A review of: Begeer, S., Bouk, S.E., Boussaid, W., Terwogt, M.M., Koot, H.M. (2008). Underdiagnosis and Referral Bias of Autism in Ethnic Minorities. Journal of Autism and Developmental Disorders DOI: 10.1007/s10803-008-0611-5

The issue of under- or over-representation of a disorder within specific ethnic groups is a complicated one. There are specific disorders that are under-represented within a specific ethnic group because of some protective factor that makes such group less likely to acquire the disorder. For example, the rates of skin cancer in the African-American population are significantly lower than in the European-American population(although this has led to increase mortality rates among African Americans due to reduced screenings leading to late diagnosis). Yet, it is possible that some disorders are under-represented within an ethnic group simply because a systemic clinical bias in diagnosis and referrals. To examine this hypothesis, the authors of this study first examined 712 case records of children referred for ASD assessment in the Netherlands. They found that ethnic minority children (Turkish and Moroccan)were under-represented in this sample of referred kids as compared to Dutch children (2.1% vs. 4.4%). But does this represent a bias or is it simply that Turkish and Moroccan children are less likely to have ASD due to some protective factor? To answer this question, the authors sent 6 clinical vignettes to 82 pediatricians. The vignettes varied in their descriptions of various autism symptoms. Three ethnic background were represented, including 1) European minority (French or English) 2) Non-European minority (Moroccan and Turkish) and 3) Majority (Dutch). However, the ethnicity was independent of the clinical vignette, so that the vignette sent to one pediatrician could describe a Dutch child, while the SAME vignette sent to another pediatrician could describe a Turkish child. The authors found that vignettes describing Dutch (majority) children elicited significantly more references to autism than did vignettes describing European minority or non-European minority children. However, the mean rate of ASD based on an objective scale was equal across all three groups. This suggests that objective assessments may help minimize any potential clinical biases.

A review of: Baird, G., Charman, T., Pickles, A., Chandler, S., Loucas, T., Meldrum, D., Carcani-Rathwell, I., Serkana, D., Simonoff, E. (2008). Regression, Developmental Trajectory and Associated Problems in Disorders in the Autism Spectrum: The SNAP Study. Journal of Autism and Developmental Disorders DOI: 10.1007/s10803-008-0571-9

Although most children with autism present very early signs and symptoms and a linear developmental trajectory, a small subset of children present a trajectory characterized by normal development followed by a loss of acquired skills or a failure to use the acquired skills. This pattern has been termed autistic regression. Possible explainations for this phenomenom have varied from a genetic effect on brain restructuring and pruning during the early stages of life, to enterocolitis due to vaccinations, to epilepsy. In this study, the authors explored differences in developmental outcomes for children with and without regressive autism, and the association between regression and enterocolitis and epilepsy. This study examined a population cohort born in the UK in 1990 and 1991. Out of 56,946 children in this cohort, 218 had and ASD diagnosis by age 10. A subset of these children were evaluated via ADOS and ADI and divided into a broad autism (N=105), narrow autism (N=53), and no autism (N=97). The narrow autism group met full criteria for autism based on ICD-10. The broad autism group met clinical consensus for autism but not full ICD-10 criteria. These children were then evaluated for history of epilepsy, gastroinstestinal problems, and developmental regression. 39% of children with narrow autism had a history of regression during development. This compared to 11% of children with broad autism, and 3% of children with no autism. On average this regression occurred around the 25 month of age. There were no differences in IQ or adaptive functioning between those with or without regression. However, those with regression classified in the broad autism group had significantly more symptoms than those without regression also classified in the broad autism group. Regression was not associated with gastroinstestinal symptoms or with epilepsy.

A brief review of: Daniels, J.L., Forssen, U., Hultman, C.M., Cnattingius, S., Savitz, D.A., Feychting, M., Sparen, P. (2008). Parental Psychiatric Disorders Associated With Autism Spectrum Disorders in the Offspring. PEDIATRICS, 121(5), e1357-e1362. DOI: 10.1542/peds.2007-2296

The journal of Pediatrics just published a population study based on the national Swedish registry, which examined the association between parental psychiatric history and autism. The authors compared the parental psychiatric history of 1,227 of children with autism spectrum disorder and 30,925 typically developing children. Children were identified as having autism spectrum disorder if they were born between 1977 and 2003 and had a diagnosis of ASD recorded in the registry between 1987 and 2003.

Parents of children with autism were 70% more likely than parents of typically developing kids to have a psychiatric diagnosis. When both parents had a psychiatric disorder, the children were 100% more likely to have a diagnosis of autism. Schizophrenia was more common in both parents among children with autism as compared to parents of typically developing kids (90% more likely for mothers and 110% more likely for fathers). In addition, mothers of children with autism were more likely than mothers of typically developing kids to have depression (70%), and personality disorders (70%).

In summary, the study suggests that in Sweden, during the last 30 years, children with a diagnosis of autism were more likely to have parents with psychiatric diagnoses than typically developing children. This could reflect a non-specific, possibly genetic, predisposition in affected families for psychiatric conditions, including autism. It could also reflect that having a child with autism increases stress in the parents possibly leading to psychiatric diagnoses. However, the association noted by the authors was even stronger if the parental diagnosis was provided before the child’s diagnosis. One important consideration, these results were based only on kids who had a history of inpatient treatment. Those with a history of only outpatient treatment were not included. It is possible that the observed link between parental psychiatric history and autism applies only, or mostly, to the most severe cases of autism requiring hospitalization.

One last comment: It’s important to note that the rate of psychiatric conditions among even children with autism were very low. For example, schizophrenia was observed among 0.6% of the mothers of children with autism (compared to 0.2% of the typically developing mothers). 99.4% of the children with autism did not have mothers with schizophrenia. Therefore, the data only suggest that there may be a familial/genetic predisposition that is related to autism among very small subset of children with autism.

…But this study is compelling in showing an association between mercury exposure and autism rates, and scientists can not just ignore it under the basis of its imperfect design and inability to make causal links – if that is the case, then only carefully controlled laboratory studies, with poor external validity, should be published and accepted as contributors to our greater scientific knowledge.

A review of: PALMER, R., BLANCHARD, S., WOOD, R. (2008). Proximity to point sources of environmental mercury release as a predictor of autism prevalence. Health & Place DOI: 10.1016/j.healthplace.2008.02.001

This fascinating, yet bound to be controversial, study hit the news yesterday as it was made available (pre-publication) by the Journal Health & Place. The study is simple, straightforward, elegant, with some powerful findings. In fact, the findings are somewhat daunting given the simplicity of the design. The researchers reviewed the amount of mercury release reported by industrial facilities and power plants in the State of Texas in 1997 from data provided by US Environmental Protection Agency Toxics Release Inventory. They compared these data against autism rates in 1997 and 2002 as measured by schools’ autism classifications provided by the Texas Education Agency. Using a specialized geographical analysis system, the authors were able to locate each source of mercury and calculate the distance between each mercury source and each school. The results:

Industrial release of mercury and distance to industrial sources independently predicted increased rates of autism. The association with industrial release of mercury was not linear, instead the statistical model fit suggested an accelerated risk. This association remained statistically significant after controlling for specific variables such as SES, urbanicity, and race.

Power plant release of mercury and distance to power plant independently predicted increased rates of autism. In this case the association was linear (not accelerated). Again, this association remained statistically significant after controlling for other variables.

It is easy to dismiss these findings as inconsequential because they are ‘correlational’ in nature, or do not really prove anything. Researchers are too often guilty of selective acceptance of research: those studies that fit the consensus are accepted while those that don’t are dismissed for their methodological flaws – even though the studies we accept are equally flawed.

In the spirit of fairness I have to say that these findings are strong. Their methodology and analytical process are not any different from what is commonly seen in social science or epidemiology research. Is it perfect? Far from it. Is it useful or informative? Definitively! The data speak very clearly: In Texas, mercury release from industrial sources and power plants in 1997, and school proximity to these sources, are associated with rates of autism in 2002 as measured by school special education classifications.

Does this mean that mercury causes autism? Not at all. In the last sentence of the previous paragraph you can not replace the words are associated with with the word cause. There is a major difference. The data, albeit strong, have limitations. For example, the most obvious (to me) alternative explanation is that mercury release and proximity to these sources is also associated with another mystery factor that is causing this apparent association and that in fact, mercury release has nothing to do with autism rates in 2002. Let’s hypothesize that these power plants and industrial sources also release another toxin – let’s call this toxin autisimic (this is a made up toxin). These sources release mercury and autisimic at the same rate, so for each pound of mercury released there is a pound of autisimic released. It is possible then that this autisimic toxin directly increases the risk for autism, and this could explain completely the strong (but now obviously inaccurate) association between mercury release and autism.

Does this study show that vaccines cause autism? Absolutely not. I know this question may sound ludicrous to some, but I pose it rhetorically because I am certain that some will make the wide leap and link these findings to the vaccine issue.

There are other problems and limitations with this study, such as how autism rates were calculated (using all children instead of only those born inor after 1997), whether the autism rates are truly climbing and not explained by other factors, whether there are other variables that could be explaining this relation, etc, etc — and yes, this study does not prove or directly indicate that autism is caused by mercury exposure (click here for a much more critical review of this study). But this study is compelling in showing an association between mercury exposure and autism rates, and scientists can not just ignore it under the basis of its imperfect design and inability to make causal links – if that is the case, then only carefully controlled laboratory studies, with poor external validity, should be published and accepted as contributors to our greater scientific knowledge. This is study is far, far, from perfect, and many changes should have been requested prior to publication, but I can say the same of 90% of what is published today.

As I read Autism related blogs, discussion forums, and news articles, I see many people using the number 1 in 150 to refer to the current rates of autism. Yet, few know or understand where this number comes from and how it was obtained. I hope this brief review can provide some clarity on this issue:

The autism rate of 1 in 150 was published by the US Center for Disease Control in 2007 as part of a weekly disease morbidity and mortality surveillance report. The 1 in 150 rate was obtained from a population-based study of 8 year old children conducted in 2002. Specifically, teams in 15 US States reviewed health and educational records of children born in 1994. Trained clinicians classified them as having an autism spectrum disorder if:

1) had a documented previous classification of an ASD (i.e., the child had either an uncontradicted record of an autistic disorder or ASD diagnosis provided by a qualified examiner or documentation of qualification for special education services during 1994–2002 under an autism eligibility category)
or
2) did not have a documented ASD classification but had an evaluation record from an educational or clinical source indicating unusual social behaviors consistent with an ASD.

However, the clinical team conducted an additional detailed analysis of the records to ensure that the accepted DSM-IV criteria for autism and ASD was met prior to classifying each child. Thus, classification was not only based on prior records, but also included a secondary analysis by a clinical team that utilized structured procedures to maximize the validity and reliability of their diagnostic process.

The results:
Overall, the teams reported a rate of 6.6 per 1,000 children as meeting the diagnostic criteria for ASD (1 in 151). Rates by State varied significantly, but this was affected by differences in the way rates were obtained. Some States were able to determine rates based on health records AND educational records, while others could not get access to educational records. As expected, States that had access to educational records had higher prevalence rates than those that only examined health records. On average, States with access to educational and health records reported an autism prevalence rate of 7.2 per 1,000 (1 in 139), while those with only access to health record reported a rate of 5.1 per 1,000 (1 in 196). The male to female ratio significantly varied by State and ranged from 3.4 to 1 in Maryland to 6.5 to 1 in Utah.

Things to keep in mind:
- This report was based ONLY on children born in 1994. Thus it is possible that the rates could not apply to other cohorts.
- The differences in prevalence rates between States with and without access to educational records could suggest that 1) the overall rate is an underestimate because some sites only had access to health records, or 2) that the overall rate is an overestimate because some sites included cases ascertained from educational records which may be less reliable than health records.
- This rate of 1 in 150 does not refer to new cases of autism, or total cases in the population. It only speaks to cases among 8 year old children in 2002.