Re: Mechanisms are Needed to Explain Cohen's Data

[Kai Kaletsch <info@eic.nu>]

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

From: "BERNARD L COHEN" <blc+@pitt.edu>

To: "Kai Kaletsch" <info@eic.nu>

Cc: "RadSafe" <radsafe@list.vanderbilt.edu>

Sent: Friday, January 11, 2002 8:40 AM

Subject: Re: Mechanisms are Needed to Explain Cohen's Data





> --This reopens the old question of whether we should measure radon

> gas or radon progeny in homes. There was a general consensus that it

> should be radon gas because factors that influence radon progeny also

> influence unattached fraction in a way that compensates, so we should not

> neasure radon progeny unless we also measure unattached fraction and that

> would be a very difficult and expensive proposition.



That leaves you open to a possible cross confounding effect with smoking.



> Note that your suggestion would also affect every other study that

> has ever been done on radon vs lung cancer, including all the case-control

> studies.



I am not trying to explain case control studies of radon in homes. There is

nothing to explain. The error bars are too big and the results are too

random at the very low levels that we are interested in. Your data is strong

enough to deserve an explanation.



(As for the old miner studies, I think the dosimetry there has bigger

problems than a small correlation between dose and smoking, even if some of

the measurements were for radon gas.)



> --Are you suggesting that a treatment is needed for non-smokers

> living in houses where there are smokers? That would be very difficult in

> any study.



Actually, I was looking for some reassurance (perhaps by a mathematical

exercise) that this could not bias your results by much.



> In case-control studies, they don't ask whether there are other

> people in the house who smoked.



They don't have to. They actually measure the radon concentration of the

house and assign it to the occupants. You are trying to use the total amount

of radon in the county and divide it up between smokers and non-smokers. I

am saying that your formula for dividing the dose may be wrong. Your table 4

might address my mechanism by varying a correlation coefficient between x

and r. I'm not sure.



> > I meant the use of filtration units that is causally related to smoking.

I

> > think in most cases the combination of smoking and filtration will

result in

> > a higher number of attachment sites than the combination of not smoking

and

> > no filtration.

>

> --This depends on the type of filtration. I have tested filtration

> and precipitation instruments that drastically reduce the number of

> attachment sites.



Even a 100 % efficient filtration unit running 100% of the time can only

cause an exponential decay in the number of attachment sites in the room,

because the cleaned air is immediately mixed with the old air. The smokers

that I know only turn on their filters once you can no longer see your hand

in front of your face.



> > I don't see how one can judge the plausibility of the correlation, if

one

> > does not consider the underlying mechanism. For example, you deal with

> > migration in your papers and conclude that it cannot explain the

observed

> > radon - lung cancer relationship. Would that conclusion still hold if

the

> > following mechanism was to be confirmed?:

> >

> > High radon in the interior of the US causes precursor to lung cancer

> > conditions in people . These precursor conditions make the dust, pollen

and

> > cold air of the interior unbearable and these people move to the coast,

> > where radon levels are low. There the precursor conditions develop into

lung

> > cancer.

>

> --I think most people would consider this to be very highly

> implausible



That was my point: plausibility of the correlation depends on the

plausibility of the mechanism.



> > I would be glad to write a letter to the Editor saying that these 20 or

30

> > mechanisms must be disposed.

>

> --That would be fine with me but they should be problems that do

> not apply to case-control studies. Also, I could not respond to 20-30

> mechanisms, so I would much prefer if you could specify a few that are

> most important in your list.



I don't have a list. My original post was intended to challenge the people

who want to save LNT to come up with specific mechanisms for cross level

biases rather that use generic arguments like "Cohen's effect is caused by

cross level biases". I gave a few possible examples to start them thinking.



Best Regards,

Kai Kaletsch





************************************************************************

You are currently subscribed to the Radsafe mailing list. To unsubscribe,

send an e-mail to Majordomo@list.vanderbilt.edu  Put the text "unsubscribe

radsafe" (no quote marks) in the body of the e-mail, with no subject line. You can view the Radsafe archives at http://www.vanderbilt.edu/radsafe/