In “Autism, Brain, and Environment” Richard Lathe argues the case for a real increase in what he describes as “new phase autism” and concludes that, “The rising prevalence of ASD (new phase autism) may be ascribed to environmental toxicity, notably including heavy metals in combination with organic endocrine disruptors and other chemical toxins.” (p210)Does the rise in prevalence represent a real rise in numbers or can it be explained by factors like improved awareness and changes to the diagnostic criteria? This question is addressed in chapter four of the book, to which I now turn.
Any discussion of the epidemiology of autism has to start with Lotter. Yet Lathe seems unaware of his work when he writes, “Over the 1950s to 1970s there were no systematic surveys of the prevalence of ASD.” (p48) This is a surprising omission as Lotter’s is a landmark study. (Lotter 1966) There was an initial screening of 77,800 children aged between 8 and 10 in the county of Middlesex. All those identified as possibly autistic were given psychological tests and their teachers and parents were interviewed.
The resulting prevalence of 4.5 in 10,000 proved remarkably robust. Beginning with Lotter, Wing(1993) compared 16 epidemiological studies carried out between 1966 and 1991. The four using Kanner’s criteria all returned rates between 4.3 and 5.0 per 10,000. Rutter’s criteria and DSM-III are based on Kanner’s criteria as operationalized by Lotter. Two studies that used Rutter’s criteria gave figures of 5.6 and 10.8 respectively. Those using DSM-III are more interesting. The two American and two European studies gave rates of 4.0, 3.3, 4.0 and 7.5. The three Japanese studies that used DSM-III all reported much higher rates: 15.5, 13.8 and 13.0. In two of these routine developmental checks in early childhood were used to flag up potentially autistic children, suggesting a more effective method of initial ascertainment or screening can have as profound an impact on the outcome of epidemiological studies as the diagnostic criteria that are used.
Leaving aside these Japanese studies we have a relatively stable picture in which autism, narrowly defined, occurs within a prevalence range of 3.3 to 7.5 in 10,000 in nine studies carried out between 1966 and 1989. The odd one out is the study that found 10.8 per 10,000 using Rutter’s criteria. This is the only study which relied exclusively on schedules completed by psychologists and psychiatrists who had clinical care of the children they had identified during the initial screening as having “infantile psychosis.” In all other studies the researchers saw and examined the children themselves.
Wing’s review includes three other studies. Two used the DSM-III-R and one used its own criteria. This was a Japanese study (Ishi and Takahashi 1983) and the English summary was not very informative about how its criteria were applied. But it did provide the highest prevalence rate of all the studies under review; 16 per 10,000 from an initial population of 35,000 6 to 12 year olds in the city of Toyota.
The two studies that used DSM-III-R both found increased prevalence. Bryson et al (1988) reported a prevalence of 10.1 in 10,000 in a population of 20,800 6 to 14 year olds in Nova Scotia. Gillberg et al (1991) reported a prevalence of 11.5 in 10,000 in a population of 40,700 4 to 13 year olds in Göteborg.
So up to and including 1991 Wing found 16 studies that satisfied her epidemiological criteria: defined populations with an initial screening procedure and systematic follow up of all potential cases. Half of these studies found rates of around 5 in 10,000 or less. The other half ranged from 7 to 16 in 10,000. Nevertheless, Lathe asserts that, “Until the 1990s ASD was diagnosed at no more than ~5 cases per 10,000.” (p50)
Lathe argues that changes in the diagnostic criteria are not a factor in explaining the rise in prevalence.
“It appears unlikely that the more recent diagnostic methods encompass a wider range of children, because all include the key features of autistic disorders with specific recognition of speech delay, social interaction impairment, eye gaze avoidance, and resistance to change.” (p51) He acknowledges that it has been argued “that DSM-III-R broadened the diagnostic concept of autism.” But “the recommendations of III-R are largely reiterated in DSM-IV, and substantially parallel ICD-10.” (p49)
The reference is to Volkmar et al. (1992) who found that, “Relative to either clinicians, DSM-III, or ICD-10 the DSM-III-R system overdiagnosed the presence of autism.” Hertzig et al (1990) report that, “DSM-III-R criteria have specifically broadened the concept of autism to include children who, although socially impaired, are not unresponsive to others.” These appear to contradict Lathe’s assertion that changes in diagnostic criteria are unimportant.
For the real story of what has happened to the diagnostic criteria we have to return to the Camberwell study. (Wing and Gould 1979) As well as coming close to Lotter’s figure of 4.5 in 10,000 for children with Kanner’s autism, Wing and Gould also found a significant number of children who, while not fulfilling Lotter’s strict diagnostic criteria, did share in the now familiar Triad of Impairments and were deemed to be on the autistic spectrum. Including these children raised the prevalence to 20 in 10,000. (It is worth noting that at the height of the “autism epidemic” in California figures of 30 in 10,000 for children of all abilities were reported. (CDDS 2003) Wing and Gould’s reported rate of 20 in 10,000 only included children with significant cognitive impairments.)
This problem of hyperspecificity was the issue that DSM-III-R was supposed to address. There were children who were plainly autistic to experienced clinicians but who did not meet the published criteria.
“The revisions introduced in DSM-III-R in 1987 were motivated by this low sensitivity [but] it soon became apparent that although sensitivity was now very high, specificity was quite low.” (Szatmari no date)
DSM-IV and ICD-10 are different. For the first time the World Health Organization and the American Psychiatric Association have tried to harmonize their diagnostic criteria. Both seek to reflect Wing’s triad of impairments while “improving its (DSM-III-R) specificity and reintroducing several non-autistic PDD sub-types.” (Szatmari no date)
This tension between the need for sensitivity and the desire for specificity has been a constant feature in the development of our understanding of autism. And there are still problems with the criteria. Asperger Syndrome is a particular problem.
Hippler and Klicpera (2003) examined the records of 74 children diagnosed with autistic psychopathy (AP) by Asperger and his team. Sixty eight percent of these met the ICD-10 criteria for Asperger syndrome. Twenty five percent met ICD-10 criteria for autism. The authors conclude that, “ICD-10 and DSM-IV criteria for AS do not quite capture the individuals described by Asperger and his team. They appear to differentiate AS from autism solely based on the onset criteria, regardless of the patient’s social impairment later in life.”
Dr Judith Gould in a conference handout dated 16/11/02, argues that nosological problems arise when we try to define categories within PDD too precisely. Autism is best understood as a spectrum disorder.
“EVIDENCE FOR A SPECTRUM
One person can show different features in different environments
One person can show different features at different ages
Members of the same family can show different features
Identical twins or triplets can show different features
The same basic principles underlie methods of education and care for the whole spectrum
DIMENSIONS VERSUS CATEGORIES
In clinical practice, it is extremely difficult to define the boundaries between different diagnostic categories, whatever the criteria used. The clinical picture found in those with autistic spectrum disorders fit better with the concept of multiple dimensions than with the concept of separate, definable categories. Individual needs are more accurately assessed from the profile of levels on different dimensions than from assigning a categorical diagnosis.”
So far we have mainly been concerned with those decades prior to 1990. Lathe regards the period since then as “the most critical period under scrutiny.” (p49) It is certainly the period that has seen the largest increase in reported prevalence.
Lotter found that 75% of children meeting his criteria had some degree of mental retardation. Wing (1993) makes the important point that when the children under consideration have mild to moderate mental retardation we see the closest concordance with Kanner’s criteria. More profoundly retarded individuals would be unable to manage the object oriented elaborate routines specified by Kanner. Those with normal or above average intelligence tend to be less object oriented and more verbal and social, especially with increasing age.
Therefore we should expect the shift away from the highly specific diagnostic criteria in DSM-III to the more sensitive criteria in DSM-III-R that catered for all children with Wing’s triad of impairments to lead to an increase in prevalence. The inclusion of Asperger’s Syndrome in ICD-10 and DSM-IV would further increase the numbers. We should also expect the proportion of those with mild to moderate mental retardation to fall even if their absolute numbers remained the same.
Basing their figures on Wing and Gould (1979) and Ehlers and Gillberg (1992) the National Autistic Society estimated total prevalence of autistic spectrum disorders as 20 in 10,000 for those with an IQ < 70 and 71 in 10,000 for those with an IQ > 70. This gives a total prevalence of 91 in 10,000. Fombonne (1997) dismissed these figures and suggested that the best available epidemiological knowledge suggested rates of 5 in 10,000 for childhood autism and 10 in 10,000 for other pervasive developmental disorders. Four years later Fombonne and Chakrabarti (2001) published the result of an epidemiological study in Staffordshire that found rates of 16.8 in 10,000 for autistic disorder and 46.8 in 10,000 for other Pervasive Developmental Disorders. A follow up study by the same authors (2005) using exactly the same methodology with a matched cohort in Staffordshire found broadly similar results. They also found that the 75% mental retardation rate found by Lotter was reversed with only 25.8% showing some degree of mental retardation.
This is consistent with a broadening of the diagnostic criteria to include children who do not display the classic autistic aloofness and are more outgoing and interested in people but are still recognizably autistic.
Lathe highly regards the overview of prevalence data by Williams et al. (2006) This includes all of the studies discussed by Wing and continues up to studies completed in 2004. It also includes three studies that Wing rejected on grounds of methodology. All gave low prevalence rates consistent with the narrow diagnostic criteria they used. Lathe remains convinced that changing diagnostic criteria cannot account for an estimated tenfold increase in prevalence since the early 1990s. Leaning heavily upon Williams et al. (2006) Lathe thinks that there has been a change in the environmental contribution to autism and this has increased over time.
Williams et al identify three factors which Lathe uses to support his case.
Urban versus rural residence.
Lathe takes this as evidence for a strong environmental influence on autism. There are more heavy metals and other pollutants in the cities. But there are also more clinicians, more specialist diagnostic units and easier access to clients and to centralized records used for screening. Williams et al mention all these factors. They do not mention environmental pollution. That is because their paper is not designed to examine the causes of autism. Its aim was to “quantitatively examine the influence of study methodology and population characteristics on prevalence estimates of autistic spectrum disorders.”
Age of Children
Rates were higher when younger children were examined. Lathe takes this to mean that autism is growing at a faster rate among young children. Williams et al suggest that, “Manifestations of autism may be more obvious in young children. Alternatively, some screening methods may be more sensitive for younger children.”
Lathe also cites administrative data from education that shows a preponderance of cases of autism among younger children as opposed to those in secondary education. This is to be expected. Howlin and Moore (1997) discuss the immense difficulties that parents faced then in obtaining a diagnosis. Typically parents faced a four year delay between expressing initial concerns and obtaining a diagnosis. These were highly motivated parents who were all members of the National Autistic Society. Many had to battle with health and education officials to obtain the necessary referrals. It is open to question how many other parents managed to stay the course.
When researching my book (Stanton 2000) difficulties in obtaining a diagnosis was the main bugbear reported by parents. The situation is improving. But the attention paid to child development in the early years suggests that younger rather than older children are the chief beneficiaries of this.
Lathe thinks that this is also evidence of an actual growth in autism over time because he does not believe that the latest diagnostic criteria are broader than before. Most commentators would not agree. Fombonne (2003) states that, “It is generally agreed that the definition of autism has broadened over the last decades, particularly at the less severe end of the spectrum.”
Fombonne also makes the point that variation in prevalence is often linked to methodology. “For example, in 4 US and 4 UK studies published recently, 14- and 6-fold variations in prevalence rates were found, respectively. Although these two sets of studies were conducted at the same time in similar age groups and in the same countries, the lack of consistency in estimates is striking and demonstrates how unique design features within each study can affect prevalence estimation. In both countries, studies relying on single administrative sources for identifying cases yielded low estimates, whereas investigations using proactive methods of case finding, that is, multiple sources of ascertainment and direct diagnostic procedures, yielded much higher rates.”
Williams et al. seem to support Fombonne rather than Lathe. The authors argue that, “The covariate most strongly associated with prevalence estimates for typical autism and all ASD was the diagnostic criteria used.” When commenting upon the higher estimates of prevalence arising from Japanese vis a vis US studies they note that, “All the Japanese studies used prospective diagnostic assessments and all but one drew on whole population rather than clinical samples.”
Lathe does look to other examples to support his claim that there has been a real increase in autism. There is an impressive graph showing a steady rise in prevalence between 1970 and 1990 followed by a massive rise following 1990. But the report from the CDDC also carries a “government health warning.”
“The information presented in this report is purely descriptive in nature and standing alone, should not be used to draw scientifically valid conclusions about the incidence or prevalence of ASD in California. The numbers of persons with ASD described in this report reflect point-in-time counts and do not constitute formal epidemiological measures of incidence or prevalence. The information contained in this report is limited by factors such as case finding, accuracy of diagnosis and the recording, on an individual basis, of a large array of information contained in the records of persons comprising California’s Developmental Services System. Finally, it is important to note that entry into the California Developmental Services System is voluntary. This may further alter the data presented herein relative to the actual population of persons with autism in California.”
This is a reminder that we are dealing with administrative data here, not epidemiological data. DDS data records numbers who are in receipt of services. Year on year rises record the net gain and show that more people are entering the system than leaving it. They do not necessarily indicate a growth in new cases. Of course year on year rises are consistent with an increase in the number of diagnoses. As I noted earlier most commentators believe that this is due to a broadening of the diagnostic criteria together with increased autism awareness and better ascertainment. Ironically, the autism epidemic hyped in the media has probably contributed to this process.
An earlier study of the California data by Byrd et al(2002) is cited by Lathe as evidence that growth in reported rates of autism in California could not be explained by a loosening of the criteria. They did this by taking two samples of children, one born between 1983 and 1985 and the other born between 1993 and 1995, and using the ADI-R interview with their parents to determine if both groups met the DSM-IV criteria. The children themselves were not tested.
There is an obvious flaw to this argument. If the newer criteria are more inclusive, of course children diagnosed with autism under more exclusive criteria will also meet the modern criteria. It would have made more sense to test the more recent cohort using DSM-III. The limitations of the Californian data and Byrd’s report are discussed more fully by Gernsbacher et al(2005) who also notes that Wing and Potter (2002) offer a similar example in which all the children from an earlier cohort who met Kanner’s criteria also met DSM-IV criteria. But only 33 to 45% of children more recently diagnosed using DSM-IV also met Kanner’s criteria. Nevertheless Lathe also cites a similar experiment in Minnesota applying DSM-IV criteria retrospectively to earlier generations to demonstrate a ten fold increase in autism.
Changing spectrum of impairments.
Lathe argues that data from the Byrd study suggest “that the pattern of presentation of the disorder has altered over the 1980s through 2000s, with autistic disorder becoming more prevalent with respect to other conditions.” (p55)
The main support for this is the drop in Asperger Syndrome from 15% to 2%. But this drop is based on parental reports of diagnosis in 261 subjects in the two samples taken from 1983-5 (n=100) and 1993-59 (n=161). None of the children were tested to confirm the parental reports. Lathe also refers to the growing cognitive abilities of autistic children in California. Again, this is consistent with a loosening of the criteria to admit more able children onto the spectrum. Lathe’s refusal to admit this possibility leads him to see a rising tide of autism where others, myself included, see a rising tide of recognition.
Evolution of twin concordance rates.
Traditionally, we find high rates of autism in identical twins and less high rates among non-identical twins. But rates are still high enough to confirm a significant genetic component. Lathe cites three studies, all of which are unpublished, to support his contention that environmental factors are impacting upon rates of autism by changing the concordance rates in twins.
The first thing to remember is that the numbers in these studies are very small. One unpublished study has 41 pairs of identical twins and 46 pairs of non-identical twins. In another study, there were only 32 identical pairs. The second thing is that in two of these studies researchers relied entirely on parental and/or practitioner reports. The third thing to remember is that these are unpublished reports. Until the rest of us can read these reports they cannot inform the debate.
Decline of a specific genetic contribution; Fragile X.
Lathe thinks that because the rate of Fragile X is constant a rise in prevalence will lead to a fall in the proportion of individuals with ASD who are Fragile X. I agree? I have already argued that an increase in prevalence does not necessarily mean an actual increase in autism. Fragile X, tuberous sclerosis, PKU have all been associated with ASD. They are also associated with mental retardation. If less restrictive diagnostic criteria are finding more cases without mental retardation it stands to reason that they will find less cases associated with Fragile X. Lathe takes this as proof of increased prevalence. But we all agree that prevalence has increased. This is not disputed. The argument revolves around whether or not the rise in prevalence represents an actual rise in cases of autism and is this rise caused by novel and potentially reversible environmental insults.
It is my contention that the evidence so far does not support Lathe’s theory of a change in the nature of autism, a new phase autism brought on by significant environmental impacts of a new and recent kind. Instead I favour the view that we have been witnessing a change in the nature of our understanding of autism.
That is not to deny the possibility that there has been an actual growth in the incidence of autism. We do not know because reliable epidemiological data comparable over time does not exist. Administrative data from school records is suggestive but as Shattuck(2006) has pointed out, this data cannot be used either to prove or disprove the theory that there has been real growth in autism. This does not stop such figures being used to fuel media scare stories as happened in Scotland recently. Autism may be growing. But there is nothing in the data to support a dramatic increase fuelled by toxins that only emerged at the beginning of the 1990s.
One feature of the closing decades of the twentieth century and the beginning of the new millennium has been an increasing awareness of and a willingness to embrace diversity as opposed to the old adherence to traditional values and social conformity. It may be that the rigidities of the past made life more predictable and easier to manage for autistic people. Now we live in interesting times that are of our making but are uncomfortable times for our autistic brethren. They and their difficulties are more apparent because of this. We owe it to them to strive to embrace their diversity and accommodate their difficulties. Perhaps the environmental influences on autism are not heavy metals. There are other toxins in society.
Bryson, S. E., Clark, B. S. & Smith, I. M. (1988). First report of a Canadian epidemiological study of autistic syndromes. JournaI of Child Psychology aud Psychiatry, 29, 433-446
CDDC (2003) AUTISTIC SPECTRUM DISORDERS Changes In The California Caseload An Update: 1999 Through 2002
Chakrabarti, S., & Fombonne, E. (2001). Pervasive developmental disorders in preschool children. Journal of the American Medical Association, 285, 3093-3099.
Chakrabarti, S., Fombonne, E., (2005). Pervasive developmental disorders in preschool children: confirmation of high prevalence. American Journal of Psychiatry, 162(6), 1133-41
Ehlers, S. & Gillberg, C. (1993) The epidemiology of Asperger syndrome. a total population study. Journal of Child Psychology and Psychiatry, 34 (8), 1327-1350
Fombonne, E. (1987) Prevalence of autistic spectrum disorder in the UK (commentary) Autism1(2), 227-229
Fombonne, E. (2003) The prevalence of Autism (editorial). JAMA 162(1), 87-91
Gernsbacher, M. A., Dawson, M., & Goldsmith, H. H. (2005). Three Reasons Not To Believe In An Autism Epidemic. Current Directions In Psychological Science, 14, 55-58.
Herzig, M.E, Snow M.E., New E., Shapiro T. (1990) DSM-III and DSM-III-R diagnosis of autism and pervasive developmental disorder in nursery school children. J Am Acad Child Adolesc Psychiatry, (1):123-6
Gillberg, C., Steffenburg, S. & Schaumann, H. (1991). Is autism more common now than ten years ago?
British Journal of Psychiatry, 158, 403-09.
Hippler, K. & Klicpera, C. 2003 A retrospective analysis of
the clinical case records of ‘autistic psychopaths’ diagnosed
by Hans Asperger and his team at the University Children’s
Hospital, Vienna. Phil. Trans. R. Soc. Lond.
lshii,T. & Takahashi, O. (1983). The epidemiology of autistic children in Toyota, Japan: Prevalence. Japanese Journal of Child and Adolescent Psychiatry, 24, 311-321
Lotter, V. (1966). Epidemiology of autistic conditions in young children: I. Prevalence. Social Psychiatry, 1, 124-137.
Shattuck P.T. (2006)The Contribution of Diagnostic Substitution to the Growing Administrative Prevalence of Autism in US Special Education. Pediatrics 117, 1028-1037
Volkmar F.R., Cicchetti D.V., Bregman, J., Cohen D.J. (2002) Three diagnostic systems for autism: DSM-III, DSM-III-R, and ICD-10. (4):483-492
Williams J.G., Higgins J.P.T., Brayne C.E.G. (2006) Systematic review of prevalence studies of autism spectrum disorders Archives of Disease in Childhood 9, 8-15
Wing, L. (1993) The Definition and Prevalence of Autism: A Review. European Child and Adolescent Psychiatry, 2 pp.61-74
Wing, L. & Gould, J. (1979). Severe impairments of social interaction aud associated abnormalities in children Epidemiology and classification. Jounial of Atism and Childhood Schizophrenia, 9, 11-29.
Wing, L., & Potter, D. (2002). The epidemiology of autistic spectrum
disorders: Is the prevalence rising? Mental Retardation and
Developmental Disabilities Research Reviews, 8, 151–162