Diagnosing Autism in Newborns: The 'Wave' of the Future?

Andrew N. Wilner, MD


April 06, 2018

Andrew Wilner, MD: Welcome to Medscape. I am Dr Andrew Wilner, and today I have the pleasure of speaking with Oren Miron, a research associate in biomedical informatics at Harvard University and the winner of the 2017 Next Einstein Competition. Oren, I have watched a couple of your video presentations where you talked about a tool for the early detection of autism. Can you tell us about that?


Oren Miron, MS: We used the test called the auditory brainstem response (ABR), which is done on every newborn to detect hearing impairment, and we modified it so that it also can detect autism. We accomplished this by looking back at databases, including thousands of brain tests, and checking which of these patients have autism today. We found that while children with hearing impairment do not have a stable response to this test, children with autism tend to have a delayed response. We found this to be a great predictor of autism and have published two studies about it.[1,2]

We are now trying to develop the test further so that it can be included with those hearing tests. Thus, we would turn those 4 million hearing tests that are done each year into 4 million autism tests. Our hope is that, together with other risk factors and eventually with more thorough and comprehensive behavioral diagnoses, this will allow earlier validation of autism, perhaps at age 2 or 3 years (now it is at 4 years), and earlier treatment with potentially better results.

Wilner: When I watched your video, I realized that you were talking about brainstem auditory evoked responses (BAERs). Back when I was a resident, we used BAERs a lot in our practice, and occasionally we use it now. I know that this test is used regularly but usually for a child who is not behaving properly, not developing properly. They use the test to make sure that the child can hear. Obviously, if someone cannot hear or see, it is very difficult to interact with the environment. The test is a way of ruling out organic disease and saying that, even though this patient can hear and see, he is not behaving normally; maybe this is autism. You have turned that on its head. You are saying that there are patterns in the ABR that suggest autism directly.

I was always under the impression that autism is probably a wiring issue of the cortex; that it is a cortical disease. ABRs are a brainstem test. Tell me about that. What is the thinking there?

Miron: I was surprised, too. When I first started learning about autism, everyone was focused on the prefrontal cortex, and the newer the cortex, the better. When I looked at the literature, I saw that there have been studies dating from the '70s. I have just published a meta-analysis that included these.[2] Dozens of studies were conducted with ABR and found this prolongation of response in children with autism. At some point, they understood the ABR so well that they started using it to screen for hearing impairment; they did not just use it for those children with problems. And now it is being used on every baby in most US hospitals. I thought that if this delayed response happens in children with autism, why don't we also check it in newborns and look for a correlation? And that proved to be successful. I definitely built on the work done by these earlier researchers, and I am very grateful for them, especially Professor Hildesheimer, who not only found the abnormality of autism in ABR for the first time but also invented the ABR device that we use. I am standing in their shadows.

We have been developing this for a while now. I joined the MIT Solve initiative and was selected as a Solver, which gave us many ways to spread out, especially to developing countries, because it is a very global organization. This test is inexpensive, which is important to me. The baby comes to you, you give the click, and 5 minutes afterwards you know if the baby hears.

That is why this test is rapidly expanding to developing countries where autism is never diagnosed. If we can use this test to help detect autism in those countries, that would be phenomenal. MRI is amazing and understands the cortex perfectly, and we can definitely work with MRI. But you cannot use MRI on every baby because it is a million dollar device that is bigger than the room I am in. It takes a lot of expertise. In southern Israel, where I started using ABR, we had one MRI for a million people. People waited in lines and came in at 3:00 AM for MRI. How could we do that for every baby?

Finding this strategy grew from my wanting to find a test that would be affordable for everyone. If I meet with a youngster with autism, and that family cannot afford this test, then what I am doing here?

Wilner: As far as I know, there are not MRI criteria for the diagnosis of autism, although every now and then I read a paper reporting this or that abnormality or structural difference. Are you looking at that? Now that you have a dataset and can identify a baby who is likely to have autism, will you pair that with imaging studies when the baby is older to see whether there is a visible structural brain abnormality?

Miron: That is something we are very interested in. We think that those studies are quite promising. The difficulty is that they have primarily been conducted on siblings because you do not know who is autistic and who is not, so you have to test those with a lot of risk to get a sample size of 10 or 20. With our test, if we could detect those at risk and begin to follow them with MRI, we will have a much better understanding.

Connecting to what you asked earlier, the cortex versus the brainstem, the studies show a definite cortical problem, but the cortex matures later on. If you want to detect autism early, you want to look at what matures early. Those studies do suggest a problem in the brainstem—in the entire lower part of the brain. Again, some studies show this, and some studies show that. Especially in autism, with all of the heterogeneity and changes over time, it is very difficult to find consistency, but there is a very good basis for it. We are not yet sure what causes autism. For us, the main thing is to predict the risk early and help the kids. I do not care if it is astrology, as long as it predicts autism risk.

Wilner: You mentioned heterogeneity, and we talk about autistic spectrum disorders. When you looked at the wave forms, what was your gold standard for saying, this is autism, this is not autism, because it is such a broad clinical diagnosis.

Miron: Results were based on the diagnosis the experts made. We did not make the diagnosis ourselves. This was a retrospective study; and, thus, there are no biases on whether or not autism is diagnosed in a given child. We used the entire birth record from the hospital and the entire diagnostic record from the autism center. We found that about 70% of the children were well defined by ABR. We are now developing other [tests] that may detect the other 30% in some different fashion. We keep getting better at it.

It seems that there is a big subgroup we want to focus on. Because autism is so heterogeneous, there will probably be groups that we cannot detect. Rett syndrome used to be considered part of the autism spectrum, but we have not found many ways to connect that with the larger group. There are probably more of these tiny subgroups, but we think that when we are able to distinguish the 70%, we will then be able to better understand and start treating the other 30% because we will know that they are different. Perhaps we will find more children with Rett syndrome or fragile X syndrome.

Wilner: Some are false-negatives; they are negative because they are phenotypically autism but with a different pathophysiology. I believe that is a huge breakthrough. Which of the five waves on ABR was abnormal? Or is it some combination?

Miron: We primarily focused on the fifth wave, which is generated by the highest part of the brainstem.

Wilner: Around the inferior colliculus?

Miron: We are trying to reach the meniscus, and now we are trying to reach even higher because the ABR has a wealth of information, and we hope to find more. Obviously, the more participants we have, the more we can combine different measures. But we need to be very careful not to overfit the data.

Having thousands of participants sounds good, but because only 1% or so of them will be autistic, we had to be very careful in these analyses, especially because there may be so many subgroups. If we are aiming to get 50 or 100 or more cases to only look at half, we do not want to overfit them with 10 markers. We say that you can predict autism based on a social security number if you just have enough of those.

Wilner: Is it a prolongation of wave V? Is there an increase in the latency, or is it an amplitude issue?

Miron: It is a prolongation.

Wilner: The ideal study will be one that will take you through the next 5 years. You are going to do a prognostic study. That is a tough, big challenge.

Miron: We are developing it. That is also going to give us a lot more control. I have become a big fan of retrospective studies, especially now that I am developing a prospective study and dealing with the bureaucracy and the funding. In the retrospective studies, we got hundreds of thousands of patients; and now, with a prospective study, how much does $50,000 get us with 100 participants?

Wilner: That will pay for the consent forms.

Miron: Exactly. The institutional review boards are much stricter now. I definitely recommend all of our listeners to explore this and similar tests. I am at Harvard in biomedical informatics under the supervision of Isaac Kohane, who founded the department. He always says to make the data work for us. Why collect data on 100 patients for $50,000 if you have not yet looked at 50,000 or 100,000 patients? We should do both, but starting with the existing data saves much money, time, and energy, and it focuses you.

It seemed crazy at first to say that we were going to detect autism at birth, but I said to myself, if I am looking at existing data, I am wasting my time but not wasting anyone's money to check it. Thankfully, it turned out. Now with MIT Solve, we are trying to push it ahead to implementation. I am looking forward to helping. The test has already been done on 98% of the [newborn] population in the United States.

Wilner: Oren, I want to thank you and to wish you luck. I hope we can talk again in 5 years or sooner, when you have the results of that prognostic study. I think it is a wonderful opportunity to diagnose and potentially help children with autism.

I am Dr Andrew Wilner reporting for Medscape.

Editor's Note: Oren Miron has disclosed no relevant financial relationships.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: