Re: Definition of Epidemiology

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

On Fri, 28 Feb 1997, Daniel J Strom wrote:

> BLC =3D Bernard L. Cohen <blc+@pitt.edu>
> DJS =3D Daniel J. Strom <dj_strom@pnl.gov>
> 
> >>>> BLC: I am not sure that my work should be called "epidemiology."  =
> 
> It is a test of a specific theory. Tests of a theory are at the heart of =
> 
> "the scientific method" and apply rigorously to all of science. The =
> 
> baggage attached to ecological studies does not apply to, or affect such =
> 
> a test.
> 
> >>>> DJS: Definition of Epidemiology.  The study of the distribution and =
> 
> determinants of health-related states or events in specified =
> 
> populations, and the application of this study to control of health =
> 
> problems.  "Study" includes surveillance, observation, hypothesis =
> 
> testing, analytic research, and experiments.  "Distribution" refers to =
> 
> analysis by time, place, and classes of persons affected.  =
> 
> "Determinants" are all the physical, biological, social, cultural, and =
> 
> behavioral factors that influence health.  "Health-related states and =
> 
> events" include diseases, causes of death, behavior such as use of =
> 
> tobacco, reactions to preventive regimens, and provision and use of =
> 
> health services.  "Specified populations" are those with identifiable =
> 
> characteristics such as precisely defined numbers.  "Application to =
> 
> control..." makes explicit the aim of epidemiology - to promote, =
> 
> protect, and restore health (Last 1995).

	---I don't think the naming is worth arguing about. I was using
a more pragmatic definition of epidemiology as "what epidemiologists do".  

> 
> >>>> BLC: It is a test of a specific theory.  Tests of a theory are at =
> 
> the heart of "the scientific method" and apply rigorously to all of =
> 
> science.  The baggage attached to ecological studies does not apply to, =
> 
> or affect such a test.
> 
> >>>> DJS: Dr. Cohen has not convinced me or other critics that "the =
> 
> baggage attached to ecological studies" does not apply to a study for =
> 
> which there are no individual exposures, no individual confounder =
> 
> measures, and no individual bias measures associated with individual =
> 
> health outcomes.  =
> 
> 
> See Figure 3 of Lubin and Boice (1997) for an example of how Dr. Cohen's =
> 
> results are statistically rejected by more cogent studies.
> 
	---The following is my response to the Lubin-Boice paper, accepted
for publication by JNCI. Note that it makes the same error that Strom is
making here.

	I am writing in response to the scathing attacks in the January
1997 issue of JNCI by Lubin and Boice (1) and by Samet (2) on my paper
"Test of the Linear - No Threshold Theory of Radiation Carcinogenesis for
Inhaled Radon Decay Products" (3). 
	Their most flagrant error is in grossly misinterpreting the
purpose and design of my study which, as should be obvious from the title,
was to test the linear - no threshold theory (LNT) of radiation
carcinogenesis. A universally accepted and time-honored procedure for
testing a scientific  theory is to develop it mathematically to derive
predictions that can be compared with experimental observations, and then
make that comparison. In my paper it was shown, utilizing rigorous
mathematics, that LNT predicts that a plot of lung cancer rates in U.S.
Counties (corrected for smoking prevalence and other confounding factors)
vs. average radon exposure in those counties, r, should have a positive
slope of +0.20 per Bq/m3; it was shown that this prediction is not, in any
way, affected by "the ecological fallacy". It was then shown that the plot
of observational data has a negative slope, -0.21 (+/- 0.02) per Bq/m3, a
20 standard deviation discrepancy with the prediction of LNT. My paper
gives very extensive treatments of data uncertainties, of potential
confounding factors (including elaborate modelling), of weaknesses in
ecological studies, etc, and finally concludes that LNT fails very badly
in the low dose region of my test, and therefore, in accordance with "the
Scientific Method", LNT should not be used in that region.
	Lubin and Boice (1) and Samet (2) completely ignore my test of
LNT, which is what my paper is all about, and choose to interpret my plots
of county-average data as risks to individuals, despite my warnings
against such an interpretation because it is logically invalidated by "the
ecological fallacy". In fact Fig. 3 of the Lubin-Boice paper, a plot of
their data on relative risk (RR) vs. radon exposure to an indivdual, r,
shows what they call my results as a straight line extending to 340 Bq/m3
at which point there is a large discrepancy with their data. Actually, my
largest r data point (which is statistically insignificant) is at r=230
Bq/m3, and the linear portion of my data ends at about r=120 Bq/m3.
	Another problem with their interpretation of my results is that my
analysis determined the slope of my data points without regard for the
intercept at r=0, and hence the line Lubin and Boice show as my results
can be  shifted up or down slightly on their plot. If this were done,
there would be relatively little discrepancy with their data. To the
extent that any discrepancy remains, it is easily explainable by the
widely recognized weaknesses of the meta-analysis technique they use, and
by the effects of "the ecological fallacy" on their distorted
interpretation of my data. There is therefore no justification for their
criticism of my paper.
	REFERENCES
(1) Lubin JH, Boice JD.  Lung cancer risk from residential radon: meta
analysis of eight epidemiologic studies. J Natl Cancer Inst 1997; 
	89:49-57. 
(2) Samet JM. Indoor radon exposure and lung cancer: risky or not? -- all
over again. J Natl Cancer Inst 1997; 89:4-6
(3) Cohen BL. Test of the linear - no threshold theory of radiation
carcinogenesis for inhaled radon decay products. Health Phys
1995;68:157-174


> >>>> BLC: I have clearly shown that testing a linear- no threshold =
> 
> theory does not require data for individuals.  If Strom disagrees with =
> 
> these demonstrations, he should say specifically why.  I can very easily =
> 
> explain why data for individuals is needed to determine a dose-response =
> 
> relationship; why can't he explain why data for individuals is necessary =
> 
> to test an LNT theory?
> 
> >>>> DJS: The county radon measurements made long after many of the =
> 
> cancer deaths apply only weakly to the individuals who died of lung =
> 
> cancer in those counties.

	---This problem applies also to the case-control studies that
Strom  extolls. In my papers I do discuss this problem and give evidence
that it is much less serious in my study than in the case-control studies


>  Most importantly, many other causes of lung =
> 
> cancer (i.e., confounders) that may be associated with (covary with) =
> 
> geographic location, e.g., air pollution, different genetic make up, and =
> 
> lifestyle factors (especially smoking), cannot be controlled in any =
> 
> meaningful way despite Dr. Cohen's attempts and his claims.

	---There is no problem in the method of correcting for smoking in
my work because BEIR-IV treats smokers and non-smokers as two different
"species", each with separate and independent risks,  and I follow their
procedure. As for other confounding factors, I have very extensive
treatments in my paper and have defended them extensively there. I have
offered a $2500 award for a suggestion of a plausible  model of how some
specific confounder can explain our discrepancy consistently with LNT, but
no suggestions have been forthcoming. I offer to provide such models to
show that any other published ecological study can be misleading; doesn't
that show that my study is different? If Strom believes that some
confounder is not controlled for in "any meaningful way", surely he should
be able to come up with such a suggestion. If the $2500 award is not
sufficient incentiive, I hereby raise it to $5000. If money isn't an
incentive, I would think that publishing his model in a journal would be
an incentive, showing that a paper that has attracted so much attention
(I would be happy to document this characterization) is flawed. 

  Dr. Cohen's =
> 
> study is useful for generating the hypothesis that radon exposure is =
> 
> causally related to lung cancer (one way or the other).  That hypothesis =
> 
> has been, and is being, tested by more cogent epidemiologic methods.  =
> 
> These simple statements have been made repeatedly in print (see, e.g., =
> 
> Lubin and Boice 1997), but Dr. Cohen doesn't accept them.

	---My response to Lubin-Boice is given above. Can Strom cite
another case where my 1995 paper has been criticized in print? or can he
respond to my response to the Lubin-Boice paper? 

  His =
> 
> persistence doesn't change the truth of the simple statements.
> 
> See Lubin and Boice (1997) and Lubin et al. (1997) for good recent =
> 
> discussions.

	---Does Lubin et al contain a criticism of my paper? Please let me
know if it does so I can investigate what they have to say.


  Stay tuned for BEIR VI, due out this spring.  =

	---So  I have to wait for BEIR-IV to be published to see why they
object to my paper. Is that the way Science is done? Isn't it scientific
tradition to give the author of a paper being criticized an opportunity to
respond? I realize that this can't be done with every paper, but I believe
mine has attracted enough attention to warrant this treatment, not to
mention the fact that I have received high awards from three national
scientific societies. I have offered to come, at my own expense, to
discuss possible criticisms  with BEIR or any sub-group of it (including
single members), but have had no response.




> Woodward, A.; Yao, S.X.  Estimating Lung Cancer Mortality from =
> 
> Residential Radon Using Data for Low Exposures of Miners.  Radiation =
> 
> Research 147(2):126-134; 1997.
> 
> The opinions expressed above are my own, and have not been reviewed or =
> 
> approved by Battelle, the Pacific Northwest National Laboratory, or the =
> 
> U.S. Department of Energy.
> 
> Daniel J. Strom, Ph.D., CHP
> Staff Scientist
> Health Protection Department K3-56
> Pacific Northwest National Laboratory
> Battelle Boulevard, P.O. Box 999
> Richland, WA 99352-0999 USA
> (509) 375-2626
> (509) 375-2019 fax
> dj_strom@pnl.gov
>