Translating developmental social neuroscience advances into solutions for the early diagnosis of autism spectrum disorder

May 25, 2023

Information

YCSC Grand Rounds May 16, 2023

2023 Donald J. Cohen Lecture

Ami Klin, PhD, Director, Marcus Autism Center, Children’s Healthcare of Atlanta and Emory University School of Medicine

ID9960

To CiteDCA Citation Guide

00:02Good afternoon,
00:08Okay. Next week we have a special
00:10speaker coming from Drexel University,
00:12Adam Bemporado, who is the author of a
00:15recent book called A Minor Revolution,
00:18Wonderful book about why we
00:19should be investing in children.
00:21So if you are not sure whether
00:22investing in children is a good idea,
00:24come next week. Further,
00:26if you don't think it's good after today,
00:29you're going to change your
00:30mind because today's talking,
00:31among many other things,
00:32is about investing in children.
00:35So it's a double honor for me today to
00:39introduce today's lecture and lecturer
00:42and my what a what a pairing we have.
00:45So today is the Donald Cohen
00:47Memorial Lecture as we sit here
00:50in the Donald Cohen Auditorium,
00:52many of us.
00:53Got to know Donald as a great scholar,
00:57leader in the in many fields developmental
01:00psychopathology and autism particularly
01:02relevant today and and also as an
01:05incredible mentor to many of us,
01:07Linda Mays,
01:08Andres Martin and our speaker included.
01:13So there's the Donald Cohen angle to
01:17today's talk and then of course there's
01:20the Ami Clinton angle to today's talk.
01:22Many of us know and yes,
01:25love Ami Klin and have for many years.
01:28Ami was here between 1989 and 2010,
01:33so he had a long run here at the Child
01:36Study Center and then 12 years ago.
01:39Alas for us and oh fortune for Atlanta,
01:44he was recruited as the founding director
01:47of the Market Center for Autism, which is.
01:50Just an unbelievable creation that
01:53you're going to be hearing about today.
01:57Ami is originally from Brazil.
02:01His schooling was both in Brazil than in
02:04Israel and ultimately his PhD in London.
02:08When we were reminiscing paying
02:11his debt for his PhD studies,
02:14he worked as a delivery boy in
02:18a motorcycle through London.
02:20And I think that the back of the
02:22thing read anything, anywhere,
02:25anytime, anything to any to anywhere.
02:29And thus a great career was born.
02:32When Donald saw this spectacle,
02:35he said you are not going to endanger that
02:38brain of yours with or without a helmet.
02:40This must stop.
02:41And Donald brought on me to New Haven and
02:45a great, great, great career was born.
02:49And I'll.
02:51I'll let Ami tell things,
02:54but I want to just add one thing.
02:56Someone said, and I'm not sure who did,
02:58but someone said that beauty is
03:01the reflection of truth.
03:03Bob King, do you know who said that?
03:06Well, OK, so Bob King doesn't know,
03:09so it's unknown.
03:09But someone said that.
03:12Let me repeat that beauty
03:14is the reflection of truth.
03:16And when I've seen Ami's studies and papers,
03:19this is true.
03:22His science is beautiful.
03:24So his science is so beautiful.
03:26You want to frame it in your living room.
03:29You want to eat it.
03:30It's delicious. It's beautiful.
03:32But it's not just beautiful to the eye.
03:35It engenders deep truths,
03:37deep truths that through all of these
03:40incredible titles and positions
03:42and platform that Ami now has.
03:45He is taking to the city of Atlanta,
03:46to the state of Georgia
03:48and indeed to the world.
03:50And we are all the more fortunate for it.
03:52So Ami via Manila, La Casa.
03:54Welcome back home.
03:54Love having you,
04:04Andres. Thank you.
04:07This is, this is home away from home.
04:09Actually, this is, this is,
04:11this is the home I grew up,
04:14certainly intellectually.
04:15Thank you for just very,
04:17very generous introduction.
04:19The fact is that I owe,
04:21I owe my my, my intellectual,
04:24my professional career,
04:25my clinical career to everything
04:27that happened between those years
04:29that you've mentioned and some of
04:31those giants that I was influenced by
04:33are sitting right now in this room.
04:35And so thank you, thank you,
04:37thank you, thank you,
04:38thank you for everything that you've
04:41done so for me being in this hall.
04:45Is Needless to say, is very emotional.
04:48There are some very important
04:49people that influence my life.
04:51You mentioned Donald.
04:52Donald, yes.
04:54I was the only dispatch writer
04:56in London with a PhD and and
04:59meeting him at Swiss Cottage,
05:01he went beyond the way that I was dressed
05:04and obviously did thanks to the larger.
05:06So she's hot and you could see that
05:08maybe there was some potential
05:10and that I could come here.
05:11But he has been present in
05:13our lives every day.
05:14And I cannot thank him for all that
05:18he's done on my behalf of the years.
05:21There are others that I would
05:24like to briefly remember.
05:26Dom and Sarah. They adopted me.
05:29They held my American wedding in
05:33their house and they cooked for it.
05:35And I miss them.
05:37I miss them really badly.
05:39And then there are the folks that
05:41I work for a long, long, long,
05:43long period of time in the trenches
05:46and I need to thank them too.
05:48Fred is not here,
05:50but I owe,
05:51I owe tremendous thanks to Fred for
05:54allowing me to share 20 years of his
05:58clinical life and that has been tremendous.
06:01And then Warren Jones
06:04somebody that I've been.
06:06Collaborating with for the past
06:092223 years right here at the
06:12CDU 1st and everything that I'm
06:14going to present to you today is
06:17a cocreation with Warren Chance,
06:20OK Disclosures.
06:22This presentation includes research
06:24related to investigational
06:26device development.
06:27Warren and myself are Invent
06:29Inventors patent holders.
06:31This has been licensed to this company
06:32called Early Tech Diagnostics,
06:34which evaluates medical technologies
06:36and hopefully provides revenue to
06:38support treatment of children with autism.
06:40This external activity is monitored
06:42by the Dean at Emory University
06:45School of Medicine and off we go.
06:47Autism is a huge public health challenge.
06:52You'll know one in 36 is the most
06:55common complex near developmental condition.
06:58And I like to to think in terms of
07:01the fact that 95,000 children will
07:03be born this year who will have autism.
07:07It's enormous societal cost,
07:09but most importantly an enormous economic
07:12burden to the individual families.
07:15The good news is that early
07:18diagnosis and early.
07:20Intervention and support can optimize
07:23outcomes that are going to make a
07:27difference in this children's lifetimes.
07:30And here is the not so good thing,
07:32Even though we've learned for the
07:34past 20 years and the American
07:36Academy of Pediatrics has recognizes
07:37at least from
07:382007, the median age of diagnosis of
07:41autism in this country is still late around
07:444:00 to 5:00 and a half, and even later.
07:48For low income families,
07:50for minority families, and to rural
07:53in families from rural communities.
07:55And one of the things contributing
07:57to this challenge is the fact that we
08:00don't have a lot of cautions in life.
08:02We don't have a lot of Jamie's in life,
08:04We don't have many extra clinicians.
08:06So in the community,
08:08Oh my God, so good to see you
08:13Kyle. What a pleasure.
08:15I'm sorry, I just saw you sitting here.
08:18OK, good to see you.
08:21We don't have a sufficient
08:23number of clinicians.
08:24Therefore, out there in the community,
08:27families may wait years and years in
08:30order to see and to get a diagnosis.
08:32And you need a diagnosis in most
08:34situations in order to access services.
08:36So autism is a public health challenge,
08:38is also a public health opportunity.
08:40This is what we've learned
08:42in the past 20 years or so.
08:44Autism is the most strongly genetic of all
08:47complex neuro developmental conditions.
08:49If I have a child of autism,
08:51as you all know, my the younger child,
08:54one to two and five will also have autism.
08:59But the problem is of course,
09:01that the genetics of autism
09:03is enormously complex.
09:05There are hundreds and hundreds of
09:07genetic variants that have already been.
09:09Implicated in autism.
09:10And yet you don't go from genes
09:12to the symptoms of autism.
09:14And here is the hypothesis is that
09:18genetic liability actually disrupts.
09:20And Kyle,
09:21you would recognize something
09:22that you've contributed and
09:24thank you very much for training.
09:27Me too.
09:28Genetics,
09:29actually,
09:30that liability disrupts normative
09:32development and and autism
09:35becomes really the results of the
09:38disruption of normative development.
09:40And for that I need to start
09:42from the beginning.
09:43And the beginning is that babies are
09:46born in the state of utter fragility.
09:49If babies were left by themselves
09:51they would die.
09:52So God or nature made sure that
09:54this would not happen and and so
09:57babies when they are born anything
10:00in their surroundings that is going
10:02to sound like give us or smell or
10:05move or look or interact like give
10:08us is going to command the attention
10:12of newborns and and biology such
10:16that everything happens in past.
10:18And so in the same way that babies
10:21are oriented to the caregivers,
10:23caregivers of course are engaged
10:25with their babies.
10:26And it's out of this mutually reinforcing
10:30choreography that brain moves forward.
10:32And that's a very important concept.
10:35In fact,
10:35these are the three concepts
10:37that I want to make sure these
10:39are coonesque set of concepts.
10:42They are Kyle Esque.
10:44Or they are Sally Prevention.
10:46I don't know how do you do that,
10:48Kyle,
10:48but all of these are concepts that
10:50have been near and dear to most
10:52of us working with young children.
10:54In this building #1 babies are
10:56born to socially Orient.
10:58Second,
10:59what moves brain forward is
11:02reciprocal social interaction.
11:04Simply placing a child in front
11:06of television is not going to
11:08make the child speak.
11:09You need a somebody on the
11:11other side engaging that baby.
11:13Reciprocal social interaction.
11:14And the third concept, of course,
11:17is the one of neuroplasticity,
11:19because this is the first two years of life,
11:22is the period of maximal
11:25neuroplasticity for our lifetimes.
11:28By the time that we celebrate
11:30the baby's 1st birthday,
11:31that baby's brain has doubled and
11:34synaptic density has quadrupled.
11:36Really beginning to reposit.
11:38Those early experiences that
11:41babies have
11:42from the first days and weeks of life.
11:46So in many different ways,
11:47how state is more than once to date.
11:50Brain structure and function is the physical
11:53instantiation of those early experiences,
11:56and we need to remember that.
11:58OK, So what about autism then?
12:02Kasha wrote at least two papers that
12:04I remember on parental concerns.
12:07So parents are concerned about their
12:08children from very, very early on.
12:10But it takes years for parents
12:13to access experts who will be
12:15able to diagnose the child.
12:18And that diagnosis is typically the
12:21ticket to early intervention, to early
12:23treatment in support of the families,
12:25the median age I already mentioned.
12:27But the issue is this is an issue of access.
12:30There are not many child
12:31study centers in this country.
12:32In fact, there are not many
12:34specialized centers in this country.
12:36Remember, 95,000 children are
12:39born every year who have autism.
12:41So this is the work that we've
12:43been trying to sort of to advance.
12:46We need to go, We don't want to replace,
12:49As for clinicians,
12:50people like even myself,
12:52but we need to to give
12:54greater access to families,
12:55to high quality diagnostic services
12:57so that the children can be treated.
12:59We need to use biomarkers,
13:02they need to be objective
13:04and quantitative dimension.
13:05They need to be standardized.
13:07They need to be cost efficient so that
13:09we can scale them and they need to be.
13:11And we need to be able to capture
13:13the core features of the condition
13:15so that when we are devising
13:17treatment plants that we are focusing
13:19on those things that are pivotal
13:21milestones for children who are
13:22going to push forward speech,
13:24language and communication
13:26and cognition and so forth.
13:29So we have Warren and I have been
13:31using this technology called eye
13:33tracking for quite some time.
13:35But the concept that we try to measure is
13:37the concept of social visual engagement,
13:40which is basically the way we look
13:43at and learn about the world.
13:45So imagine social visual engagement
13:48in babies will be.
13:49What they are looking at and learning
13:52about in their immediate social environment,
13:56and here is just an example.
13:57So you see that cross there.
13:59That cross is basically a toddler
14:02looking at this two toddlers
14:04interacting K and you can then measure
14:08120 times a second moment by moment
14:12how that toddler is hi waiting,
14:15the perception of that experience that
14:17is happening right in front of them.
14:19So it's a Co creation.
14:21That's why we talk about
14:23transactional biomarkers,
14:24because what's being created
14:25is not only on the baby side,
14:28it's not only on the other,
14:30it's actually the mutually reinforcing
14:32choreography between the two is that
14:36baby creates an experience that is
14:38what's happening in front of them,
14:40a transactional biomarker, OK.
14:42So,
14:42but one of the big sort of technical
14:45problems that we've sort of
14:47encountered is how do you quantify
14:49social visual engagement particularly
14:50for groups of people.
14:52And so we started this way,
14:54this is what we want and see if it
14:59so this is the battleground
15:01of brain development in tops.
15:13I die for your thoughts. Here we go.
15:19This is how children learn about paying
15:22attention to people's faces and facial
15:24expressions and nonverbal gestures.
15:26And and and communication and and
15:29and and maybe calling somebody
15:31else in conflict with such.
15:33I mean, you name it,
15:34This is what we want to know.
15:36How does a child create?
15:39An experience out of the simulator
15:40that I just put in front of you.
15:42So this is what we did.
15:44This is a frame of video,
15:45the 30 frames per second.
15:47And then we basically measure a say a
15:502 year old one child looking at that
15:53frame of video at that moment in time.
15:56And what you see there is an equation
15:58and that equations basically shows the
16:00amount of visual resources that that
16:03toddler is dedicating to that point
16:04of the screen at that moment in time.
16:07And that equation comes from the mapping
16:10of photoreceptive cells in in primate eye.
16:13OK, Now if you have 45 two year olds
16:17looking at that frame at that moment in
16:19time and they're looking at the same spot,
16:21what you see is that the colors
16:23are getting hotter, OK,
16:24And that means that there is a convergence
16:27of attention across all these children.
16:29Remember,
16:30there is a frame of video there.
16:31Now let's take a different
16:33kind of perspective.
16:34Let's take the bird's eye view of that frame,
16:37and what you see in front of you now is
16:39something that we've called a salience map.
16:42And what you see is that the hot color,
16:44like the Reds,
16:45means that more children are converging
16:48attention on and the Blues you don't
16:50see many children paying attention to.
16:52And now what you can do is to basically
16:55watch the same video that I showed you
16:58before through the eyes of 45 two year olds.
17:01And you see that this is
17:03changing all the time.
17:04Sometimes it is is dispersed,
17:06but sometimes something really
17:07important happens in middle of the
17:09screen or in the side of the screen
17:10and they all going to focus on that.
17:12And we were sort of fascinated by the
17:15fact that when we're doing this work
17:18with typically developing toddlers,
17:20they would watch our videos and
17:2180% of the time they'll be looking
17:23at the same spot at the same time.
17:25And so these are what a wonderful child
17:29psychologist from Seattle One said.
17:31This is a hotspot of socialization.
17:34So Andy Meltzer decided to name it
17:36this way because this is a momentary
17:38thing that happens in front of the
17:41child that all children will focus
17:43on because they're all entrained
17:45to that hotspot of socialization,
17:48something they need to learn.
17:50In order to understand social interactions,
17:53OK, let's take a different perspective.
17:55Now let's bring in another axis.
17:58And this axis here is the axis of time.
18:01And what do you see are lots of frames,
18:03the frames of the videos.
18:05Now let's take that salience map
18:07and let's carve it
18:09all the time. OK, And So what you
18:12have now is a space-time distribution,
18:14and when he gets red, when he gets hot.
18:18It means that this is something that
18:20we've called an attentional funnel.
18:23Basically children are looking at
18:24different spots and that all of a
18:27sudden something important happens
18:28and there is a funneling of the
18:30attention of all of these children.
18:32And now what you can do is to watch this
18:35video that I showed you through the funnel.
18:38It basically that funnel right here is,
18:42is is basically encircling.
18:44What is really important for these
18:47children as they are going through
18:50that particular video now let me
18:53just show you here as you see.
18:57Right here, they're not looking at
19:00the faces of those two toddlers.
19:02They are looking at something that
19:04is in between those two children.
19:06And that's the object that they
19:08are fighting over.
19:09Because all of those two year olds
19:11were able to grab the gay skew that
19:13comes from one of the toddlers,
19:15and they are focusing on the object
19:18of shared attention.
19:19And this is happening in
19:22intensive milliseconds.
19:23OK, so keep that in mind now.
19:26This is now.
19:27What we call a normative funnel of
19:31attention as all of these typically
19:33developing 2 year olds are traversing
19:35that particular frame of video.
19:37OK,
19:37and so now what you see are skin paths of
19:42two year olds of autism traversing the same.
19:45The same frame,
19:47and this is what the typically
19:49developing children are focusing on.
19:51And this is what the children
19:52with autism are focusing on.
19:54So they are exposed to exactly the same,
19:57the same stimuli, OK,
19:59those toddlers playing,
20:00but they Needless to say,
20:02they are reconstructing internally
20:04a very different experience.
20:07And what we found out is that those
20:10things happen hundreds and hundreds
20:12of time within 5 minutes of free
20:15viewing of a video like that.
20:18To the extent that we could
20:20measure that our our kids,
20:21the ones with autism,
20:22the two year olds of autism,
20:24were diverging over 570 times from
20:27the experience of their peers.
20:30Now this was happening in 5 minutes.
20:32They are not only diverging
20:34during our experiments,
20:35they are diverging in their lives.
20:37And so if you extrapolate,
20:39imagine thousands of missed
20:42opportunities of social learning,
20:45OK, in a few hours,
20:47and really millions in the
20:48first two years of life.
20:49So this is our conception of autism.
20:52If you miss thousands and thousands
20:54and thousands of opportunities
20:56for social learning,
20:58you will become socially disabled.
21:00OK, now remember that.
21:07These are the experiences that the
21:10children may be having on in their lives.
21:13And if we go by Joe Le Doux's
21:16sort of subtitle of a great book,
21:18that our brains become who we are,
21:20those early experiences are basically
21:23becoming the brains of the children.
21:27That's the reason I meant that the
21:30brain becomes a physical instantiation
21:32fact of those early experiences.
21:35Now this biomarker social visual
21:37engagement has become very important
21:39to us. We needed to show first of all that
21:41he was a biomarker and what we found out.
21:43This is work with our good friend
21:46John Constantino and just.
21:48So we did studies with twins, OK?
21:51And what we found out is that
21:54something as ephemeral as eye looking.
21:56You know, eyes are very important
22:00false mechanism of socialization.
22:01The eyes are not only the window to the soul,
22:05they're the window to to the social brain.
22:09So much so that the more,
22:11say an animal needs caregiving,
22:15the greater the contrast of the sclera.
22:18So, for example, primates have sclera,
22:20humans have most of all,
22:22but a rodent like a chinchilla does not.
22:24That's not the way that
22:26caregiving happens with rodents.
22:27So eyes are very important.
22:29So just thinking in terms of eye looking,
22:31when we had twins that were basically
22:34watching videotapes in two separate rooms.
22:37OK, identical twins,
22:39MZ twins and fraternal twins,
22:42DZ twins.
22:42What we found out is that social visual
22:45engagement is very strongly influenced
22:47by genetic variation with a heritability
22:50of over 90% in fact .91 in this study.
22:54And so the concordance rates
22:56for identical twins was .91,
22:59for fraternal twins it was only .35,
23:02and for just the age and sex
23:05match was only .15,
23:07so under strong genetic influence.
23:10And interestingly,
23:11those highly heritable assays were also the
23:14assays that were separating the children,
23:17the tablets of autism,
23:18you know,
23:19the most efficiently from their
23:21typically developing peers in
23:23one cohort and in another cohort
23:26in a replication cohort now.
23:28What to us was remarkable is that
23:32that genetic control of social
23:34visual engagement was not only on
23:36measures that were summary measures,
23:38like I looking over a period of time
23:40that a child was actually watching a video.
23:43It was moment by moment to the extent
23:46that identical identical twins,
23:48they were more likely than fraternal
23:50twins to shift their eyes.
23:52At the same moment,
23:53in the same direction,
23:55onto the same semantic targets and
23:57microscales of 10s of milliseconds.
24:00So we are really talking about two
24:02synapses from retina.
24:03So this is the extent to which
24:06God or nature decided,
24:08going along with Kyle and Donald,
24:10that we are born social beings,
24:13we are the product of our relationships
24:16and nothing has shaped really nothing
24:18has shaped the primate brain as much.
24:22As sociality.
24:23OK, so this is now what Warren did here.
24:28Even though the twins watched those
24:30videos in different in different rooms,
24:32Warren brought together their eye
24:34tracking signal which is this
24:36crosshair from 2 videos onto the same.
24:39So this is now fraternal twins
24:42watching this video.
24:43You can see the concordance.
24:45Remember they are watching this
24:47video in different rooms
24:50and now. These are identical twins
24:53watching this video in different rooms.
25:01So an agreement of .91.
25:05Well, how early does that assay begin
25:10to segregate children of autism from
25:13their typically developing peers?
25:15This is now in slow motion,
25:18a typically developing 5 months
25:20old looking at the caregiver
25:23falling in love with her eyes.
25:25And this is a 5 month old who
25:28was later diagnosed with autism.
25:31A very different way of looking
25:34at that speaking face, now over 3.
25:39Cohorts, we follow children from the age
25:42of two months to the age of 24 months.
25:45Densely sample collecting eye tracking
25:47data monthly until six months and
25:49every three months thereafter.
25:51And what we are able to see is
25:53that this in blue here is the
25:56growth chart of eye looking for
25:58typically developing children.
26:00You see it goes up, down and up,
26:03up, OK and this goes from
26:05two months to 24 months.
26:07For the children who are later
26:09diagnosed with autism over 3 cohorts,
26:11they start over here and it's free
26:14fall for the first two years.
26:16So this is autism unfolding
26:18right in front of your eyes.
26:20We have differences in the first
26:22six months that are predictive of
26:25diagnostic classification as well
26:27as level of disability at the ages
26:30of 24 and 36 months experiences in
26:32the first six months of life. OK.
26:34Another thing that we found out,
26:36everything that we do with human infants,
26:38we do with infant monkeys because
26:40they are an extraordinary platform
26:42for studying the social brain.
26:44So with colleagues at the Emory
26:45Primate Center,
26:46John Celine by Chevalier and Marsanches,
26:48what you see here is basically
26:51I'm looking in both human infants
26:53from 2 to 24 months,
26:55and here you seen black infant monkeys.
26:58And you need to do a transformation.
27:00You need to divide by 4 because monkeys
27:03mature much more quickly, of course.
27:05But you see that the pattern
27:07of I looking is very similar.
27:09It serves the same adaptive value.
27:11This assay is highly phylogenetically
27:14conserved.
27:15So that's the reason why we're
27:17excited about this biomarker.
27:18But could we use it?
27:21To provide greater access to
27:23early diagnostic services.
27:24So this is what we did.
27:25We took some of this lab and gosh,
27:28I might remember this lab well.
27:30And then we started building
27:31a prototype of something that
27:32would be more portable somebody,
27:34something that could actually
27:35be deployed in the community.
27:36And we ended up with with this
27:39prototype and here is what it
27:41looks like these days.
27:47It's just like the car,
27:48it's just like the cars
27:52hanging in here,
27:57so it's entirely automated.
27:58The whole idea is to make it
28:01as efficient as possible so a
28:04minimally trained technician can
28:07actually perform this procedure.
28:09So this is what the the technician does
28:12while the child is inside watching.
28:18This calibration is automated,
28:19but this is what the child sees. OK,
28:28and
28:32that's the end.
28:33He did such a good job.
28:37So we built this prototype in
28:40order to conduct some studies,
28:43and these studies would make use
28:46of this quantification procedure.
28:48Our follows of attention what
28:49we what we did is that we create
28:52large bodies of normative data
28:54for looking behavior and typically
28:56developing children as they watch
28:58all these things so that we would we
29:01could create this moment by moment
29:03normative benchmark against which we
29:05could compare individual children
29:06And the the first aim of this work
29:10was to derive quantitative indices
29:13for the classification of autism.
29:16And so as you see here, the children,
29:18each individual child diverges from
29:21that normative benchmark hundreds of times,
29:24hundreds if not thousands of times.
29:26Within the context of this procedure is
29:29about between 8:00 and 12:00 minutes.
29:31And so we would mind this thousands
29:33of divergences, as it were,
29:36in order to create a.
29:38A diagnostic classifier, OK,
29:39And the diagnostic classifier is
29:41always going to be the reference standard,
29:44the gold standard.
29:45So expert clinicians conducting the
29:48diagnosis independently of course,
29:50from this experimental procedure.
29:51But we wanted to do a little more as well,
29:54because for those of us who do
29:57those diagnostic evaluations,
29:58it's not only the diagnosis.
30:01Autism is a huge spectrum is
30:03very critical for us to have at
30:05least three additional measures.
30:07One is the level of autism or
30:08the level of social disability
30:10typically measure with the Ados,
30:12the Autism Diagnostic Observation Schedule,
30:14as well as the level of verbal
30:16learning and nonverbal learning
30:18and that of course coming from the
30:20Mullen Early Scales of learning.
30:22We wanted to be able to derive that too.
30:26So we conducted 3 studies.
30:27The first one packaged in,
30:30one we call the feasibility trial,
30:32had a discovery sample and
30:35a replication sample.
30:36And then there was a pivotal,
30:37a national pivotal trial.
30:39The goal we focus on 16 to
30:4130 month old children,
30:43primarily because of the
30:46the the federal mandate,
30:47the Part C services,
30:49the early division services that
30:50most of the children don't access to
30:52because they don't get a diagnosis.
30:54So we prioritize the age of 16 to 30 months.
30:57We wanted to know if this
31:00would accurately assess autism.
31:02Proxy in fact the diagnostic process
31:04executed by an expert clinician
31:07and we wanted to see the extent to
31:10which we could create those those
31:12measures of severity proxying anaidas
31:14and the two measures of the moment.
31:17The the discovery sample was over
31:22700 children primarily at Marcus.
31:25And the replication sample was
31:29370 toddlers and one was at Wash
31:31U with my friend John Constantino
31:34and the other part of it was in
31:38a community Health Center about
31:4050 miles north of of of Atlanta.
31:42We wanted to we want to see does this
31:45work in the real world K not in the lab.
31:48And then the second study was a
31:51pivotal trial national clinical
31:53trial with 335 toddlers.
31:55And we conducted that as Seattle Children's,
31:57Cincinnati Children's, UCSF,
32:00Rush Phoenix and Emery Ki.
32:04Know this is a little small,
32:06but this is from the visibility study.
32:09This is the discovery data
32:11and the replication data,
32:13the clinical characterization sample.
32:15And what you see here is that the
32:18kids were around 21 to 24 months.
32:21They represented the full spectrum.
32:23Of social disability through the Atos
32:25and the full spectrum of verbal and
32:28nonverbal ability using the Mullin.
32:30So some children were severely delayed,
32:32but some children are actually precocious.
32:34And this is the clinical characterization
32:37sample for the national trial And
32:39you see here the classification
32:42was autism versus not autism.
32:45It was not autism versus typical.
32:47OK, so of the 185 toddlers
32:51with non autism diagnosis.
32:5487.6% had other non autistic
32:57developmental delays,
32:58only 13 kids had no diagnosis.
33:01And of the $150.00 of Voltas and 42.7%
33:06did not have developmental delays.
33:08OK, so we had the full spectrum on
33:12both sides and here are the results.
33:15Ignore the discovery study because
33:17this is where we develop our mode.
33:20This is where develop our.
33:22Our mathematical model,
33:24but in the replication sample,
33:27the diagnostic classifier was able,
33:29we're able to achieve over
33:3180% of sensitivity,
33:32specificity,
33:33positive predictive value and negative
33:36predictive value and and and those
33:39numbers were fairly replicated in the
33:42national in the national clinical trial.
33:45What about our indices of severity?
33:48Remember that the ADAS,
33:50the higher the number on the ADAS,
33:52the more socially disabled you are and
33:55it's different in our eye tracking metric.
33:57That's why you have this direction here.
34:00But on the replication,
34:01say for example,
34:02we're able to capture 72% of the variance
34:06of ADA scores in that 8 to 12 minute
34:10procedure and in the national clinical trial.
34:14About 74% of that variance.
34:17OK,
34:18so you remember takes,
34:21it takes a well trained person to do aid us,
34:23a well trained person to do the Mullen.
34:25It takes a long time to complete that.
34:27What about the verbal ability
34:28on the replication sample,
34:30we're able to capture 58% of the variance
34:35and 65% of the variance of verbal
34:39ability and for nonverbal ability.
34:42About 64% of the data.
34:44So we are very happy with those results.
34:48But we needed to make sure particularly
34:50if you're dealing with FDA that you
34:53need to conduct studies that are called
34:55repeatability and reproducibility.
34:56So you need to quantify measurement
34:59precision, measurement error.
35:01So you need to put children through
35:03this ordeal which is the same child
35:06needs to complete the the procedure
35:08three times on the same device.
35:11And then three times again across devices,
35:13OK, measure,
35:14these are the infamous RNR studies,
35:17but the data are good.
35:19So that we could see that repeatability
35:22and representability variance was
35:24accounting for very little of
35:26what we've got and most of the
35:29variance really was due to between
35:32subject participant variance.
35:34So we are very,
35:36very happy with.
35:38This low measurement error
35:39because it gives us more precision
35:41what we do. And if I combine all of the
35:44studies together this is this is the data.
35:47These are the sensitivity, specificity,
35:50PPV&amp;PV and accuracy of the device
35:52right now and and here are the the
35:56amount of variance that they are
35:58capturing of social disability.
36:00The aid us verbal ability to Mullen,
36:03nonverbal ability from the Mullen as well.
36:05So how can I summarize this?
36:08Happy that we got high sensitivity
36:11and specificity when comparing those
36:13eye tracking assays with expert
36:17clinician diagnosis with sensitivity
36:21and specificity and PPV&amp;NP V / 80%.
36:24And very happy that they are also capturing
36:27large variances of those indices.
36:29Those reference standard indices
36:32of severity and the goal here is
36:36never to replace the clinician.
36:38The goal here is to free the
36:40clinician to spend more time with the
36:42parents so that we can do what the
36:45diagnostic process is supposed to be.
36:47Is only supposed to be a window to treat.
36:50So we need to make sure that clinicians have
36:52more time to spend time counseling a family,
36:55supporting the family,
36:56and most importantly,
36:57doing the care coordination so that
37:00we can translate diagnostic results.
37:02Into actual treatment.
37:04So the good news about this procedure
37:08is that this was done under FDA
37:13sort of rules for past 6-7 years
37:16and they cleared this device as a
37:19clinical tool since June of last year.
37:23So this is now a clinical tool which is very,
37:27very, very exciting.
37:29The idea is to support a public
37:31health system that does not have
37:33enough extra clinicians.
37:35The idea is to deploy tools that are
37:38cost effective and will increase access,
37:40particularly for those families
37:42that are marginalized,
37:44particularly low income minority
37:46and rural families.
37:48Autism.
37:48Well, it's it's the same in most healthcare,
37:52but in autism in particular,
37:54the healthcare disparities are stark.
37:57If you are a black child
37:59with autism in this country,
38:00you are at double the risk of
38:03intellectual disability than a white
38:04child with autism in this country.
38:06And that is so not because, believe me,
38:09we have a whole grant with our our,
38:12our colleagues at UCLA and Wash
38:14U and what not.
38:16It's not the genetics,
38:17it's really the access is those
38:19early experiences that those
38:21families are being the pride of.
38:24So we need to solve this problem.
38:26And with early identification
38:28leading to effective early treatment,
38:31we can change lifetime outcomes.
38:33So the good news also is that
38:35this was the size of, you know,
38:38is half a fridge okay.
38:40This was not something that could
38:42be deployed, you know, very broadly.
38:44So this is what it looks now.
38:49Oops. So this is what it looks now.
38:51So we were able to miniaturize that.
38:53So that's now in a tablet version.
38:56So there is a built in eye tracker and
39:00that tablet can be deployed anywhere
39:03there is Internet connectivity because
39:06the technician can sit anywhere and
39:08that can actually be shipped by a UPS.
39:11So very happy with the way
39:13that things are going.
39:14Now we're very interested in,
39:17in in tailoring those
39:18devices to context of use.
39:20So for those of you who sort of you
39:23know like boring stuff like positive
39:25predictive value and negative
39:27predictive value and Needless to say,
39:29the moment that you sort of
39:31maximize negative predictive value,
39:33you sort of minimize positive
39:35predictive value and vice versa.
39:37And this is basically where our device is,
39:39is right here.
39:41So for the context of a
39:44clinical context of use,
39:46just like the clinical trial that we did,
39:48these are all specialized clinics.
39:50For that then we have that, OK,
39:53sensitivity and specificity are both
39:56optimized and what the results is
39:58that we can actually make statements,
40:01very precise statements of
40:03probabilistic statements of a child
40:06having autism or not having autism,
40:08OK.
40:10One thing that was a serendipitous
40:12kind of a thing we didn't expect
40:14that to happen is that in the sample
40:17for the national clinical trial what
40:19we found out is that if you go by
40:22clinician reference standard diagnosis,
40:24we found that they were both ethnic
40:27and racial bias as you can see here
40:32when we went by the by the device.
40:35Diagnostic classifier.
40:36Both that racial and and and
40:40and ethnicity bias disappeared.
40:43We don't know what that is due to,
40:45but this is something that we
40:48certainly are going to to to focus on.
40:50But let me say something about the
40:52possibility of using this device in the
40:54context of population based screening.
40:56There pediatricians tell us all
40:58the time they don't want to tell a
41:00family that your child has autism,
41:02they want to tell a family your
41:03child does not have autism.
41:05They that's what they want to say and so
41:07you maximize negative predictive value.
41:10So where we to use this device in the
41:13context of population based screening,
41:15we might be able to make those kinds
41:18of of of of statements of probability
41:20including to be able to get to not
41:23over 99% of probability reassuring
41:26the pediatrician you can say to the
41:28family your child does not have autism
41:30basically come out of the system.
41:32Let's focus on those who need our attention.
41:35OK,
41:36very quickly now we we have an NIH
41:39grant now that basically downwards
41:43extends the utilization of this device.
41:47We are pairing with our college pediatricians
41:51pediatric practices at nine month old,
41:54nine month well well baby checkups.
41:57And so the baby goes and finishes
41:59the nine month well baby check
42:00up and then steps into our van
42:02and completes this procedure.
42:04We are now using it truly as
42:08a quantitative population
42:10based screening device.
42:14I want to say one final thing
42:16about something that I've
42:17learned here which is you know,
42:19Sally Province is to say.
42:21Go and spend one year in that daycare,
42:25the hospital daycare,
42:27sit in the corner and do nothing.
42:31Observe, right.
42:33So this is what we've been doing.
42:35This is McDonald and Jim and and
42:39and all of you have taught me is.
42:43You want to be a walking
42:45laboratory of social engagement.
42:47You want to be somebody who
42:48can actually sense how much
42:50children are getting from this.
42:52But can we quantify,
42:53Donald is the one who said,
42:54I mean, this whole business of,
42:56you know, you're a good clinician,
42:57but can you elevate it to
42:59the plane of science?
43:01And Donald is the one who's sort of
43:03prompting us to do this kind of thing.
43:05So here is what life is like, K.
43:09Now this is what typically
43:12developing toddlers are looking at.
43:14This is what a child devotism is looking at.
43:18Let's look now at pointing,
43:19social monitoring,
43:20joint detention being so important.
43:22OK, so this is what the typically
43:24developing children are looking at.
43:26And this is what one single
43:28child devotism is looking at.
43:29And here is now facial affect and
43:32you see where most children are
43:34looking at and you see what the
43:36child devotism is looking at.
43:38And here is another one facial effect.
43:42Can you see and this is where
43:44the child devotism is looking at
43:46and now we can basically there
43:48are hundreds and hundreds of
43:50those examples in our videos.
43:52And So what we're trying to do now is
43:55to derive indices that are skill based,
43:58something that we can actually
44:00give to the interventions and say,
44:03well this is where the child is.
44:05In terms of processing social affect
44:08or facial gestures or or or or or
44:11joint attention or or anything of that.
44:14So one of the things that we're
44:16trying to do now is to derive this
44:19quantitative indices of skill based
44:21milestones for treatment because I
44:23mean you guys know that one of the
44:26things that people who work if a BA
44:28and what not in early intervention.
44:31And treatment is that they spend
44:33a lot of time collecting a lot
44:35of data that is not standardized
44:37that drives payers crazy and takes
44:40hours upon hours upon hours.
44:42And so they need to document
44:43progress or not of the child.
44:45This is time that they are
44:47not providing services.
44:48So that's what we're trying to do as well.
44:50We are trying to basically advise
44:54develop those quantitative
44:55indices of those important skills.
45:00You know, what I've learned in the past
45:0320-30 years in fact working in this
45:07field is that autism is not a disease.
45:10Autism is not a disease.
45:11Autism is a trait.
45:12Autism is a genetic trait.
45:15Autism is something that may
45:16or may not lead to disability.
45:18And whether it does so depends
45:22very much on us if we want to know
45:25where those 95,000 babies that are
45:27born every year who have autism.
45:30Where are they going to end up?
45:32One of the largest treatment programs
45:34that we have at our center is treatment
45:36for severe behavior challenges.
45:38Is a self injury, is a lovement, is.
45:40Is, Is aggression is,
45:42I mean you name it,
45:43K We would like to put that program out
45:47of business because what I've learned
45:49is that the greatest burdens of autism,
45:51the intellectual disability,
45:53the languages ability,
45:54and the severe behavior challenges are not
45:56part of the definition of this condition.
45:58They are not inevitable.
46:00They are the results,
46:03particularly for some children
46:04whose access to any supports come
46:07only during their school years.
46:09So that's what we want and ultimately
46:12we would like to change the narrative
46:14of this condition from one of disability
46:17to one of possibility and promise.
46:21These are all the different foundations
46:23that have been supporting over the years,
46:25the wonderful people who have been
46:28working in all those projects.
46:30Doctor Martin,
46:42I assume that if people would like to
46:45ask questions, I'm here. If we have you
46:47assume correctly and if you are
46:49on Zoom, please let us know.
46:50And if you're in the crowd,
46:52hold on a second.
46:54Doctor Crowley is got it started.
47:02Doctor Clan brings back memories
47:05and it was a great pleasure to see
47:07you head out of the park again.
47:10So I'm going to save some of
47:12my questions for later today.
47:14But my specific question
47:15is the Ados is a gold
47:17standard for assessment of autism and
47:20you're presenting these assessments to
47:23predict to that and I think you said.
47:2568% of the variance was accounted
47:27for and was that right and what
47:28I was thinking little over
47:29that, little over that,
47:30what was it again, what was the
47:33747474? But what I was
47:34thinking was what if the autism
47:36diagnosis is actually or that that
47:39measure is actually imperfect and
47:41your measure actually accounts
47:43for variability in impairment and
47:45functioning and long term things And
47:46do you have any data to speak to that?
47:48So Michael, I will have you speak of the FDA.
47:52Whenever you are developing any
47:55medical device, any medical procedure,
47:57but certainly a medical device procedure,
48:00you need to emulate a reference standard.
48:03You cannot be better than your reference
48:06standard because you're going to be wrong.
48:08You understand what I mean?
48:10So for A, it doesn't matter if
48:12you're working in oncology or or
48:15in diabetes or in in cardiology,
48:17if you are trying to develop a new procedure,
48:20you need to match the reference standard and
48:23in this case the AIDOS is gold standard.
48:26One thing that we never,
48:27we never discusses the fact that the
48:30AIDOS was never approved by the FDA.
48:31You understand that the sensitivity
48:33and specificity of the AIDOS.
48:35Was actually calculated by by
48:38basically looking at the concordance
48:40between the clinician's diagnosis
48:42and the ADAS when the clinician used
48:45the ADAS to make the diagnosis.
48:47So I know it's a that's one thing
48:49that's that you need to consider.
48:50And the second thing is that the ADAS
48:54is not perfect and so there is error
48:57measurement actually that it's quite,
48:59it's quite a bit and yet FDA
49:02disregards this entirely.
49:04ADOS is the gold standard.
49:06You better get close to it.
49:08The the funny part and Kasha would be
49:11able to speak that to that more eloquently
49:13than I in our national clinical trial.
49:17You understand that 29.5%
49:21of all of our sample.
49:23Remember this was a very complicated sample.
49:26These are not kids with without diagnosis.
49:29They have autism or non autistic
49:31developmental delays of all kinds.
49:3329.5% of the cases.
49:34The clinicians from UCSF,
49:36from Seattle Children's,
49:37Cincinnati Children's, Rush,
49:39Emory and and and and and Phoenix had
49:43suboptimal confidence in their diagnosis.
49:46These are 16 to 30 month oldest.
49:48FDA says OK, no,
49:51it should be 100% you should.
49:53You basically they assume the truth,
49:56but the ground truth in our field.
49:58I mean this is a behaviorally
50:00defined condition, so you have to.
50:02Deal with those imperfections.
50:03But absolutely I my heart is
50:06goes of your statements,
50:07but if you want to get it cleared by FDA,
50:11they wouldn't listen to what you just said.
50:17Adam.
50:20Hey Yami, that
50:20was amazing as always.
50:24A lot of questions,
50:24but I'll stick with one,
50:25like very concrete.
50:27Can you talk a little bit about the
50:30curation process for the videos,
50:32for the standardized test,
50:34the video stimuli? Yeah, But how
50:37you selected which ones and
50:38what you were looking for?
50:40Well, we had several generations of videos.
50:45I remembered that,
50:46sort of the one that we used in
50:49all of our experimental studies.
50:51The behavior, genetic studies,
50:53the early developments,
50:55they they were either caregivers and
50:57we remember sort of a lot of actresses
51:00coming to the child study center
51:02sometimes from Calvin Hill that you
51:04come here and they record our videos.
51:06The toddlers came from a preschool actually
51:12an after school program at Guilford Guilford,
51:14Guilford Green.
51:16You probably remember that family as well.
51:19They came from that,
51:21but subsequently we became to enrich
51:23those particularly in terms of Guilford
51:26is not very diverse and so we needed
51:28a little more of diversity in our
51:30videos and things of that nature.
51:31But the funny part and if this is what
51:34you're driving at is that we have a very,
51:37very large program in our center focus
51:40entirely on the Latino population,
51:42particularly non non-english
51:44speaking children and families.
51:47And we also have large grants focused
51:49entirely on African Americans.
51:51I remember Atlanta is 40%,
51:52forty, 7% African American.
51:55So we we have the only NIH genetic
51:59grant genetics grants focused
52:01entirely on African Americans.
52:03And what we're finding out is that,
52:05number one,
52:06those essays are not biased by
52:09either language or race, ethnicity.
52:12And I think the idea is that
52:14probably we're tapping into assays.
52:17That the language that in the
52:18stimuli are not very important,
52:20that are things that are
52:22a little more universal,
52:23but that's absolutely something
52:24that we're going to be working on.
52:26Were there also like certain
52:28kinds of sorry, were there also like
52:32specific kinds of social interactions
52:33that you're hoping to capture?
52:37To be truthful, no.
52:40We started by basically filming children.
52:43Well, we have caregivers.
52:45Directed infant directed speech,
52:46we have a whole program of
52:48research focused on live.
52:50Live tracking for example
52:52is bidirectional tracking is
52:55caregiver infant interaction live.
52:58And so that's live of course,
53:00but in terms of the videotapes that we use.
53:03They were basically trying to get as
53:07varied as varied as sampling of of
53:11toddler interactions as we could.
53:13It has to be developmentally
53:15appropriate of course.
53:15But what I remember one thing
53:18that I think Kyle once said we
53:20were sort of being trained on the
53:22on the Gazelle schedules and it
53:23was like gosh this looks so much,
53:25you know that how interactive the
53:27psychologist he looks so much like
53:29this test and that test and I think
53:31you said something the effect.
53:32Yeah, come on.
53:33Toddlers don't do that much right?
53:36So you're going to have a lot
53:37of commonalities.
53:38We try to grab as much as possible with that.
53:41Doctor Pruitt thinks
53:42toddlers do as much or more
53:47toddlers. They're the best.
53:48It's the last time.
53:49The mind and body are good friends.
53:52That's why we love them so much.
53:55After that, the mind takes over
53:56and it's downhill forever.
53:59And we cannot free ourselves from our
54:01mind the rest of our lives. Kyle,
54:03I ami you.
54:07I've been I've been wondering
54:08for years how we let you go.
54:10Now I realize why you had to,
54:12so that you could do this.
54:13And this is such an amazing gift
54:16to clinicians particularly.
54:19The burden that you lift from
54:21Pediatrics to say so many parents
54:23worried about their children,
54:24there's something off.
54:25What is it? Are they autistic?
54:27To be able to say no liberates
54:29both the physician and the family
54:31and lets them get on their path
54:33so the pediatrician can pay
54:34attention to the ones that need.
54:35It's it's an amazing gift.
54:37I'm struck by one of the small little
54:40quotes in your side that said you
54:42know that these these children create
54:45their own experience and to the parents.
54:48They're not part of it and that's
54:51a very bitter pill to swallow,
54:53to treat it as a trait and to
54:55say we can help you with that,
54:57let's get started and let's get started
55:00together is such an immensely repairing,
55:03phenomenal interpersonal experience
55:05between the clinician and a parent,
55:08as opposed to waiting for 3-4 five years.
55:12The kid you know explodes
55:14kindergarten and say, wow,
55:15there's something really
55:16wrong with this child.
55:18To have eliminated,
55:19you know,
55:20hundreds of thousands of negative
55:23experiences from ever happening with
55:26your wonderful toy is quite a gift.
55:29And it's not just exciting,
55:32it's a real game changer.
55:33And I'm thinking of the institutions
55:35that need to know about this
55:37yesterday so that it can wind up in
55:39the hands of people in the country,
55:42and I mean in the country.
55:44Wi-Fi access remains in the issue.
55:47They'll have to get busy with
55:49their legislators to get it.
55:50But this is worth the this is
55:53worth the candle. And I thank you.
55:56And I remember Sally saying,
55:58we we do not understand autism.
56:00It starts so early.
56:01We are completely at the wrong
56:03end of this dog. We have start.
56:05We have to start working with the head,
56:07not the tail. And you did that.
56:09Thank you.
56:10Kyle.
56:11I must say this. I know I am in New Haven.
56:14I know that I yell when I receive an e-mail.
56:17From a former patient
56:19whom I saw for 15 years,
56:21every Tuesday that person was first seen.
56:25He's my age. He was served.
56:27First seen at the Health Child Study Center
56:30at the age of 20 months by Sally and by Sam.
56:35And then Donald was involved.
56:37I assume Jim must have involved.
56:39I don't know if Bob was involved.
56:41Eventually he ended up with me.
56:44And he wrote to me the most heart warming.
56:49I I'm talking about Tom and
56:53what a distinguished family.
56:55Yale family. But my goodness me.
56:57But what I'm saying here is this.
57:00This is just a toy.
57:02The concepts.
57:04The concepts came from you.
57:06It came from Linda, it came from Jim,
57:08it came from Donald, it came from Fred.
57:10And I mean this center to me is
57:14really a monument to the impact
57:18of relationships in one's life.
57:21When, when, when, when?
57:23My my good friend Dr.
57:25Martin was editing that book.
57:27That book, Donald's book,
57:29was all about relationships,
57:31life being a correct collection thereof,
57:34to my mind, Brain is.
57:36Exactly.
57:36That is the instantiation
57:38of those experiences.
57:39But let me speak just very briefly
57:41substantively to one thing that you said,
57:44yes, they are creating that
57:46experience and their parents,
57:47their children were born with an
57:49attenuated sense of the other.
57:51So the parents are not what
57:53they are cocreating.
57:54Having said that,
57:56we can engineer that environment to
57:58empower families so that we can strengthen
58:01them mutuality that that that sense of of.
58:06Typical connection and and
58:08and and and engagement.
58:10But here is what we found out because
58:13you work with that population,
58:15you know with teen moms and very,
58:18very,
58:19very,
58:19very low income mothers were
58:22extraordinarily fragile.
58:23They leaving situations of stress and
58:25gales when you are telling them that's
58:28something that they can't actually do,
58:30meaning they can engage that baby.
58:33And that engagement will promote
58:36optimal development in that child.
58:39The sense of empowerment
58:41that one conveys is huge.
58:43So nothing that I said today is new
58:47because it came from all of you.
58:49Thank you.
58:50Is that the final word?
58:51That's such a beautiful final word,
58:53but there may be a final, final word.
58:54There's two questions.
58:56There's more questions that I think we're,
58:58I think we're going to be
59:00respectful of your time and ever.
59:01And I mean you know your your
59:03love note to to the center.
59:05We feel it and we send it back and
59:07come back again and again and again.
59:09Thank you.
59:10Thank you very much.