Notes on the prevalence of autism spectrum disorders

(Paper originally published for the autism99 Internet conference. The site for this conference is no longer available)

The National Autistic Society is often asked "How many people are there with an autism spectrum disorder?". Trying to answer this is fraught with difficulty. There is no central recording by the UK Government and, indeed, very few epidemiological studies on which to make informed predictions. The available studies into this population are infrequent, expensive and problematic.

The problems arise from three main causes.

1. There are many difficulties in diagnosing people with these disorders. There is no medical test that can determine whether or not a person has autism. Diagnostic criteria are in terms of descriptions of behaviour. The earliest criteria, suggested by Leo Kanner, were very narrow. The current standard classification systems (ICD-10 and DSM-IV) are much wider, even for the sub-group of 'childhood autism' (or 'autistic disorder' in DSM-IV). Professionals differ in the way they apply the criteria, even if they are, theoretically, using one of the standard systems. Diagnoses may be recorded in different ways in case notes and centralised data collections.
2. Diagnostic terms tend to be used in different ways. Sometimes the term 'autism' is used to mean Kanner's original group, sometimes it refers to the wider group called 'childhood autism' in ICD 10 and sometimes the whole autistic spectrum, including the individuals described by Asperger. In any case, there is a very great deal of overlap among all the sub-groups named in ICD-10 and DSM-IV and many individuals fit more than one diagnosis within the spectrum.
3. In epidemiological studies of prevalence, case finding methods vary. Those that involve seeing, assessing and diagnosing every individual in the sample to be examined will tend to find higher numbers than studies that rely on using case notes of individuals who have already been given the diagnosis in local clinics.

For these reasons, it is very difficult to make comparisons among studies done by different workers, at different times, in different places, using different definitions, and different methods of case finding and examination. Thus, the earliest epidemiological studies used Kanner's very narrow criteria and found the often quoted prevalence rate of 4 to 5 in 10,000 children. Later studies have used wider criteria, so it would be inappropriate to calculate a mean prevalence based on results from the earliest and the later studies.

Knowledge about prevalence is, however, vital if effective services are to be planned and provided at the correct points in individuals lives. The expense of making appropriate provision for those people within this population makes it the more surprising that governments have not been willing to fund new epidemiological research, covering individuals of all levels of ability, using soundly based scientific methods, which would enable accurate planning.

Until such research has been carried out and published, the NAS response to the question is based on two epidemiological studies of prevalence; the first by Lorna Wing and Judith Gould involving a cohort of children born between 1956 and 1970 in the former London Borough of Camberwell. (Wing and Gould 1979) This looked at children virtually all of whom had IQ levels below 70. The second study focused on the children with IQs of 70 or above and was conducted in Gothenburg. (Ehlers and Gillberg 1993) This study looked for Asperger syndrome and high-functioning autism, and is the only one concerned with these sub-groups known to the present authors. These studies are quoted, first, because they were very intensive and, second, because, unlike most other studies, they covered the whole spectrum of autistic disorders. The researchers screened a defined geographical population of children and then examined and diagnosed all the children who passed the initial screening.

The NAS is not aware of any published studies of the incidenceof autistic disorders. (Incidence is the number of new cases occurring in a specified population in a specified time, such as one year). Nor is it aware of any studies of prevalence that have covered autism and Asperger syndrome and pervasive developmental disorder not otherwise specified (PDD-NOS) in individuals of all levels of IQ. (Prevalence is the number of cases in existence in a defined population at any one time). In order to give an estimate of overall prevalence, the NAS combined the rates from Camberwell (20 per 10,000) with those from Gothenberg (35 per 10,000 or 71 per 10,000 if all those with social impairment were included). (See later for more details of these studies).

Many clinicians feel they are now seeing many more children with autistic disorders than in the past. Some recent small scale but intensive studies have shown high prevalence rates for autism - for example, Arvidsson et al (1997) found a rate in 3-6 year old children for ICD-10 childhood autism plus autistic-like disorders of 31 per 10,000. This did not include Asperger's syndrome. The question is frequently asked "Is the prevalence of autistic disorders rising?" Interesting evidence on this question comes from the following report from California.

The California Health and Human Services Agency (California 1999) has examined the numbers of people with autistic spectrum disorders entering the California Developmental Services system each year from 1987 to 1998. The figures have been published in a most interesting, detailed and balanced report. The diagnoses have been made using DSM-IV criteria. There has been a rise in each of the years studied. The author points out that it would not be appropriate to use the figures for calculating prevalence. However, on pages 13-14 they make a provisional calculation of the rates of new intakes based on the annual live birth rate. From this, they draw the conclusion that the number of new intakes has, over the last few years, exceeded the annual estimate of persons likely to be newly diagnosed with autism spectrum disorders, if published prevalence rates are correct.

As noted in the report, these conclusions should be considered with caution, since the estimated annual rates for California are based on the annual number of live births as the denominator, whereas the numerator is the number of new cases diagnosed each year. These include a wide range of ages, presumably from about 18 months upwards.

It is relevant and interesting to examine more deeply the calculations in the California report of the numbers of new cases to be expected each year. The highest published prevalence rate for autistic spectrum disorders in individuals of all levels of ability among those quoted in the California report was 20 per 10,000, but no reference was given. A search of the epidemiological literature found only one possible source for this figure. This was the study by Wing and Gould, mentioned above, published in 1979. As noted previously, they found a prevalence of 20 per 10,000 in children with IQs less than 70, who had been born in the years from 1956 to 1970 inclusive . The Gothenburg study, by Ehlers and Gillberg (1993), which examined those with IQs of 70 or above and looked for Asperger syndrome (diagnosed using Gillberg's criteria, not DSM IV criteria) and high-functioning autism, found a prevalence of 36 per 10,000 in children born 1975 to 1983. The authors also found an additional 35 per 10,000 with social impairment of the autistic spectrum type, but not all the criteria for Asperger's syndrome or high functioning autism. Adding figures from both studies gives a prevalence for all IQ levels for all autistic spectrum disorders of 56 per 10,000, or even, if all those with social impairment of the autistic spectrum type are added, a total prevalence of 91 per 10,000.

The figures from the California report can be compared with those from the prevalence studies quoted above if certain assumptions are made, as follows.

1. The number of new entries to the California developmental services diagnosed as DSM IV autistic disorder was 1,685. This did not include other PDD (Asperger's syndrome etc.) so did not cover the whole spectrum. From personal communication with the author of the report, it appears that the number of new 'other PDD' individuals was, at the most, 189. This gives a total of 1874 for all autistic spectrum disorders. The numbers of live births in California in 1998 was estimated to be 526,501.
2. On page 11 of the California report, it is noted that 42% of those with autistic disorders known to the developmental services functioned above the level of mental retardation. This information was used to estimate the level of ability of the new entries to the services,

Using the method of calculation in the California report, the figures for new cases entering the Developmental Services in California in 1998, and the estimate of the proportions of individuals with IQ <70 and IQ 70+, the rates for new entries with all autistic spectrum disorders in1998 in California are as follows:

• IQ <70 = 20.6 per 10,000
• IQ 70+ = 14.9 per 10,000

The above rate for those with IQ <70 is close to that found in Camberwell in children born between 1956 and 1970 (that is 20 per 10,000). The rate for those with IQ 70+ is less than half that found in Gothenburg for children born 1975 to 1983 (35 per 10,000 for Asperger syndrome but 71 for the whole spectrum). The California report states that it is the number of individuals without mental retardation (IQ 70+) that has risen most steeply. However, from the above, it seems there is still some way to go to reach the Swedish rates.

These calculations are fraught with problems and are epidemiologically unsound. Furthermore they do not include any estimate for children with language and communication disorders who have not yet been diagnosed. The conclusions depend upon the accuracy of the results of two epidemiological studies on small populations. Nevertheless, the calculations do not disprove the possibility that the reported increase is due to a widening of the definitions of autistic disorders plus improved case recognition, rather than a true increase in incidence. There has been a marked increase in awareness of autistic spectrum disorders in recent years so it is not surprising that cases are now being diagnosed that previously would have been missed. In the Camberwell study conducted in the 1970s, only about 8 of the 74 children identified by the researchers as having an autistic spectrum disorder had previously been diagnosed as autistic by the clinical services. At present, it is not possible from the California study to draw any definite conclusions concerning the apparent rise in rates. The author of the California report comes to the same conclusion. The report emphasises the complexity of the problem and the need for properly designed research.

The feeling that there is an increase in the numbers of people with an autistic spectrum disorder could be related to changes in referral patterns and in diagnostic criteria, and greater awareness of the different manifestations of autistic behaviour. Oliver Sacks, in his introductory interview for this conference, makes an interesting comparison with Tourette's disorder, which is now known to be about 1000 times more common than originally thought. It could also be the case that there are real changes in prevalence, either locally or nationally, for reasons not yet understood. However, without properly resourced epidemiological studies, the situation will remain unclear. Are the figures the NAS puts forward generally accepted? Is it helpful to suggest an overall figure for autism spectrum disorders of nearly 1 in 100 (91 per 10,000)?

The figures for those with IQ 70+ mask a wide range of ability and disability. Many of those within this range will not require services to help them to function within society; because they are able to make use of the advantages associated with autistic traits, such as the ability to concentrate and ignore distractions. The key issue for planners is to determine how many people with autistic spectrum disorders have support needs and what those support needs are.

For those who do not have specific support needs there are major ethical issues to consider. Everyone has (or should have) the various human rights won and accepted over the years; rights to privacy and respect under the law. Many people with an autistic spectrum disorder, whether or not diagnosed, will be able to function appropriately and independently within society. Although they may appear different, even eccentric, in other areas of society diversity is welcomed, even celebrated. So it should be for those with apparent social differences. Our role should be to lobby for tolerance of difference.

Recent studies, like that in California, are tending to give much higher prevalence rates for autism than the traditional figure of 4-5 in 10,000, which, as noted above, was found for classic Kanner's autism very narrowly defined. The real problem is to interpret the studies done in the past. What sorts of children were included? How good were the case-finding methods?

What evidence have we got of what went on in the past?

One of us (LW) was involved in the planning of the study by Vic Lotter of 78,000 children aged 8, 9, and 10 years living in the former English county of Middlesex. This was the study in which the 4-5 in 10,000 prevalence rate was first found. I (LW) know what sorts of children were included as classically autistic because I was one of the small group (Neil O'Connor, John Wing, Vic Lotter and myself) who decided on the criteria. In those days we were interested only in really classic Kanner's syndrome and Vic was determined to keep the criteria as narrow as possible. Later, in the Camberwell study described above, Vic was shown case histories of the children Judy Gould and I thought fitted Kanner's descriptions - to our surprise, Vic said we ought to exclude some because they were not classic enough! I think it is fair to say that, when Vic specified narrow criteria, they were NARROW.

When Judy Gould and I started the Camberwell study, we still thought that Kanner's autism could easily be differentiated from other developmental disorders.. By the end of the study our ideas had been turned upside down. We had learnt from direct experience that the psychological dysfunctions underlying autism were manifested in many different ways, far beyond the boundaries of Kanner's syndrome. We developed the hypothesis of an autistic spectrum based on the triad of impairments of social interaction, communication and imagination. Because we concentrated on the children with learning disabilities (IQ under 70) we saw very few with the pattern described by Asperger. We had to wait for the study by Christopher Gillberg in Gothenberg to find out how many children with IQ of 70 and above were also in the autistic spectrum. As described above, combining the results of these two studies gave an overall prevalence rate for the whole autistic spectrum, including those with the most subtle manifestations, of 91 per 10,000 - nearly 1% of the general population.

There have been two follow-ups of the Camberwell children. The latest study, now that the participants range in age from 29 to 43, is being carried out, not by Judy and me, but by independent workers from the University of Kent. They have seen nearly 100 of the 140 individuals from the study who are still alive, and they have confirmed our diagnoses of autistic spectrum disorders.

We are as sure as we can be that, in both studies, reliable counts were made of the individuals concerned. The questions we cannot answer with certainty are, first, was the finding of a high prevalence due simply to chance, and, second, how representative of other areas, in respect of the numbers of people with autistic spectrum disorders, were Camberwell and Gothenberg at the relevant times. Camberwell was a former mainly working class inner London borough. Gothenberg, at the time of the epidemiological studies, was representative of Swedish towns with populations of more than 100,000 . It was partly industrialised but did not have slums.

It would have been a great help if, in the past, many studies using exactly the same methodology and definitions and with intensive case finding had been carried out in a wide range of different types of areas. We would then have been in a much better position to compare rates being found now with those in the past. As things are, the only conclusion is that we do not know if the prevalence of autistic spectrum disorders or of one or more sub-groups has risen or is still rising. There is evidence both for and against. Nor do we know if there are differences in rates or changes in rates in different parts of the world. We have no idea of the prevalence, past or present, in developing countries. There is no way, short of a Wellsian time machine, that we can revisit the past. Examining the possibility that there is a continuing rise in rates of autistic disorders would require intensive studies carried out repeatedly over a long period of time, stretching way into the future - not a very likely project given the present state of funding for research. What is of major importance is improvement in the training of relevant professionals and provision of services so that autistic disorders are diagnosed early and the children and adults, and their families, can be given the help they need.

Postscript:

Kadesjö et al (1999) report a study in Karlstad, a Swedish town. Although this was small scale it was very intensive (over 50% of the 7 year old children seen and assessed personally by the first author). The study found a prevalence for all autistic spectrum disorders for all levels of IQ, of 1.21%!!! Children were followed up four years later and had the diagnoses confirmed.

Arvidsson, T. [et al.] (1997) 'Autism in 3-6 year old children in a suburb of Goteborg, Sweden.' Autism, 2, 163-174.

California Health and Human Services Agency: Department of Developmental Services. (1999) 'Changes in the population of persons with autism and pervasive developmental disorders in California's developmental services system: 1987 through 1998: a report to the Legislature, March 1 1999.' Sacramento, CA: California Health and Human Services Agency, 1999.

Ehlers, S. and Gillberg, C. (1993) 'The epidemiology of Asperger syndrome: a total population study.' Journal of Child Psychology and Psychiatry, 34 (8), pp. 1327-1350.

Kadesjö, B., Gillberg, C. and Hagberg, B. (1999) 'Brief report: Autism and Asperger syndrome in seven-year-old children: a total population study.' Journal of Autism and Developmental Disorders, 29 (4), pp. 327-331.

Wing, L. and Gould, J. (1979) 'Severe impairments of social interaction: and associated abnormalities in children: epidemiology and classification.' Journal of Autism and Developmental Disorders, 9 (1), pp. 11-29.