<<Disclaimer: Verify this information before applying it to your situation.>>

I wanted to comment on an exceptionally detailed post that appeared a
few issues ago regarding twin studies.


> The occurrence of CD in twins will be important to the understanding
> of any
> role commensal microflora play in the onset of CD.  Ideally, it is
> easiest
> to study the effects of specific commensal bacteria on the gut and gut
> development through the use of specially bred germ-free (gnotobiotic)
> laboratory animals.  This allows researchers to study the effects of
> only
> one or a few microbes at a time.  There are typically some 500
> varieties of
> bacteria colonizing the human gut which makes studying the effects of
> specific microflora difficult.  Since no gnotobiotic humans are
> available,
> comparisons of the microflora in identical (monozygotic) twins, where
> one
> has CD and the other does not, provides one approach to identifying
> microbes which may play a role in the onset of CD.

A comment here not to argue but to illuminate the complexity.  While
400 to 500 species can colonize a typical person, each person has a
complement of about 250 species.  The actual mix for each person is
different.  While the bactria that colonize are thought to be largely
dependent on cell surface factors that promote adhesion of one bacteria
or another I suspect there is a random component to the selection as
well.  I come across patients who are not ill, but have travelled to
another part of the world and ever since then the bowels have changed.
I suspect a shift in the flora of some of those people.

One problem with the approach is that 90% of the bacteria cannot be
grown in the lab.  The research was started by NASA during the Apollo
program, figuring if you were going to detect a germ coming back from
the moon you needed an inventory of what we were brining there.  Those
studies provided the estimates of 500 species total possible with 250
species per person.  The identification of the flora was 90% based on
the microscopic shape of the germs rather than growth in pure culture.

> Twin CD studies have been pretty small.  Even this largest study has
> only
> 47 pairs.  How much confidence can be placed in such small sample
> studies?

47 is huge!  Remember these are twins.  The usual reason to study
groups is to average out the genetic influences.  With a twin study the
genetic differences don't exist within the pairs for identical twins.

> In this study it was concluded the risk of being concordant for celiac
> disease estimated for the non-index twin of monozygotic pairs was 17
> (95%
> confidence interval 2.1-134).  If I understand this correctly, that
> means
> that the risk of having CD for one identical twin if the other twin
> has CD
> is 17 times more than the risk rate for CD in the general population
> (1/133
> in the USA) or 17/133.  What is disturbing is that the 95% confidence
> interval ranges from 2.1/133 to 134/133, or a risk rate anywhere
> between 2%
> to 100% or so.  That doesn't seem very confident to me.

It is most likely that the rate is 17 times higher.  The confidence
interval says that if you did the study 100 times over, with different
twins selected every time so no study had the same people in it that
every time you did the study the results would be in the range of 2.1
to 134.  Now, given a bell shaped curve, there will be ver few studies
where the result was less than 5 fold risk, and very few where there
would be over 50 fold risk.  The real utility of the confidence
interval is that the number 1 is not in the interval.  If it were, the
study would not have reached statistical significance.  Kudo's to the
researchers that they picked a number of twins to study that they got a
stastically significant result, but didn't waste a lot of effort
getting a really narrow confidence interval.

But we haven't addressed that interesting stool question.  Twin
concordance studies are not the issue with the question of flora
contribution.  What you want is a study of the colonic bacteria in
identical twins discordant for celiac disease.  The really insightful
question raised here was not just whether twins show Celiac has a
genetic contribution, but what are the factors that allow identical
twins to be discordant when there is such a strong genetic influence?

Stephen Holland

*Support summarization of posts, reply to the SENDER not the CELIAC List*