, Jay Kunze, Vern Rogers, and others took
EPA's state level radon data, and the American Cancer Society lung cancer
data, and showed that the lowest US radon area, the Pacific Northwest states,
had a lung cancer rate at 390% (4 times) the BEIR IV prediction, and the
highest US radon area, the Rocky Mountain states, had a lung cancer rate 14%
(1/7th) the BEIR prediction. Of course, the study of radon in France showed
the same, the women in the uranium mining areas of Germany showed the same,
and on, with small areas like the Misasa Spa area of Japan and Yangjiang
Province in China not showing statistical significance in lower lung cancer in 
smaller populations, but there is also no increase as would be predicted by
BEIR. 

[It should also be noted that the power of the statistics of even a small
ecological study (of a million+ people) is much more powerful than any human
case-control study. This is simply because the power of a case-control study
using small populations relies entirely on knowing the dose (in medical
experiments, usually thru injection or ingestion). In the case of radon
case-control studies, residential measurements are not directly related to
doses to individuals, since any individual in a "radon pool" will inhale a
different dose, like in what side of the bed she sleep on, or where the
washing machine is located, etc etc (eg, in the French uranium miners,
individual dose monitors did not correlate well with area monitors, with 2
persons in the same area getting different doses). Therefore residential
measurements are a level around which residents have a distribution of doses;
and the radon case-control study depends on averages in these data (perhaps
excepting, eg, cases like the women with lung cancer in the smaller, more
controlled residences/populations, with a full year of monitoring, in Shenyang 
China, with a negative correlation). But the populations are so small that,
unlike the ecological study, the averages have large errors. So the results of 
individual studies are all over the map. They become essentially useless data
vs the quality of the data in a good, and there aren't many, ecological study, 
like Cohen's.] 

Regards, Jim Muckerheide
muckerheide@mediaone.net
jmuckerheide@delphi.com
Radiation, Science, and Health