Eric:
The issue for me is not really whether BEIR-VII disparaged findings at variance with their view. Even their sleight of hand dismissal of findings from ecologic studies or in favour of hormesis is not disquieting me most. It is the flippancy they are treating those data with which superficially back their LNT postulate.
If you want to assess by yourself the scrutiny which the BEIR VII committee applied to peer reviewed data (attention: I mean data and not fit parameters) just pick two of the publications which figure very prominently (of course 'positively') in their argument.
1) Preston D L et al., Radiation effects on breast cancer risk: a pooled analysis of eight cohorts. Radiat. Res. 158(2002)220-225
and
2) Cardis E et al., Effects of low doses and low dose rates of external radiation: Cancer mortality among nuclear workers in three countries. Radiat. Res. 142(1995)117-132
In 1) go to table 4 (pp.224-5) and pool the rows HMG and HMS - as the ICRP did which in turn is quoted by BEIR-VII, e.g., Figure 1-7, SwHaem. Then draw in a linear-linear plot of this pooled breast cancer incidence rate together with the confidence limits in units of (10^5 PY)^ 1 against organ dose in mGy.
Since we are in the radiation protection business we are interested in (chronic) exposures below 1000 mGy.
Then ask yourself who in his right mind would represent these data by a straight line through the origin with a positive slope in order to estimate the risk in this dose range from these data(!).
If you wish to learn more about BEIR-VII scrutiny; plot the SwHaem rows for the whole dose range in a log-log plot.
What do you (in contrast to BEIR VII) see?
And then ponder: Breast cancer is supposed to belong to the more easily induced radiogenic cancers; babies are considered to represent the most radiation sensitive age.
These data represent the only truly chronic medical exposure analysed so far.
If you are not yet satisfied you might wish to look at the earlier original publications of the haemangioma data (Lundell M et al., Breast cancer risk after radiotherapy in infancy: ... Radiat. Res. 151(1999)626-632.
On page 628 you will read "The statistically significant ERR was driven mainly by cases with doses >4.0 Gy."
In Lundell M et al., Breast cancer after radiotherapy for skin hemagioma in infancy. Radiat. Res. 145(1996)225-230 you will read on page 229 "It was the contribution of subjects with breast doses >1 Gy that produced a positive association between dose and the subsequent breast cancer risk.".
BEIR-VII obviously did not want to see or know.
In 2) you pick from Table III (p.125) any row you like and this time plot the SMR given together with the 95% confidence limits.
If you don't want to waste your time you concentrate on all cancers, all leukemia, leukemia except CLL, colon cancer, stomach cancer, prostate cancer, and lung cancer.
After drawing these SMRs in a linear-linear plot ask yourself what these data(!) tell you about the cancer risk of chronic low dose exposures, say below 400 mGy, and in particular how these data compare to the BEIR-VII LNT estimate in this dose range.
Any student venturing to offer me such a misrepresentation of his data(!) for a master thesis - not to speak for a PhD thesis - without providing a compelling theoretical justification I surely would have sent back to square one.
Now search BEIR-VII for such a compelling theoretical justification of LNT.
You can repeat this exercise for ref. 1) and 2) virtually for any published epidemiological study for chronic/occupational exposures where these 'raw´ data are provided and you will find the same evidence.
Unless you are preoccupied with the notion that LNT must hold, you would never approximate these data with such a line.
Unfortunately editors increasingly are satisfied with statements of fit parameters instead of presentations of data (the new Cardis paper in BMJ is a very dire case in point).
If this trend continues, you soon will be unable to come to your own conclusions and you will have to rely on committees like BEIR-VII etc.
Upon request I can supply gif-files of the graphs whose construction I described above.
Regards, Rainer