Dr. Sally Ozonoff is the Principal Investigator of the Infant Sibling Study Team at the UC Davis MIND Institute. We talked with Dr. Ozonoff about her autism research work, how the Infant Sibling Study operates, and what the study does and does not mean. This interview took place before yesterday’s CDC announcement about autism prevalence rates, but Dr. Ozonoff’s team had some similar findings within their own study, especially regarding earlier/more intense scrutiny leading to earlier and increased rates of autism diagnoses.

How did you become involved with the Infant Sibling study? Is it in line with your previous areas of research?

I have always been interested in diagnostic issues within the autism spectrum. I began my work in the autism field researching individuals with Asperger syndrome, trying to understand if and how they differed from individuals with autism. This was back in the early 1990s, when Asperger syndrome had just been added to the DSM-IV. I also spent about a decade doing research on the neuropsychological strengths and weaknesses of children on the spectrum, looking at their executive function skills, theory of mind abilities, visual processing, etc.

In 2001, I was recruited to the MIND Institute. Part of going to a new institution is the opportunity to re-tool your laboratory, since you have to buy new equipment, remodel space to suit your research needs, hire and train new staff, etc. I had always been very interested in early detection and when the earliest signs of autism were apparent. However, these were difficult questions to address in a lab that I had built around the needs of high functioning older individuals with autism and Asperger syndrome. So, when given the opportunity to change course somewhat as I moved to UC Davis, I began to study infants at high risk for autism and haven’t looked back since!

It’s been very interesting work. I am trained as a clinical psychologist and have always worked with families in clinics throughout my career. One aspect of the infant sibling research that is the most satisfying is that it has direct clinical applications. The things we are learning in the lab are directly applicable to families who are participating in research and, I hope, are helpful to them … in addition to advancing science and helping others.

How were Infant Sibling Study participants recruited?

An Infant Sibling Study participant, and her autistic brother.
Photo © Shannon Des Roches Rosa

Participants were infants (under the age of 18 months old) who had either an older sibling with ASD (we call this the High Risk group) or older siblings who were typically developing (Low Risk group). The High Risk group was recruited through the MIND Institute, Regional Centers, and various community organizations serving children and families with ASD. The Low Risk group was recruited through flyers posted around the community and the obstetric departments of local hospitals.

The media has been very interested in your study and its results so far. Do you feel the right messages are getting out?

Yes, I do. The bottom line is that the results have very important implications for how primary care providers follow and manage the care of infants who have older siblings with ASD. Our study showed that they are at much higher risk for ASD than the general population. These are high risk patients and need to be treated as such by their doctors. Just as internists would perform special tests and surveillance of a patient with risk factors for cardiovascular disease (family history, high cholesterol), pediatricians should be aware that they have a child in their practice who is at very high risk and will need special watching. This is a message that the media attention has really helped us get out.

Can you explain for a lay person how your study controls for families who stop having children due to having an autistic child?

Most previous estimates of how likely a family was to have another child with ASD were based on the rate of families who had two or more children with ASD (divided by the total number of families who had any children with ASD). This did not control for the fact that some of the families included in these calculations had only one child or had their child with ASD last, followed by no more children. For those families, there was no chance that they could have had a second child with ASD, so they shouldn’t be part of the calculations.

The right way to do a recurrence risk study is to only include families who have chosen to have another child, after the child with ASD. That is the research design used by infant sibling studies, so they are better able to estimate the rate of recurrence. I should say, however, that even this study design is not without its flaws. For example, parents may choose to join the study because they are already worried about their child’s development — if this happens, it can bias the study toward finding a higher recurrence rate than exists in the total population of families with ASD because we might be studying a more affected subgroup of families who have self-selected to participate.

The ideal study design is to enroll families during the pregnancy, before the infant is even born. Other studies at the MIND Institute that I am part of, the MARBLES and EARLI investigations headed by Dr. Irva Hertz-Picciotto, use these methods to best estimate recurrence.

Were the siblings in the study diagnosed substantially earlier than average? Has this provided a way to explore the value of early intervention?

I cannot speak for all twelve sites that were part of the Baby Siblings Research Consortium study, since we did not analyze the age of diagnosis in the first study we published from this large dataset. I can respond based on the smaller sample that comes from the MIND Institute: Yes, our sample is diagnosed significantly earlier than average (which, across studies in the US in the last five years, is between 3 and 4-½). The average age of diagnosis at our site is 24.6 months (ranging from 12 months at the earliest to 36 months at the latest).

You’re absolutely right, these kinds of studies will help us now study the effect of age at intervention to see whether initiating intensive treatments before age two has a bigger impact on development and functioning than starting them later. At several sites around the world, including the MIND Institute, studies of this type are currently underway, since the infant sibling studies they are associated with have the opportunity of identifying children so much earlier.

Has there been any exploration of the role of epigenetics? Would you expect to see manifestation of trans-generational epigenetics in the sib study?

The studies I have done so far have not conducted genetic or epigenetic analyses, but this is clearly a fruitful area of future investigation.

Do you think that in these families with more than one autistic child and who might also have other autistic relatives, there’s a stronger genetic component than, say, in a family with three children, only one of whom is autistic and no other known autism? If so, are there studies looking at these as differentials?

Our study did look at the difference in rates of ASD recurrence if a family had one child with ASD versus more than one child with ASD. The overall rates were different, with families with more than one child with ASD having, on average, a 32% recurrence rate, while those with one child had, on average, a 20% rate. There were too few families in even our very large study to examine the specific scenarios you ask about but all the sites continue to collect data. Someday I hope we will have a sample large enough to ask specific questions about whether the recurrence rate is elevated if there are second or third degree family members (cousins, uncles, etc) with ASD.

Are you looking for or noting Broad Autism Phenotype characteristics in the sibling study? How and, if possible to discuss, with what results?

The broader autism phenotype is not a diagnosis or a classification with a specific definition, but it is a concept that describes milder, subclinical features of ASD, which do not cause significant impairment. We are very interested in how often this occurs in our sample and when such characteristics might first become apparent.

We study the infants in a variety of ways — we examine their language, social, cognitive, temperament, attention, and behavioral characteristics, looking to see if difficulties in any of these areas occur more often in the High Risk group than in the Low Risk group. If so, this might be evidence of the broader autism phenotype. So far, we have found that some mild social and language delays are seen in some of the siblings, such as delayed development of joint attention and delayed language.

What is interesting is that delays in these areas are usually thought of as early signs of ASD (which they can be), but what we are finding is that they do not always inevitably predict ASD. Some children will show delays of this type but then go on to have non-ASD outcomes (which we might call the broader autism phenotype). So the good news for families is that early delays in these areas do not always translate into a later diagnosis; many of these children will turn out to be developing pretty much in the normal range at age 3, with possibly just some mild “quirks” or stylistic differences. This can be reassuring news to worried parents.

Regardless, we always help families find appropriate interventions right away. This is unpublished data that has not gone through peer review yet (we are preparing the paper right now), so I’m afraid that I can’t share many other details at this point.

Are you seeing any differentials in siblings of children with what they call “regressive” vs non-regressive autism? If genetics analyses are involved, are there any patterns in terms of genetic hits for siblings of autistic children stratified by subsets of any kind?

The work out of our infant sibling study at the MIND Institute suggested that many more children experienced a regressive-like course than previously thought. We intensively studied a small sample of 25 children from birth through age 3, taking detailed measurements of their social abilities every 6 months or so. What we found is that the majority of the children (over 80%) showed fairly good social abilities, in the average range, at 6 months, but over time they lost interest in people and became less socially interactive.

This data could be interpreted as evidence of a regression in these children. It suggests that regression may be much more common than previous studies have found. It was just easier to detect using our study methods, since we were giving the children detailed examinations every few months, rather than asking parents to recall detailed information from several years ago.

In terms of genetic analyses, this is something that the Baby Siblings Research Consortium is very interested in studying and investigations are in progress to collect and analyze this data. The work is ongoing but hopefully we will have some answers in the near future … check back!

Disclosure: My youngest daughter participates (with variable compliance) in the Infant Sibling study. Sincere thanks to Emily Willingham and Matt Carey for contributing to the question set. -SR