5th Congress Autism-Europe Articulos / Proceeding Autism-Spain An Epidemiological study of Autism at 18 months of age John Swettenham Department of Psychology Goldsmiths College University of London New Cross, London SE 14 6NW INTRODUCTION This project is a prospective epidemiological study of Autism and Pervasive Developmental Disorders. Funded by the MRC, it involves researchers from four centres in London. Simon Baron-Cohen at Cambridge University; Tony Cox, Gillian Baird, Auriol Drew, Kate Morgan and Natasha Nightingale at Guys Hospital; and Tony Charman at University College London and John Swettenhafn at Goldsmiths. In addition there are also 300 Health Visitors and 30 GPs assisting with the data gathering. So what issue could possibly bring so many people together on one project? Well, autism is widely regarded to be the most severe psychiatric disorder in childhood, and our main concern is that autism is currently diagnosed relatively late - rarely before 3 years of age -and this is despite the fact that there is a consensus that the disorder almost always has prenatal onset. There are several possible explanations for this late diagnosis. Firstly GP's and Health Visitors (HV's) are not specially trained to recognise autism early. Secondly, there is currently no screening instrument for autism in the current routine HV check. At present HV's only screen for motor, intellectual or perceptive development. Thirdly, it is a rare disorder so that even if a GP does see a child with autism in his or her practice it may be several months before any one comes along. And finally, the deficits are quite subtle - so that autism can often be mistaken for other disorders. It is difficult to assess abnormalities in social and communicative development in the pre-school period Our study is attempting to diagnose autism early. In order to give you a general picture, here is a overview. Screening takes place in two stages. Stage one began on April 1st 1992 and continued for exactly a year. We used an invent called the CHAT (which will be describes in a moment) to screen a large sample of 18 month old children. The idea was that the CHAT instrument could be used by HV's or GP's during their routine 18 month check. We now have data from this instrument for 16 000 children screened in 9 Districts around the South of England The second phase is now well underway, and involves a second invent which we've called the CD which will be used to screen the same 16 000 children again during their next routine HV check at around 3 and a half years of age. This will take place through 1994. So what are the main aims of the study? The first aim is to lower the age of diagnosis of autism to the 2nd year of life. This have consequences for families, who would not have to wait so long for diagnosis. Earlier diagnosis also of course means the possibility of early intervention. Although we cant he sure of what appropriate intervention would consist of at the moment, it is a step in the right direction at least to be able to identify autism early. Finally, toddlers with autism would be of considerable research interest in that this may help us to understand more about the early development of social understanding. The second aim is to test the reliability, the validity and the sensitivity of the CHAT instrument. In other words, is there a difference between test and retest? or between HV and GP use? Is the CHAT pickling out what we want? Is it clinically valid? And what is the hit rate like? Are we making too many false hits?. Is our predicted "high risk for autism" profile picking out too many children? The third aim is to identify what patterns of abnormality predict Autism at 18 months. and to assess how well particular psychological test indicators discriminate children who receive a diagnosis of autism from other forms of developmental delay. Finally in this section it is worth mentioning that the basic intend of this method is that early detection of autism may now he possible because results from experimental psychology have shown us what to look for in toddlers if we want to detect autism early. The CHAT Form Our screening content (shown in Table l.) is called the CHAT, which stands for "Checklist for autism in toddlers." It consists of 2 sections, one for parents' answers and one for HV«s observations. Table 1: The CHAT (Reproduced from Baron-Cohen et al (1992).British Journal of Psychiatry.) (Ed.Note: Not available here) The instrument has been designed so that it is quite difficult to fail. In a pilot study using the CHAT with a group of normal children visiting a health clinic more than 80% passed all the tests. We have been careful to ask HV«s to avoid raising anxiety - and to point out that this is a research project so far at this stage we don't know the significance of a fail on these items. Also, on a practical level, it is crucially important that the CHAT is easy and quick to administer - given the time constraints on HV«s and GPs. Considering the CHAT itself Each question refers to different aspects of development. First lets look just at section A. Question 1: "Does your child enjoy being swung or bounced on your knee?"- looks specifically at one type of play, "Rough and Tumble play." Question 2: "Does your child take an interest in other children?" - looks at social interest, so we would predict that this might be abnormal in a child with autism Question 3: "Does your child like climbing on things, such as upstairs?' looks at motor development. Question 4: "Does your child like playing Peek-a-boo and hide and seek?" looks at social play. Here again we might predict that children with autism would fail those item and be reluctant to engage in this type of social interaction. Question 5: Considers the child's ability to pretend. There is evidence of abnormality in this area in autism - by pretending we mean using objects as if they have other properties or identities Question 6: looks at pointing to ask. This is an important preverbal communicative gesture, where a child gets another person to do something by pointing. Some autistic children do show this behaviour (ie pointing for non-social purposes) Question 7: looks at the ability to point for interest (ie use the proto-declarative gesture) which has been shown to be abnormal in autism. This is an important joint attention gesture where the child makes assertions about objects (ie Indicating an object to a person as an end in itself). So, we may find that the 18 month olds fail to point for interest but pass pointing to ask. In this way, the CHAT tests whether relates forms are predictive. Similarly the distinction between pretend play, functional play and rough and tumble play can be made. Question 8. looks at functional play ability Question 9. looks at Joint attention and the Showing gesture, which may be absent in autism. These items are not currently screened, and yet both pretend play and joint attention behaviours are universal milestones of development at 18 months. Our hope is that the absence of these behaviours could be clear, specific indicators of autism or related disorders. The final point to make about section A is that items were arranged in order to control for YES/NO bias, in other words, they were arranged so that the parents would he able to say YES without having to produce a long string of NO's Now lets turn our attention to section B, this is the observation section to be filled in by HV«s or GPs. The first point to make is that some of these items correspond to items in section A. For example Biii is a question about pretending, where the HV attempts to get the child to display some pretend play using either a tea set or toys of their own choice. This corresponds to the parent question concerning pretending. Biv check the child's ability to produce a protodeclarative gesture. So this reliability check in section B is a useful one as it checks whether parents are overestimating or underestimating their child's performance Also in section B, a question concerning the use of eye contact, thought to be abnormal in autism. And finally, a rough indication of general development from whether the child can build a tower of bricks High Risk Study The CHAT has been check in an earlier small scale study with a small sample of 41 high risk toddlers (who were siblings of children with autism). It was found that the children who failed pretending (items A5 (validated by Biii) who failed pointing for interest (item A7 (validated by Biv) and who failed gaze monitoring (item Bii) turned out to be autistic. So our prediction based on the small scale study was that these 3 behaviours would turn out to be the crucial psychological risk indicators for autism. The present study though is clearly on a far greater scale involving 9 District in the South east of England. Risk Groups We predicted 2 risk groups. Firstly, arising from the pilot study with siblings of autistic children, we predicted that children who failed three items may have autism ("protodeclarative pointing, "Gaze monitoring" and "pretending" (validated in section B). That is, CHAT items A5, A7, Bii, Biii, and Biv). Children with this profile enter the Autism Risk Group (ARG) Our second predicted group consists of those children who are failing protodeclarative pointing (PDP) (A7 and Biv), or PDP and pretend play (A5, A7, Biii, and Biv). Critically, children in this group had to pass Gaze Monitoring (Bii). We predicted that these children may be at risk for Specific Language or General Developmental delay (without autism). This we call the Delay Risk Group (DRG) Our Third group consists of those children who pass protodeclarative pointing, gaze monitoring and pretend play. We predict this to be a Normal Group (NG). Any child who showed either risk group profile at 18m was given a second CHAT either by phone or home visit - this usually took place about one month later Like most screening tests in public health surveillance, a positive case is identified if a child consistently fails on an initial test and on a subsequent retest. Our interest was only in those children who were consistently failing key items since this was likely to be due to significant developmental causes more than situational causes (such as the child's current physical state), or very mild developmental delay. We have now finished phase 1 of the project end collected data for the 16000 children at 18 months of age. Here are some of the findings so far. Results Table 2. shows the numbers of children, from our sample of 16 000, who showed these pattern of behaviour. Only 12 children out of 16 000 are in our autism risk group. Table 2: Number of children falling into the 3 Risk Groups, in the total population (n=16000). Risk Group N Autism Risk Group 12 Developmental Delay Risk Group 44* Normal Group 15,944* Autism Risk Group: Failed A5, A7, Bii, Biii and BIV. Developmental Delay Risk Group: Passed Bii, but either failed A7 and Biv, or failed A7 Biv A5, and Biii Normal Group: Passed all of A5, A7, Bii Biii, and Biv. *=Estimate based on retesting approximately 20 cases in this Group at the initial CHAT So, what were the children in the Risk Groups like? There were 12 children in our "High Risk for Autism" group, all of whom we saw at our Clinic at Guy's in London. In addition we saw 22 children from the Delay Risk Group and 16 children from the Normal Group. They were each assigned to one of 3 diagnostic groups. Definition of Statistic Group We used 5 very experienced clinicians, who worked independently of one another and who were blind to the risk groups the children were in. Autism Children who met criteria on at least 2 out of 3 diagnostic methods: o The Revised ADI o ICD-10 criteria, from interview with the child o ICD-10 criteria from as rated from videotapes of all 50 children Developmental Delay Children with less than 5 words and/or more than 4 months delay on the Griffiths assessment Normal Group Children who fitted into none of the above categories were assigned to the normal group. It is worth noting that children with severe developmental delay were not included in the population, since such children are already clearly identified by 18 months of age,. Our aim was to assess if autism could be distinguished from mild or moderate forms of developmental delay (without autism). Table 3. shows the diagnosis of the children in each of the risk groups. First let's consider the diagnosis of the children in the Autism Risk group. 10 out the 12 children in this group received a diagnosis of autism. The 2 remaining children were thought to have Developmental Delay at 18 months. However, now we have seen these children again at 3 years of age we strongly suspect that even these 2 children are also in the autism spectrum , It's worth mentioning that the 10 children presumed to have autism at 18 months have now had diagnosis confirmed at 3 and a half years of age. Table 3: Number of children failing all 3 key items, or only 1 or 2 of the key items, who given different diagnoses. (Risk Groups) Number of Key Items Failed Diagnosis Autism Developmental Delay Normal (n = 10)(n = 17) (n = 23) Fail all 3*: (n = 12) 10 2 0 Fail 1 or 2**: (n = 22) 0 15 7 Fail 0: (n = 16) 0 0 16 *=Failed Gaze Monitoring; Protodeclarative Pointing; and Pretend Play **=Failed Protodeclarative Pointing; or Protodeclarative Pointing and Pretend Play Of the 22 children in the delay risk group, none received a diagnosis of autism, 15 were thought to be developmentally delayed. So less than half the children in the Developmental Delay Risk group were diagnosed as Normal. Finally, of the 16 children in the normal control group, none were considered to have any problems. CHAT profile of children with autism at 18 months Table 4. shows how the children with autism were doing on all the CHAT items (i.e. the percentage of children out of 10 in the autism group passing each item). Table 4: Percentage of subjects consistently passing each item on the CHAT, by Diagnosis. Autism Developmental Normal item Delay (n =10) (n = 17) (n = 23) Al 100 100 100 A2 50 94 91 A3 100 100 100 A4 90 100 96 A5(PP) 0* 53 100 A6 10 53 87 A7(PDP) 0* 0 70 A8 100 100 100 A9 60 65 91 Bi 70 100 100 Bii(GM) 0* 65 100 Biii(PP) 0* 65 100 BIV(PDP) 0* 0 70 Bv 90 82 91 * = This unique pattern identified 10 true positive cases of autism in 16,000 and 2 false positive cases of autism (they actually had developmental delay without autism) As can be seen, all the children failed our key predicted items (A5, A7, Bii, Biii, Biv), 9 out of 10 children also failed "pointing to ask" (A6), and 5 out of 10 failed "showing an interest in other children" (A2) (nb. these are parents reports only, not validated in section B). It is also of interest that 7 out of the 10 children did pass Bi "Does the child make eye contact with YOU?" Concluding remarks So the take home message (at least at this stage of the project) is: "Falling gaze monitoring, protodeclarative pointing and pretending (reported by parent, validated by HV and followed up 12 months later) carries a high risk for autism at 18 months." As you will remember from the overview the entire 16 000 are currently being re screened using the CD at 34 years of age. PHASE 2 should provide us with a more complete picture. In fact this will be essential in order to establish the rate of false negatives and the accuracy of the diagnostic predictions. The idea of phase 2 is to collect diagnostic information from multiple sources (Paediatricians, Speech Therapists, Health Centres and from the CD). Like the CHAT the CD contains a section for HV to fill in themselves and a a section with questions for parents that might alert us to children with Autism or PDD All that remains now is for us to wait for the CD results and wonder. What has happened to the children we have seen in our Risk Groups? How many children with autism did we miss? What is the CHAT profile of the entire group of children with autism in this sample?