RSH > Documents > RSH Comments on NCRP 136 > Bernie Cohen Comments
Comments on NCRP 136

by Dr. Bernard Cohen

Dr. Cohen is

COMMENTS ON NCRP REPORT 136

Bernard L. Cohen
University of Pittsburgh

    I am writing to convey some of my criticisms of your NCRP Report 136 on "Evaluation of the Linear-Nonthreshold  Dose-Response Model for Ionizing Radiation". One very important criticism of that Report is that it did not include as a "Research Need" in Sec. 9.5 a thorough study of health impacts on residents of the Taiwan apartment buildings who were exposed for many years to significant levels of radiation from Cobalt-60 in the structural steel. However, while I have many other reservations about NCRP-136, I will confine my discussion here to its dismissal of my study of lung cancer rates vs average radon exposure in U.S. counties [1].

At the outset, let me say that I do greatly appreciate the NCRP-136 statements that my "study was carefully and thoroughly done". But let me explain my objections - I will be as brief as possible here, but offer to provide an expanded explanation on request.

    As a general statement, I am puzzled by the fact that NCRP-136 references several papers that offer qualitative suggestions about how my study can give erroneous results, but in each case I have published papers showing by quantitative calculations that these suggestions are not productive, and my papers are not referenced. That is surely a violation of standard scientific procedures.

     I next list three specific issues on which I object to NCRP-136.

1. NCRP-136 implies that my study suffers from the same problems as other ecologic studies

    It references several general papers about weaknesses of ecologic studies -all published before my papers -and broadly implies that they apply to my work. While I agree with all of those papers, I vehemently deny the latter implication. I have offered to show how any other published ecologic study can give misleading results, but I have been unable, after years of effort, to show that for my work, and I have been able to show that all suggestions from others have similarly failed.     One major difference (not noted in NCRP-136) is that my study was not designed to determine the dose-response relationship between lung cancer and radon exposure, but rather as a test of the linear-nonthreshold theory (LNT). This point is obvious from the titles of my papers, and the difference is clearly emphasized in them. It is important for several reasons, including avoiding what has historically been called "the ecologic fallacy", and avoiding the false assumption that my data, reproduced in Fig. 9.6 of NCRP-136, can be interpreted as a dose-response curve for individuals -- although even with that interpretation, there is no significant disagreement with other data [2].

 
 

2. NCRP-136 claims that my study did not exert adequate control for smoking.

    In Sec. 9.2.2 it states that percent of population that smokes is "seriously inadequate to control confounding." But in the BEIR-IV theory I was testing, it is the only thing that matters. BEIR-IV considers smokers and non-smokers as different species of humans, with different lung cancer risks from radon; each "species" then makes a separate and independent contribution to the lung cancer statistics for a county, and all that is needed in combining them to obtain total county statistics is the fraction of the population belonging to each "species".     Recently, BEIR-VI suggested that a possible problem was a strong negative correlation between radon exposure and intensity of smoking, and proposed a model for testing that suggestion. I thoroughly performed that test and even extended it to include a strong  negative correlation between radon exposure and percent of the population that smokes; I found that no combination of these two correlations that is not completely implausible can explain more than a small fraction of the discrepancy between my data and the predictions of LNT [3]. This, incidently, shows that the statement on page 176 of NCRP-136 that my results might "be caused by small correlations between radon levels and smoking" is wrong. (I had previously shown why the referenced origin of that NCRP-136 statement was misguided.)

    NCRP-136 on page 176 faults my study because I did not have county-by-county data on smoking prevalence, and claims that the "surrogate variables [I used] probably have only a modest positive correlation" with the true smoking prevalence. However, in my original paper I showed that even a perfect negative correlation in this situation would eliminate only half of the discrepancy with LNT predictions. Moreover, an analogous study was done on a state-by-state basis, for which there are reliable data on smoking prevalence, and the discrepancy was even larger [4].     My treatments of smoking have been extensive and varied. A very brief summary of my seven largely independent approaches up to 1998, only one of which is recognized in NCRP-136, is given in Reference [5].

3. NCRP-136 claims that my study did not exert adequate control for age distribution in the populations of counties

    NCRP-136 states on pages 176 and 199 that potential confounding by age is an important issue in my study. For any confounding factor to influence the results, it must have a strong correlation both with lung cancer rates and with radon exposures. The bulk of the problem is handled by my using age-adjusted lung cancer rates, but I treated as potential confounding factors the percentage of the population in various age ranges, 35 age ranges in all. For none of these was there a significant correlation with either radon exposure or with lung cancer rates. I challenge anyone to propose a not completely implausible model by which age distribution can possibly explain the discrepancy between my data and LNT predictions in the light of this finding.

    I can well appreciate that the Committee that prepared NCRP-136 did not have time to carefully examine and evaluate my work. However, it is very important that it not be dismissed without such careful examination. I therefore offer to personally finance an NCRP study to settle the matter with some finality. If you will give me an estimate of the cost and time frame for such a study, I will provide the necessary funds.

REFERENCES:

1.  Health Physics 68:157-174;1995
2.  Health Physics 76:437-439;1999
3.  Health Physics 78:522-527;2000
4.  Environmental Research 64: 65-89;1994
5.  Health Physics 75:23-28;1998

 
 

RSH > Documents > RSH Comments on NCRP 136 > Bernard Cohen Comments

For more information please contact the RSH President Jim Muckerheide

For website problems please contact the Webmaster 		 

Google Scholar

07/13/05