String of Pearls

In my relatively short tenure as a blogger, I have posted a fair bit about autism and mercury. This is not necessarily by design - I have simply followed where the current led. During this time, I have had a number of people try to show me the error of my ways - that there is data supporting the hypothesis that mercury causes autism.

These people have been, in turn, collegial, insulting, threatening and accusing. A few have even been comical, although I don't think that was their intent. The majority of them (at least, those that tried to use reason) have presented a series of studies and other findings which they believe show that mercury causes autism.

What they are doing is known by a number of different names - the name I am most familiar with is data chaining. In mathematics, this is a solid theorum - if A = B and B = C, then it follows that A = C. In the biological sciences, this is not always so straighforward, but it can be deduced that if A causes B and B results in C, then A can cause C (although A may not be the only cause of B or C).

In biology, the use of data-chaining has led some investigators to make claims they later regretted, usually because they failed to consider the following caveats:

[1] Even if A causes B, it may not be the only cause.
[2] If A and B occur together, one is not necessarily the cause of the other.
[3] Biology is probabilistic, not stochastic. (In other words, organism vary, even within a sub-species)
[4] A biological organism - even the simplest - is a complex of interconnected systems that are only partially known.

The autism-mercury supporters tend to use data-chaining in a very cavalier way, as in the following simplistic example:

[1] Mercury is known to cause brain injury. true
[2] Autism is a brain disorder. also true
[3] Children are exposed to mercury in the form of thimerosal in vaccines. was true, is less true now
[4] Ergo, autism is caused by mercury.

Now, this is admittedly a very easy "chain" to break, but it is also a widely quoted one, so the example is valid. To "break" this chain, all you have to do is note the lack of solid connection between the data used. In this example, it is true that mercury causes brain injury, but the injuries documented over the past hundred years or so have not looked like autism. Some of the individual symptoms or signs may share a certain similarity to autism, but the presentations as a whole are vastly different.

By showing the lack of connection between the "facts" used in this data-chain, it becomes apparent that what we actually have is a series of separate "facts" with an implied connection, something that another writer called a "string of pearls". A "string of pearls" is a beautiful group of data without any visible connection.

Another string of pearls appears in Dr. Hornig's presentations to autism groups. She found that different strains of mice had different susceptibilities to thimerosal. Not suprisingly, a strain that was known to be susceptible to autoimmune disorders did poorly - although they did not specifically look for signs of autoimmune disorders. Curious.

This is her string of pearls - it's a bit more complex:

[1] Children get thimerosal-containing vaccines at ages 2, 4, 6 and 12 month.
they used to, not any more

[2] The developmentally-equivalent ages in mice are 7, 9, 11 and 15 days.
true for certain aspects of development - human infant brain development is much different from that of a mouse in many ways

[3] Giving infant mice thimerosal at ages 7, 9, 11, and 15 days is equivalent in timing to giving thimerosal to infant humans at 2, 4, 6, and 12 months.
only in timing - the metabolism and elimination of thimerosal has not been shown to be equivalent

[4] Humans are genetically diverse.

[5] The SJL/J, C57BL6/J, and BALB/cJ strains of mice are genetically different.
true, but not in ways that are necessarily equivalent to human genetic diversity

[6] Some autistic humans show decreased interest in their environment.

[7] The SJL/J (autoimmune disorder susceptible) mice treated with thimerosal showed decreased locomotion and exploration of their environment - this is similar to what is seen in autistic humans ([6]).
this is a pretty big assumption - a better explanation for the decreased locomotion and exploration is that the SJL/J mice were ill from autoimmune disorders that were not tested for

[8] Some autistic humans are averse to novel situations.

[9] Thimerosal-exposed SJL/J mice showed increased exploration in a novel environment.
it is hard to see how this can be seen as an analogue to [8], but that was the implication

[10] Self-injurious behavior is seen in some autistic humans.

[11] Some of the thimerosal-exposed mice repeatedly bit their tails.
this is also seen in partial denervation in mice and rats and may simply be a sign of mercury toxicity. Humans do not bite partially denervated limbs


In the final analysis, Dr. Hornig has found out that SJL/J strain mice - who are known to be susceptible to autoimmune disorders - also react poorly to mercury. The other mouse strains in this study did not. She also - and I know I'm being repetitive here - did not test the SJL/J mice for signs of autoimmune disorders. That could have explained many of the results.

Interestingly, it appears that the SJL/J strain of mice has a much lower ability to excrete mercury, which would then lead to a higher tissue level. It's a pity Dr. Hornig didn't get hair and blood samples from these mice - we would then have a chance to see if the string of pearls proposed by Holmes et al (and here) has any validity. Oh, well.


So, the next time somebody says that they have "incontrovertible proof" of something based on a study that shows A causes B and B is seen in cases of C, look for the string connecting the pearls. It may not be there.


Prometheus.