As I talked about last time, a new twin study entitled "Genetic Heritability and Shared Environmental Factors Among Twin Pairs With Autism" was published earlier this month. Since this is probably going to be a memorable study, I thought I would go into exactly that it did and didn't show.
First of all, the study and an accompanying editorial are open access so the full text is freely available. If you are interested in the subject, I strongly suggest you go read both for yourself (see links at the bottom of this post).
The purpose of this study was to look at how common it is for identical and fraternal twins to both have autism. The conventional wisdom up to this point has been that if one identical twin has autism the other is almost certain to have it while if one fraternal twin has autism the other has roughly the same risk as any other sibling.
To that end, researchers used the records from California's Department of Developmental Services (DDS) to identity all twins both between the years of 1987 and 2004 where at least one of the twins had a qualifying diagnosis of autism. A qualifying diagnosis was either receiving services for autism, a code that indicated suspected autism, an actual diagnosis of autism, a code indicating intellectual disability of an unknown source, or another developmental disability.
A set of twins was eligible to be included in this study if the mother lived in California at the time of birth, at least one of the parents could read and speak either Spanish or English, both twins were currently living in California, both twins had a mental age of at least 18 months, and there was no history in either twin of "neurogenetic conditions that might account for autism" (such as fragile x, down syndrome, or tuberous sclerosis).
A total of 1,156 twin pairs met the these criteria.
Of these twins, 384 could not be contacted, 330 declined to participant, and 10 pairs were ineligible (not sure why), leaving 432 pairs of twins. Of these twins, 90 additional pairs declined to participate leaving 342 pairs (by my calculations at least, the study text has 333) that could have been evaluated in the study.
In the end, 202 out of the initial 1,156 twin pairs were evaluated. Each twin was evaluated with both the ADOS and ADI-R to establish whether they had either a narrow form of autism (autism), a broad one (asd), or no form of autism. These groupings would roughly correspond to classic autism (narrow) and pdd-nos and/or Aspergers (asd). The criteria that were used to group the twins into these groups were pretty conservative. In order to be included in either autism group both the ADOS and ADI-R had agree.
After the evaluation, another 10 twin pairs were excluded because neither twin met the definition of even the broad form of autism. This brings the total number down to 192 twins pairs.
As an aside, if you look in the supplemental material you can see that the majority of the twin pairs excluded (6 out of 10) had received services for "full spectrum autism" and another for "suspected autism". The study doesn't say whether it was one or both twins that had the diagnosis, but the fact that at least some some of this children once had a diagnosis of autism but no longer is yet more confirmation that it is possible to loose the label.
Although on the flip side, the researchers did diagnosis some of the twins with autism who did not have a diagnosis before. Which shows that it is possible for autism to be missed even when another child in the family is one the spectrum.
But getting back to the study, each of the remaining 192 pairs of twins were genetically tested to determine whether they were identical (MZ) or fraternal (DZ) twins.
For those of you who don't know, MZ twins come from one fertilized egg that, for reasons unknown, splits into two organisms sometime in the first two weeks after fertilization. Identical twins are, by definition, genetically identical - at least up until the point that they separate from each other. DZ twins happen when two eggs are fertilized at the same time and are basically the same as siblings except that they are conceived and born at the same time.
There are a few other important differences between MZ and DZ twins to be aware of. MZ twins normally share more prenatal resources (placenta and amniotic sacs) than DZ twins do. MZ twins are basically random and happen in all races around the world in about 1 out of 250 pregnancies. DZ twins, on the other hand, are strongly dependent on the mother's age, race, and family history. The chance of having DZ twins goes up as the mother gets older and the rate of DZ twinning can vary pretty drastically depending on where you look.
So, out of the 192 sets of twins, 45 were found to be male-male identical twins, 9 female-female identical twins, 45 male-male fraternal twins, 13 female-female fraternal twins, and the last 80 were male-female fraternal twins.
The researchers then looked at how many of each twin pair either both had a narrow version of autism or a broad one in common. The results were somewhat surprising. The following numbers are in terms of how many of the twin pairs both had the condition (pairwise) rather than the more confusing notion of probandwise. If you want to see the numbers the other way, look at table 2 in the study text.
In the male-male identical twins -
42.5% both met both met the criteria for strict autism
64.4% both met both met the criteria for broad autism
In female-female identical twins -
42.9% both met both met the criteria for strict autism
33.3% both met both met the criteria for broad autism
In male-male fraternal twins -
12.9% both met both met the criteria for strict autism
20.0% both met both met the criteria for broad autism
In female-female fraternal twins -
20.0% both met both met the criteria for strict autism
30.8% both met both met the criteria for broad autism
In male-female fraternal twins -
3.6% both met both met the criteria for strict autism
6.3% both met both met the criteria for broad autism
There are few notable things about these numbers.
First, the concordance rate for male and female identical twins is lower than would be expected from other estimates in the past.
Second, the concordance rate for same sex fraternal twins is much higher than would be expected.
Third, if you look at the difference between identical and fraternal female-female twins and compare that to the number for male-female fraternal twins, you will see something odd. It looks like having a female co-twin instead of a male greatly increases your chance of having autism. I could understand that happening with identical twins but I don't understand why fraternal twins have the same result. This might just be a result of having a small number of female-female twins (22) but I am skeptical.
Finally, the concordance between opposite sex fraternal twins should give us a good baseline of where typical siblings would fall using the same criteria. It isn't an exact comparison because there are risk factors associated with twin births (maternal age, prenatal complications, premature birth) that have been shown to increase the risk of autism. This range of concordance (3 to 6) agrees with other non-twin sibling results in the past.
The researchers used a variety of statistical models which took into account a number of different factors in an attempt to explain the results. They also cross checked the twins who participated in the study against those who didn't to ensure that they didn't get a biased sample. They found no glaring problem that they could see, so their conclusion, in their own words, was -
The results suggest that environmental factors common to twins explain about 55% of the liability to autism. Although genetic factors also play an important role, they are of substantially lower magnitude than estimates from prior twin studies of autism. Nearly identical estimates emerged for ASD, suggesting that ASD presents the same liability spectrum as strict autism.
On in other words, the researchers think that their data shows that autism is more environmental than genetic. This result is obviously at odds with what other twin studies have shown and some consider it to be controversial.
Now, there are a few possible problems with the study.
First, a substantial part of the twin population did not participant in this study. It is possible that the twins that did participate were not representative of the entire twin population or that there was something different about the twins who did participant.
Second, the number of identical twins in the sample is rather small. If there was any skewing in which twins participated, it could have been amplified by having only 54 identical twin pairs in the study. On the other hand, not many other studies in the past have had even 54 confirmed sets of identical twins.
Third, the study excluded known genetic causes of autism. Doing so could have artificially lowered the number of identical twins who both had autism because both identical twins would presumably have the same genetic problem and could have raised the number fraternal twins who did so because, again presumably, fraternal twins who would not have been co-morbid for this other defect were excluded.
Finally, it is possible that the results could be explained by some risk factor specific to twins. All types of twins are more likely to have prenatal problems, be born premature, or have a lower birth weight. All of these factors are thought to increase the risk of autism. Fraternal twins are far more likely to be born to older mothers than either identical twins or non-twin siblings. Older maternal age is also another risk for autism.
However, if this were a major factor, I would have expected the concordance of opposite sex twin pairs to be much higher than it was. This is the largest single group of twins included in the study and yet it has the lowest concordance.
There are a few strengths of the study as well.
First, the twins included in this study were a sample drawn from a large population and should be more representative of twins in the general population than other studies in the past. California's DDS is thought to serve the majority (greater than 80%) of children with autism in the state, so it is likely the initial sample included the majority of twins with autism born during the relevant period.
Second, each pair of twins included in the analysis were genetically tested to see whether they were identical or fraternal twins. Other recent twin studies have either relied on parental reports or have used estimates to determine which twins were fraternal and which were identical. Both of these methods suffer from serious flaws that could easily miscategorize identical twins as fraternal or fraternal twins as identical.
The only reliable way to tell the difference between the two types of twins is if they are the opposite sex or if they are genetically tested.
Finally, this study used a far more rigorous and conservative way of determining which twins had autism than did past studies. This may have caused the concordance numbers to be lower than they would under a broader definition, but if it did that would bump the numbers for the fraternal twins as well as the identical ones.
All in all, this is a pretty solid study that casts doubt on the idea that autism is primarily a genetic condition.
1. Hallmayer, Joachim et al. 2011. “Genetic Heritability and Shared Environmental Factors Among Twin Pairs With Autism.” Archives of general psychiatry 1-8. http://www.ncbi.nlm.nih.gov/pubmed/21727249.
DOI : 10.1001/archgenpsychiatry.2011.76
2. Szatmari, Peter. 2011. “Is Autism, at Least in Part, a Disorder of Fetal Programming?” Archives of general psychiatry 1-2. http://www.ncbi.nlm.nih.gov/pubmed/21730328
DOI : 10.1001/archgenpsychiatry.2011.99