Re: If you do Science, use the Scientific Method!

[Kai Kaletsch <eic@shaw.ca>]

Peter Thomas wrote:



>  It is a simple fact that the risk to an individual cannot be determined in such a study. 



The data seems to indicate that the risk of lung cancer TO AN INDIVIDUAL is negatively related to average county radon levels. The lower the average county radon level, the higher the risk of lung cancer for INDIVIDUALS living in the county.



What can be debated is if this is due to people's cumulative exposure to radon, some other radon exposure related parameter (such as a few short very high exposures that stimulate the immune system), something not related to radon at all (confounding from smoking, age, altitude, terrestrial gamma etc.), .... . The risk to an individual living in a low radon county seems to be quite high, whatever the cause might be.



Unfortunately, I am not aware of any case control study that was designed to investigate the risk of lung cancer TO AN INDIVIDUAL as a function of AVERAGE county radon levels. (This could eliminate some of the more obvious confounders.) Surely, the technology exists to investigate the risk to an individual posed by a group level factor. All the case control studies that I am familiar with try to investigate the relation between risk and people's cumulative exposure to radon, which could be completely irrelevant to the ecologic finding.



Kai



----- Original Message ----- 

  From: peter.thomas@health.gov.au 

  To: Fritz A. Seiler 

  Cc: RISKANAL Mailinglist ; RADSAFE Mailinglist ; Joseph L. Alvarez 

  Sent: Friday, September 19, 2003 1:56 AM

  Subject: Re: If you do Science, use the Scientific Method!







  I probably shouldn't be doing this but ... 



  Fritz Seiler wrote ... 





  [snip] 



   The discussion on RADSAFE, on the other hand, is an endless series of

  variations on the Radon/Lung cancer theme.  I must say that I am slowly

  getting tired of a scientifically totally aimless technical debate about

  epidemiology in the presence of confounders etc.  I have now several times

  asked the people involved to state clearly just what they are really doing

  in a scientific sense. 



    So here I go again, and to make it easier I propose

  the following alternatives 1) The aim is to determine the risk of lung

  cancer for Americans that are exposed to various levels of Radon and its

  daughters, taking into account that a third of these people smoke at various

  levels; or 2) The aim is to determine the purely academic risk coefficient

  for lung cancer due to an exposure to Radon and its daughters alone.  I do

  not believe that all of the members of the discussions have stated clearly

  what the scientific goal of their study is.



  [snip] 

  I think that nearly everybody (and especially Bernie Cohen) would agree that 

  Cohen's ecological study is NOT about determining the risk of lung cancer 

  due to radon.  It is a simple fact that the risk to an individual cannot be 

  determined in such a study. 



  What I understand Bernie to be saying he is doing in the study is to test the 

  Linear No-Threshold (LNT) hypothesis by assuming its truth (which allows 

  the use of aggregate data by virtue of its linearity) and then demonstrating 

  that this doesn't reflect reality. 



  Lubin, Field and others (in my understanding) maintain that Cohen's result 

  is the product of "cross-level bias" (I admit I don't really understand what 

  that is, I am not trained in epidemiology and have only been trying to 

  make sense of the debate as a part-time out-of-hours exercise in the recent 

  past) and confounding factors. 



  Bernie appears to me to maintain that he performs extensive analysis 

  of confounding and that he can find no plausible confounding to explain 

  his result.  My reading of Lubin's paper (J. Radiol. Prot. 22, 141 2002) 

  is that he thinks only a very weak correlation between smoking and 

  radon (as one example) is required to explain the result.  Hence the 

  endless arguement over confounding etc. 



  Finally, Field and others maintain that among the better ways to determine 

  the risk of lung cancer due to low radon levels for individuals is via a case-control 

  study and the results of such studies (on the whole) show a positive (i.e. 

  cancer causing) effect that is broadly consistent with a downward 

  extrapolation of studies of miners exposed to high radon levels (albeit 

  with error bars that include no effect). 



  My A$0.02 (which rounds up to A$0.05 unless transacted electronically) 



  Peter Thomas 

  Medical Physics Section 

  ARPANSA 



  The ideas expressed above are my own and do not necessarily reflect 

  those of ARPANSA etc.