Re: Is radon as dangerous as the EPA says?

["John Moulder" <jmoulder@post.its.mcw.edu>]

> The EPA states categorically that 20,000 deaths per year from lung 
> cancer in the USA are caused by indoor radon.

The most recent EPA doc I have (EPA 400-R-92-011, May 1992) says 14000/yr with 
an uncertainty of 7000/yr to 30000/yr.  An earlier EPA doc says 20000/yr with 
an uncertainty of 5000-40000/yr.

> So - their model 
> predicts that number of deaths. Their model doesn't predict "no 
> increase in lung cancer should be detectable for the radon levels to 
> which people are exposed in the vast majority of residences." If you 
> could give me a reference for that quoted statement, I would appreciate

To know whether the projected risks are detectable, you have to look at the 
relative risk, not the total deaths.  From a different EPA doc (402-K92-001, 
May 1992), the EPA shows the following life time radon-attributed lung cancer 
risks:
Radon level     Non-smokers      Smokers
<0.4 pCi/l      <1/1000          3/1000
2 pCi/l          1/1000         15/1000
4 pCi/l          2/1000         29/1000
8 pCi/l          3/1000         57/1000

Now lets design a perfect case control study, and assume:
1)  Life-time radon exposure level is known
2)  Smoking history is known and "former" smokers are excluded
3)  You can get a control group with life time radon exposures below 0.4 pCi/l
4)  Non-smokers in the control group have a 0.2/1000 lifetime risk of getting 
lung cancer
5)  Smokers in the control group have a 141/1000 (male) or 56/1000 (female) 
lifetime getting lung cancer

Now you can do a power calculation and ask how many cases and controls you 
would need (I did the calculation assuming equal numbers of cases and control) 
to detect an elevated relative risk with a confidence interval that excluded 
1.0

For nonsmokers you would need about 4000 cases and controls at 2 pCi/l, 2250 
at 4 pCi/l and 1450 at 8 pCi/l.

For male smokers you would need about 7000 cases and controls at 2 pCi/l, 1500 
at 4 pCi/l and 400 at 8 pCi/l.

For female smokers you would need about 3500 cases and controls at 2 pCi/l, 
750 at 4 pCi/l and 250 at 8 pCi/l.
 
Any violation of assumptions 1 and 2 causes these numbers to escalate.  You 
could do the power calculation for various violations, but I have not.  

If you use EPAs upper level risk estimate (30000/yr) the numbers get a bit 
better.  If you use the lower level risk estimate (7000/yr) the numbers get a 
lot worse.

I would argue that the above case control studies cannot actually be done, and 
hence that the predicted increases are "nondetectable".  The studies cannot be 
done, because the assumptions cannot be met and because sufficiently large 
numbers of life-time high exposure cases could not be found.  The fact theat 
no one would pay for it is secondary.

> With respect to your definition of "health hazard" 
> number 3 -- It is impossible to demonstrate that any agent is 
> absolutely safe. Those who are in camp number 3 are not living in the 
> real world (even though they vote). We should do everything in our 
> power to tell the public that they can't have number 3.

No disagreement there.  

> As for number 2, the only body that can establish an acceptable risk 
> level for the US is Congress. Although EPA is driven by Congressional 
> Law, EPA has interpreted the law when it assigns a numeric value 
> to acceptable risk. Congress has never mandated a numerical value 
> of acceptable risk.

That I did not know.  So where does the EPA "acceptable risk" come from?

> Someone should challenge the EPA's numerical value 
> in the Supreme Court to see if it is constitutional. I'll bet the 
> Court would say "no." 

That would be interesting.  But as I understand the legal issues (caution: 
I've testified on cancer risks, but I'm a biologist, not a lawyer) this would 
require someone who has been "damaged" by a specific risk estimate.