UK children are nearly 5 times more likely to have autistic conditions than adults according to the results of the first ever UK government survey to assess the numbers of adults with the condition: ["Autism Spectrum Disorders in adults living in households throughout England - Report from the Adult Psychiatric Morbidity Survey 2007" The NHS Information Centre.]
But the British public were told instead that the survey shows rates in adults and children have always been the same so that that vaccines like MMR cannot be to blame for the increasing rates in children. The UK National Health Service funded the survey and made these claims.
Survey Authors Couldn’t Find Enough Adult Autistics
There were early rumours in 2008 the authors could not find numbers of adult ASC cases matching numbers in children.
The survey failed to find a single case of an adult with “typical” or “classic” autism, found in approximately 30% of ASC children. “Typical” or “classic” autism is a type of ASC controversially claimed in 1998 might be associated with the MMR vaccine.
The authors also failed to find sufficient adult ASC cases overall – just 19 in 7,451 adults, being “higher functioning” ASCs – mainly Asperger’s Syndrome. This represents an overall rate of just under 1 in 300 potential adults cases and not the officially claimed 1 in 100 for children.
The authors say that if they had carried out the survey differently they might have found four times as many adults - a further 46 or 47 ASC cases to come up with the same figure as for children. Using already estimated data they claimed to be able estimate-on-the-estimate that the adult autism rate for the entire UK is the same as in children [1 in 100].
The survey’s failure to find adult cases is troubling. The survey’s original design was broadly based calling into question the authors’ claims to be able to estimate for such a large number of claimed-to-be-missed cases. An unlikely 72% of the adult participants were selected as potentially having one of four mental illnesses [rather than just ASC]. The study was designed so that those most likely to have an ASC were interviewed and those unlikely to were excluded.
Autism In Children Already Higher Than 1
A problem with the survey is that two recent formal peer reviewed journal published studies have found the rate in children is not 1 in 100 but much higher. A 2006 study [Baird] found a rate in children of 1 in 85 and a recently published study by Baron-Cohen et al of 2005 data found a rate of ASC’s in children of 1 in 64 when children not yet diagnosed were taken into account. This gives a rate of ASC in children 5 times higher than the 1 in 300 indicated by the survey’s results. Neither of these studies are mentioned in the survey although members of the same team were involved in the Baron-Cohen study, and Professor Baron-Cohen was a consultant to the survey.
Impossible for Adult ASC Rate to Be Same as Childrens’
The authors inexplicably also fail to explain how their estimate of ASC rates in adults is the same as in children when their results show ASC rates in adults and children are significantly different:-
- Aspergers Syndrome rates in
adults on these results are 40%
higher than in children;
- correspondingly Aspergers Syndrome cases in children compared to adults would have to have substantially decreased for this to be so;
- the rate of “classic” autism has leapt from zero in adults to 30% of ASC cases in children;
- adults with “classic” autism do not exist.
Identical Figures – A Remarkable Coincidence
The addition of the authors’ hypothetical “estimated” cases provides exactly the same rate of 1 in 100 previously accepted for children. This is despite the authors themselves warning the results should be approached with caution because of the inaccuracies. In such circumstances if rates were the same then a similar figure might be expected but not exactly the same one.
The 1% figure for children was stated in pre publicity announcing the start of survey: University of Leicester to lead audit of adults with autism Eurekalert 9-May-2008. “Author bias” in favour of finding a pre-determined figure favourable to the study funder is a well-known phenomenon in formal medical literature.
Breaching the codes of practice
prohibiting the release of national
statistics for government policy
purposes the UK National Health
Service’s publicity put out to the media
on publication of the survey focussed on
claiming ASC rates in adults and
children were identical and that
therefore the MMR vaccine did not cause
ASCs. This was not the stated purpose
of the survey and the authors make no
mention of the vaccine-autism issue.
Media stories included:
Autism rates back MMR jab safety
Michelle Roberts BBC Tuesday, 22
September 2009 and
Autism just as common in adults, so MMR
is in the clear Sarah Boseley The
Guardian Tuesday 22 September 2009.
Survey Does Not Live Up To Its Title
The survey was originally announced to establish the UK adult autism rate for the first time by finding what were then being claimed to be “missed” adult cases of autism.
The publicity claimed “The new prevalence study now underway will give the first ever accurate picture of how many adults have the condition.” [Emphasis added]: University of Leicester to lead audit of adults with autism Eurekalert 9-May-2008. The purpose was to inform UK authorities in planning service provision for adults autistics.
The authors used “second-hand” data collected in 2007 which covered only potential adults cases of Asperger’s Syndrome unsuitable for the claimed purpose. The authors planned an additional part of the survey to collect data on adults with other ASCs such as “typical” or “classic” autism but failed to mention this or any resulting data or results in their recently published survey report. They stated in 2008:-
The prevalence study will make use of new data collected in 2007 by NatCen and Professor Brugha’s team to record the number of adults with Asperger’s syndrome and high functioning autism. There will also be an additional part to the study on the number of people with autism who have more complex needs and learning disabilities. The aim of the combined research will provide good epidemiological information in terms of prevalence and the characteristics and problems of this group.“
University of Leicester to lead audit of adults with autism Eurekalert 9-May-2008.
The standard approach under what is called “DSM IV” to diagnose ASC’s was applied but the authors based their assessments on a broader test for ASCs used in a manner neither accepted or relied by other professionals nor scientifically validated. Inexplicably the survey’s results of the DSM IV diagnoses were neither used nor published.
1 in 100 figure for children is based on
formal diagnoses using accepted methods
and published in formal peer reviewed
medical literature. The authors’ survey
has not been independently reviewed nor
has it been assessed for compliance with
accepted statistical standards.
The 19 potential adult ASC cases in 7,451 participants included 17 male and 2 female, a rate of 1 in 8.5 whereas 1 in 5 cases in children are girls.
Other problems with the survey include:-
- Inclusion in the study was based on the ability of selected members of the public to answer questions on the telephone followed by cooperation with a complex psychiatric assessment. The participants were therefore self-selecting and appear unrepresentative of the general population
- The 19 cases claimed included no one from an ethnic minority, and a very low rate in females (the ratio of females with ASD is known to be 1 in 5 cases in children, but the survey results indicate it is half that in adults at 1 in 10)
- Cooperation of more than half the 14,000 households approached is a remarkable and unexplained achievement. The questionnaire screened for psychotic disorder, Asperger syndrome, borderline personality disorder, and anti-social personality disorder. Were the participants told this and would they have participated had they known?
The survey authors give a 95% “confidence interval” to their claim their estimate of adult ASC rate is the same as in children. A confidence interval is a measure of trust in the reliability of the results but despite being the broadly based the survey failed to find the “missing” cases.
- The survey results were published despite not having been through peer review nor verified for conformance to accepted standards for government statistics
- Adults with “typical” autism [associated with the MMR vaccine] remain missing despite being a category claimed in the prior publicity to be part of the survey.
Dr Carol Stott commented on the methodology in Age of Autism:-
Whatever claims are made to the contrary this report tells us very little about the number of adults with ASDs – in England or anywhere else. The main problems with the study are with (a) case-definition, (b) ascertainment (c) diagnostic instruments (d) case identification and (e) statistical power.
There is no clearly stated case-definition anywhere in the report. The cases reported are defined loosely in terms of an initial score on a shortened unstandardised non-peer reviewed version of ascreening tool together with a semi-structured clinical interview – the Autism Diagnostic Observation Schedule (ADOS) – that was not designed to act as a stand-alone diagnostic instrument.
The sample selected is not representative of the ASD population. All participants were verbally fluent, living in ordinary households, and able to complete a self-report questionnaire.
The choice of measurement tools is inadequately justified and badly referenced. No details are provided, nor are any sources referenced, on the psychometric properties of the initial screening instrument (AQ-20). The standardized scoring criteria for the ADOS were not followed (using a total cut-off of 10 for Communication + Social Reciprocity, rather than three cut-offs (respectively) of 3, 6 and 10 for Communication, Social Reciprocity and the two combined). Additionally the authors over state the validity of the ADOS as a tool for use in adult populations. Module Four (used in the study) was standardized on a sample of only 70 adults aged between 16 and 44. Ages in the study sample range from 16 – 75.
The technical appendix, which is intended to provide information about the derivation of the AQ-20 is statistically naïve, unclear and potentially inaccurate. It is not clear for example, what is meant by the phrase on pg 16 of Appendix C “….once the final set of predictors had been selected, a regression equation was available for predicting the prevalence of ADOS.” Neither is it clear whether General Linear Modelling, Linear Regression or both were used to derive final items.
Finally, inferences are made about the lack of a significant association between age-groups and ASD prevalence without reference to statistical power. In a study of this size, with only 19 identified (unweighted) cases, the likelihood is that the study was underpowered to detect such differences.
A valid and reliable study of the population frequency of a disorder requires clear and robust case definition, validated instruments, standardized procedures and adequate statistical power. An initial evaluation of this report suggests it fails on all counts.