Wednesday, October 6, 2010

Jabberwocky Of The Day : Estimates Are Not Major Discoveries

Earlier this year, a paper was released entitled "Social Demographic Change And Autism" that attempted to explain the part of the rise in the rate of autism.  The authors arrived at three "major discoveries" that were supposed to change our thinking about autism.

While the ideas in the paper have some merit, I can't resist pointing out one of the flaws in one of the "major discoveries".  The discovery that I am talking about is this one -
First, the estimated heritability of autism has been dramatically overstated.
In very rough terms, this statement is basically saying that autism is not something that your parent's pass down to you but rather something that is caused by something else.  The paper goes into some specific about what this could be, but I am going to focus on this one point for now.

To prove this point, the researchers analyzed some data  -
To anticipate the main results of this article, we first demonstrate that autism heritability — defined in the narrow sense as the difference in concordance for autism between monozygotic (MZ) and dizygotic (DZ) twins is not as significant as typically believed.
Of course, it always comes back to twins.  Most other studies have shown that identical twins are more than twice as likely as fraternal twins to both have a form of autism.   But here, the researchers are basically saying that this is not true - that identical twins are not much more likely to both have autism than fraternal twins.

If this result were true, it would signal a a major change in how we think about autism.  But, as I was looking through the paper to see what numbers that the authors arrived at, I ran across this statement -
The administrative data we work with do not have a direct measure of zygosity, so we do not know from these data whether twins are MZ or DZ, which is central to the estimation of genetic influence.
Wait, what?

One of the major findings of the paper is based on the difference between identical and fraternal twins but the data in the study does not have the ability to tell you which twins are identical and which aren't?  So, how can any conclusion be reached about the differences between the two groups when you don't know what the groups are?
A simple and well-established rule that has been shown to give robust zygosity estimates has been developed for this purpose and has been widely used in research on twinning rates and a range of other research questions.
So one of the "major discoveries" of the paper - one that disagrees with most other twin studies out there - is not based on actual data but a guess of what the data might look like?

It is a good thing that the proportion of twins that are identical is a constant (oh wait, it isn't) and that factors influencing twin births weren't changing drastically during the study period (oops, they were) and that we know that identical twins aren't more likely than fraternal twins to have autism in the first place (no clue).

There are so many unknowns and possible other factors involved that without this one little, tiny yet fundamental piece of information - whether a set of twins were identical or fraternal - the conclusion based on that data becomes nothing more than a educated guess.

Conclusions should be based on data, not guesses.


  1. Hi MJ -

    I think you might be being a bit hard on Bearman. While the limitations you point out are real, they do have two big advantage; sample size and a randomized sample not subject to perils of study selection, i.e., 'We solicited for twins with at least one sibling with autism'.

    According to Bearman, they used five times the number of twin pairs with autism than our other studies combined. They don't list which papers they use as a platform on which to base this statement, unfortunately.

    Another finding that isn't subject to exact zygosity measurements was the finding of a very high concordance between opposite sex twins, who must be dizygotic.

    The casewise concordance rate for OS twins is 18.4%,
    fully 2.6 times higher than the recurrence risk in OS full siblings (7.2%). Because OS twins
    and OS full siblings have the same genetic relatedness, the difference in recurrence risk
    indicates that in addition to genetic factors and also perinatal and prenatal care, social and
    other environmental factors must contribute to autism.

    We don't have to worry about estimates here, but they still seem to observe some numbers that indicate something other than genes is participating.

    I posted some links on LBRB on this, but there have been other studies that suggested that just being part of a twinning process raised the risk of autism (other findings were negative). From a competition of nutrients or additional stress on the mother standpoint, this would seem to have a logical foundation.

    It is a good thing that the proportion of twins that are identical is a constant (oh wait, it isn't)

    Really? That's pretty interesting. Can you post a link to something about this? What is the proposed mechanism? IVF?

    factors influencing twin births weren't changing drastically during the study period (oops, they were)

    IVF again?

    that we know that identical twins aren't more likely than fraternal twins to have autism in the first place (no clue).

    I believe the previous studies on twins in autism are used as evidence for this point. I suppose you'd have to figure out how badly you think the problems Bearman raised are with those studies. I'm not sure. (?)

    - pD

  2. pD,

    I can write more about the twin aspect in the future, but the simple answer is that the the rate of MZ twins is pretty much a constant while the rate of DZ twinning has been growing rather rapidly (almost double form 1980 to 2006).

    Identical twins are basically 1 in 250 births in pretty much every race, world wide (with some few exceptions). DZ twins have been becoming more common every year, even over the small time frame of this study and especially in the US.

    I don't have the numbers in front of me right now, but when you look at the first year of the study, you would expect about 65% of same sex twins to be identical but by the end of the study, that number drops to about 50% because DZ twins had become that much more common.

    If you are interested, I can find the exact numbers again and post some links to the original sources.

    As for why this is the case, IVF is one reason, but there are others. But whatever the reason, DZ twins are become more common and the proportion of twins that are MZ are shrinking by comparison.

    The skewing from that change would not be small. The authors briefly address the change in the ratio between MZ and DZ in footnote 10 on page 12 of the study.

    This is all assuming that MZ twins don't have greater risk factors than DZ twins do, which is far from a given. MZ twin are more likely to have prenatal complications because of their typical sharing of resources in the womb. They are also more likely to be born premature than DZ twins, which has already been shown to be a risk factor for autism. MZ twins would also be more likely to be vulnerable to the same environmental factors where DZ twins might not be.

    There is also a significant bias in type of twin by age - older parents are more likely to have DZ twins than younger ones. Since one of the other main findings of the study had to do with parental age the type of twin becomes even more important to the findings.

    Regardless, for these reasons and more, I think it is far from a given that the proportion of MZ to DZ twins in this population is a constant and is the same as the general population. I think if you are going to base one of the conclusions of your study on the difference between MZ and DZ twins that you had best have a way of telling them apart, especially when you are looking a small subset where the two populations could very easily be over or under represented.

    Take for example of the years 1992 to 1994, there were only 35 sets of same sex twins. If even a few of the pairs were misidentified as MZ instead of identical, that could skew the results badly. 1 twin pair misidentified is about a 5% change and it would be very plausible for even random chance to shift that number when we are talking about so few twins, let alone any actual bias. And I would suspect that there is some real bias here.

    As far as this being a random sample, you have a point. But, this is not the only large twin study in the past year, and this one - - found different results using 277 twin pairs AND had a way of separating out MZ and DZ twins. So the current study had more twins, but not by much, and not by a factor of 5. I found it interesting that this paper did not reference one of the most recent results on the subject.

  3. (Have I ever mentioned that the comment system on blogger is terrible?)

    One other thing, this line from the study -

    "Because mutations are rare events, the same de novo mutations should always be present in both MZ co-twins and lead to high concordance for the expression of autism."

    struck me the wrong way, because I know that it is not true. My MZ daughters have difference CNVs, which is goes against what the authors suggest should be the case. And they didn't share any more resources in the womb than DZ twins do (they were Dichorionic-Diamniotic MZ twins) so I have to wonder about this statement as well -

    "Because de novo mutations are very unlikely to be shared by DZ twins, such sporadic
    mutations should generate discordant DZ twin pairs"

    Because my daughters have different mutations, they had to have happened after they split - which means the same exact mechanism that caused their mutations could cause similar problems with DZ twins.

    The other numbers for twins vs ordinary siblings are interesting but aren't anything new. Even the CDC acknowledges that DZ twins are more likely than ordinary siblings to both have autism. If anything, the numbers (once you exclude the unknowable MZ vs DZ split) confirm what other findings have shown in the past.

  4. And this was peer-reviewed? perhaps what was really needed was a decent editor.

    I really believe that a lot of studies start with the premise "if the title contains the word Autism we are guaranteed funding/publication."

    I also really believe that no one is bothering to teach proper writing techniques, and the importance of same has been tremendously devalued. Finding citations should not be so difficult! -where did this appear originally? who are the authors? further research should not be needed for answers to these basic questions...