Well,
Here is an story that you may find of interest.
In the U.S. Navy, you had to take a special medical physical to qualify for work associated with nuclear reactor duty and repair. (By the way, there are several physical requirements for such things as submarine and flight duty in the militray.) Part of the physical included a medical history that asked typical questions such as a personal or familly history of cancer. The medical officer was required to review and comment on all findings. Generally, a familty history of breast or lung cancer was not considered a disqualifying factor. If there was a questionable comment, a copy of the medical exam was sent to Washington for review and determination. This practice, as far as I know, is still in effects, and I believe that it applies to civilian shipyard workers.
When I was in the Navy, we had a sailor who had a white cell count that was just above normal. Reviewing his records for the prevous eight or so years, his white-cell counts were always just above normal, except for the time he had his first "radiation physical," when it was just below the high end of normal. It seened obvious that he was one of those who fell outside of the 95% "normal" values. Needless to say, he was removed from radiation repair work.
With regard to the NSWS, I am beginning to believe that the problem is in comparing the NNW and the NW. The Navy had a big asbestos problem that was ignored for many years. Consider the the older ships and non-nuclear submarines probably had more asbestos in them. The newer ships, nuclear submarines and surface vessels, probably little or no asbestos in them. My guess is that assignment to new vs. old ship repair may have also been a confounding factor.
Kai Kaletsch <eic@shaw.ca> wrote:
Chris,
I think you raise a very important point. If we can identify a mechanism by
which sick people are actively moved out of one group and into another, the
results of a comparison will be difficult to interpret. You raise the
possibility that people predisposed to cancer would have been taken out of
the nuclear cohort and moved to the non-nuclear one by the medical screening
process. I was hoping that someone more familiar with the study would
comment on this.
Is it really possible that someone could "fail" the medical screening
because of family history? I am familiar with two types of occupational
screening: You could fail the physical if you were not fit enough for the
work (nothing to do with radiation). The other type of screening was
intended for early detection of cancers and had nothing to do with job
placement.
Of course, it ! could also be that people with a family history of cancer
might be more worried about it and refused to work in the nuclear yards. It
is probably not possible to control for everything and no one study can
provide all the answers. I think the shipyard worker study is one more
reasonably strong data point.
It should also be pointed out that moving sick people from one group into
another can be an issue in all types of epidemiological studies, whether
they are ecological, cohort or case control. For example, if lung cancer
changes lifestyle, so that one keeps doors and windows shut just a few
percent more (too small a change to be measurable on a questionnaire) and
this raises radon levels in the living areas, then you move sick people from
the low radon group to the high radon group. I have yet to see anyone
quantitatively address that problem.
Overall, I think we have a tendency to throw data into the data analysis
technology t! oo quickly, without examining possible cause and effect
mechanisms first. The solution to the seemingly contradictory data sets will
not be found by calculating yet another correlation coefficient or p value
or by talking about non-identifiable cross confounders.