A recent (2003) study of effects of dental amalgams
http://www.ehponline.org/members/2003/5879/5879.htmlwhich found no harmful effects appears to have been
designed not to find any such effects.
Basically what they did is to perform a series
of neuropsychologic performance tests on people
with various numbers of amalgams.
They did find direct linear correlation in blood
and urine levels of mercury metabolites with
number of amalgams (or surfaces). Clearly the
amalgams do cause mercury absorption and show
linear dose/response relation.
Regarding the neuropsychologic tests, they cheated
by "stratifying" the subjects into categories such
as educational level, professional status,
socioeconomic status and numerous other "confounders"
(using unspecified adjustment). Then they compared
mental performance of subjects with and without
amalgams _within each subset_. The problem with
this approach is that after you adjust for all
"confounders" you get subset which by definition will
be have level of mental performance corresponding
to the parameters held fixed for the subgroup
(e.g. professional level, socioeconomic
status...). For example if amalgams (which were
all placed in teen years) have caused drop in IQ
of 30 points, a person who might have been 130
would become 100 and achieve professional status
of people with QI=100. Hence someone who might
have been a scientist (with IQ 130), may have
turned out due to amalgams an average person.
Similarly a person with pre-amalgam IQ 160, who
might have been Nobel laureate, would only become
a regular scientist corresponding to IQ 130.
Hence, by fixing great many "confounding variables"
within each group, the authors had in effect fixed
the equivalent of the mental performance within that
group. Comparing subjects within each subgroup will
then measure that same fixed level of performance
corresponding to that subgroup. The experiment was
in effect a tautology with guaranteed outcome. It
was completely insensitive to the effect on mercury
on shifting persons from one "stratum" to another.
To give analogy, consider we wish to measure effect
of childhood starvation on height of adults.
We pick 1000 adults, divide them into subgroups
based on weight and height. Then _within_ each subgroup
we compare how the remaining small height variations
correlate with the data on childhood starvation.
Obviously, if we're looking within the 6 ft +/- 10%
subgroup, we won't find out how tall a 6 ft person
with history of starvation would have been had he
not been starving in childhood.
The entire study was obviously rigged -- it was
devised to be entirely insensitive to the question
of amalgam effects on mental performance. The fact
that they had to cheat using such cheap statistical
sleight of hand means that the amalgams do cause
harm and any genuine study would show it.
