[ RadSafe ] Cameron's refutation of "Alara Does Work"
hflong at pacbell.net
Wed Jun 28 13:04:51 CDT 2006
Ranier, your careful review of the Boice commentary on Matanoski's presentation of the vast Nuclear Shipyard Study, confirms my belief that healthy worker effect cannot explain the 3 years of life apparently added by exposure of Gulf Coast workers to 0.5 rem extra.
The only credible (although minor) healthy worker selection I have seen proposed, was here on Radsafe. It was that a doctor giving employment exams might unconsciously have directed persons of less vigorous health away from imagined hazards of radiation exposure, to otherwise identical work.
Rainer.Facius at dlr.de wrote:
Thank you for providing the Boice Jr. (2001) paper commenting on Matanoski's Nuclear Shipyard Study. I took some time to compare his remarks with the original report and therefore my reply is somewhat late.
First of all my comparison leads me to conclude, that Boice delivered a very compressed but overall fair summary of Matanoski's report.
Secondly, his assessment of the high general quality of Matanoski's work and report is in line with Sponsler and Camerons's (2005) evaluation.
Boice does nowhere claim that the observed significantly lower total mortality of the >= 5 mSv nuclear as compared to the non-nuclear worker group is the result of bias.
He addresses some points of concern without calling into question the principal quality of the results.
Boice's concentration on leukaemia data - though understandable - is somewhat moot, given the small number of cases and indeed he does not draw any conclusions from discrepancies with his expectations.
Regarding race, Boice states "It appears that race was not considered in the analysis which could be an important confounder for certain cancers". It is true that Matanoski could not stratify their data with respect to race since the personnel files did not contain this information. It is not true that this confounder was not considered. Mantoniski et al. discuss in quite some length this problem and present their reasons to conclude that their findings would not be seriously affected.
An objection sometimes raised against the validity of Matanoski's study is the (apparent) absence of the healthy worker effect frequently seen in occupational studies. Boice also does state this absence without drawing the usual conclusion - most appropriately so.
Although the literature does differentiate a dozen or so 'healthy worker effects' the bulk of this selection bias can be attributed to two mechanisms, (i) the "healthy worker hire effect" and the "healthy worker survivor effect". The first bias excludes potential workers whose illness prevents them from seeking or finding employment. The corresponding biased SMRs ususally scatter around 0.90 (90%) and hardly ever fall below 0.80. Furthermore this selection bias dilutes within the first years of employment. The "healthy worker survivor effect" results from "differential job migration" either into another less demanding occupation or out of the workforce altogether due to illness occurring during employment.
Although the healthy worker hire effect is regularly invoked when radioepidemiological studies find massively lower cancer mortalities (and they usually do) to explain this discrepancy with expectations, this argument rarely exceeds the standing of an ad hoc explanation. So far I never have seen this 'explanation to be offered and then accompanied by an obviously mandatory discussion of which checks of employment medicals do have prognostic power to select against employers more likely to develop cancer than others.
Regarding the healthy worker survivor effect I have so far not found quantitative data as to its typical size regarding non cancer diseases, but here too such arguments - where needed to explain too low radiogenic cancer mortalities - do never accompany its invocation with a proper discussion of what amount of mortality reduction might reasonably be ascribed to it. Apparently, "anything goes". Furthermore, since cancer is predominantly a disease of advanced age, the age dependence of the healthy worker survivor effect should be addressed too. Finally, the design of some radioepidemiological studies - such as the studies on European aircrew - excluded its influence altogether - which did not prevent some of the investigators to invoke it nevertheless.
After all, the Matanoski study does exhibit the healthy worker effect for diseases were you would expect to see it, i.e., for mortality from diseases of circulatory system - the dominant single cause of mortality in the age of workforces. The diagram in the attached PDF-file (again available upon request if the moderator intercepts it) clearly demonstrates that the healthy worker effect is there as it properly should be. Regarding its absence from total cancer mortality, the diagram reveals enhanced mortality from cancers of the respiratory system as the cause. Matanoski's report as well as Boice's discussion appropriately invoke known workplace related agents which cause this enhanced cancer mortality which in turn masks the otherwise reduced mortality from other cancers.
Therefore, the objection - by others than Boice - of a missing healthy worker effect is misplaced and futile.
In summary, none of the limitations as he enumerates them make Boice to conclude that the observed reduced cancer mortality can be explained as arising from bias.
Boice's final 'overcautious' assessment "While the study is 'consistent' with no effect follow low-dose radiation exposures, the relatively small doses and narrow dose distribution limits interpretation." spells out the reason why personally I do hesitate to publicize the Matanoski data as a 'proof' for radiation hormesis.
Nevertheless, your discarding these data altogether as irrelevant is neither backed by statements from Boice's paper that you kindly provided to me nor by their limitations - as they were properly spelled out by Matanoski et al. in the first hand.
Kind regards, Rainer
Von: radsafe-bounces at radlab.nl im Auftrag von John Jacobus
Gesendet: Fr 21.04.2006 22:38
An: Facius, Rainer; hflong at pacbell.net; jjcohen at prodigy.net; mike.bohan at yale.edu; radsafe at radlab.nl
Betreff: Re: AW: AW: [ RadSafe ] Cameron's refutation of "Alara Does Work"
Here you go.
Have a good weekend.
--- Rainer.Facius at dlr.de wrote:
> Thank you John for your speedy reaction.
> Do you have a copy of this paper or do I have to
> order it through our library?
> Kind regards, Rainer
> Von: John Jacobus [mailto:crispy_bird at yahoo.com]
> Gesendet: Fr 21.04.2006 22:18
> An: Facius, Rainer; hflong at pacbell.net;
> jjcohen at prodigy.net; mike.bohan at yale.edu;
> radsafe at radlab.nl
> Betreff: Re: AW: [ RadSafe ] Cameron's refutation of
> "Alara Does Work"
> I would say that the Boice paper, J Radiol Prot.
> Dec; 21(4);400-3, detailing the limitations of the
> shipyard worker study should shed doubt on Dr.
> Cameron's assertions.
More information about the RadSafe