AW: [ RadSafe ] Re: LNT/threshold/hormesis - what do HPs really

[Rainer.Facius at dlr.de]

John:

Here are data from another study for many cancers.

>From the data in the Cardis 1995 paper I attach three graphs giving their SMR data with the 95% confidence limits for all cancers, for all cancers except leukemia, and for all radiogenic leukemia. The data from this so far largest epidemiological study yielded as yet the most reliable empirical risk estimates for most tumours from occupational chronic radiation exposure.

Given that in no interval the SMR is significantly different from 1, the most appropriate form to represent these data is to lump them together. At least, that is what every textbook on statistics would advise. The pooled average SMR again with its 95% confidence limits in the given exposure range is indistinguishable from 1.

Finally, lines representing the various respective LNT postulates allow you to compare, which representation yields a more reliable risk estimate. When doing this comparison, you have to keep in mind that these risk coefficients are themselves associated with large - some people would say huge error bars - which are not reflected here. Depending on the number of factors considered to contribute to the uncertainty of the LNT risk coefficients, their resulting indeterminateness is surely in the order of 100%.

What degree of reliability would you assign to risk estimates derived from the LNT postulate in comparison to those derived from the data themselves?

Regards, Rainer

Dr. Rainer Facius
German Aerospace Center
Institute of Aerospace Medicine
Linder Hoehe
51147 Koeln
GERMANY
Voice: +49 2203 601 3147 or 3150
FAX:   +49 2203 61970

-----Ursprüngliche Nachricht-----
Von: John Jacobus [mailto:crispy_bird at yahoo.com] 
Gesendet: Donnerstag, 21. Juli 2005 23:28
An: Muckerheide, James; A Karam; Facius, Rainer; radsafe at radlab.nl
Betreff: RE: [ RadSafe ] Re: LNT/threshold/hormesis - what do HPs really

I think it is also fair to note that this is only based on one study.  However, if one looks at all the types of studies, e.g., thyroid cancers, lung cancers, incident rates, etc., what model fits all of the various types to some degree of reliability?  It is certainly possible that the LNT does not fit the results for one type of cancer.  How well does it do for ALL cancers collectively?

Again, the BEIR VII says that leukemia cancers are better fitted to a linear-quadratic.