It is now more than 18 months since the UK’s General Medical Council found Andrew Wakefield guilty of dishonesty and other serious professional misconduct; and it is nearly a year since the BMJ concluded that his now retracted Lancet paper linking the measles, mumps, and rubella (MMR) vaccine with autism and bowel disease was an “elaborate fraud.” At that time, January 2011, we called on Wakefield’s former employer, University College London (UCL), to establish an inquiry into the scandal. Ten months on, no inquiry has been announced.

Our coverage in January showed how Wakefield manufactured the appearance of a link between the vaccine and regressive autism while employed by lawyers trying to build a case against the MMR vaccine, and while negotiating extraordinary commercial schemes that would succeed only if confidence in the vaccine was damaged. … [and] …. that the conflicts of interest were not confined to Wakefield. They drew in his then employer, the Royal Free hospital and medical school. …

Now we can report that it is not only Wakefield and UCL’s administrators who have a case to answer. This week we publish new information that puts the spotlight on Wakefield’s coauthors. Previously unpublished histopathology grading sheets apparently completed by Amar Dhillon, the senior pathologist on the paper, remove any remaining credibility from the claim that the Royal Free doctors had discovered a new inflammatory bowel disease associated with MMR. Along with UCL’s failings during and after Wakefield’s tenure, this evidence also raises wider concerns about the prevailing culture of Britain’s academic institutions….

It gets worse. Here.

The evidence used in this study come primarily from two sources: Epidemiologic data and clinical or other biological studies (which the study refers to as “mechanistic”). The epidemiologic findings from previous studies indicate that there may be an effect of interest, provides information about its frequency or link to various risk categories, and allows an assessment of the likelihood that the putative effect is really linked to the vaccine, as opposed to showing up by coincidence. An example of this might be the question of severe flu-like illness as an effect of getting a flu shot. Since people tend to get their flu shots at the beginning of flu season or later, there is a certain chance that they will actually have the flu at the time they get the shot, or some other similar illness. If one person tells you “I got the flu shot once and it gave me the flu” you have not learned anything about the link between flu shots and flu-like side effects. You need to see if among a large number of people a disproportionate number get severe flu-like symptoms after they’ve gotten the shot.

The clinical or biological studies address the mechanism of the effect. No matter what the epidemiological evidence may be, if there is not a sensible biological mechanism behind an effect, it is hard to say what is happening, or even to verify that anything is happening at all. Chance alone allows for what would seem like an inordinate number of instance of a particular effect. In other words, if you look at a possible side effect for a particular medical treatment in 100 studies, it may well “show up” as sample-wide phenomenon in one or two case by chance, with nothing actually happening biological; A false positive. But if there is a known underlying biological mechanism, then the epidemiological data can be more sensibly assessed, and if the biological model that emerges from these “mechanistic” studies is sufficiently detailed, it might be possible to improve the field studies by collecting more appropriate data.

Each of these categories of information, epidemiologic and mechanistic, were assessed for each putative link between vaccination and adverse effect. Each study was assessed for its strength with respect to sample size, research design, or other factors. The result of this was a measure of “weight-of-evidence” for each type of study. These were then combined into a third assessment about the causal relationship between the vaccine and the effect.

For epidemiologic evidence, the following categories were used:

• High: Two or more studies with negligible methodological limitations that are consistent in terms of the direction of the effect, and taken together provide high confidence.
• Moderate: One study with negligible methodological limitations, or a collection of studies generally consistent in terms of the direction of the effect, that provides moderate confidence.
• Limited: One study or a collection of studies lacking precision or consistency that provides limited, or low, confidence.
• Insufficient: No epidemiologic studies of sufficient quality.

For mechanistic evidence, the categories were:

• Strong: One or more cases in the literature, for which the committee concludes the vaccine was a contributing cause of the adverse event, based on an overall assessment of attribution in the available cases and clinical, diagnostic, or experimental evidence consistent with relevant biological response to vaccine.
• Intermediate: At least two cases, taken together, for which the committee concludes the vaccine may be a contributing cause of the adverse event, based on an overall assessment of attribution in the available cases and clinical, diagnostic, or experimental evidence consistent with relevant biological response to vaccine. On occasion, the committee reviewed evidence consisting of at least two cases that, taken together, while suggestive, are nonetheless insufficient to conclude that the vaccine may be a contributing cause of the adverse event. This evidence has been categorized as “low-intermediate.”
• Weak: Insufficient evidence from cases in the literature for the committee to conclude the vaccine may be a contributing cause of the adverse event, based on an overall assessment of attribution in the available cases and clinical, diagnostic, or experimental evidence consistent with relevant biological response to vaccine.
• Lacking: No clinical, diagnostic, or experimental evidence consistent with relevant biological response to vaccine, regardless of the presence of individual cases in the literature.

The categories used for causation were:

• Evidence convincingly supports a causal relationship –This applies to relationships in which the causal link is convincing, as with the oral polio vaccine and vaccine-associated paralytic polio.
• Evidence favors acceptance of a causal relationship–Evidence is strong and generally suggestive, although not firm enough to be described as convincing or established.
• Evidence is inadequate to accept or reject a causal relationship–The evidence is not reasonably convincing either in support of or against causality; evidence that is sparse, conflicting, of weak quality, or merely suggestive–whether toward or away from causality–falls into this category. Where there is no evidence meeting the standards described above, the committee also uses this causal conclusion.
• Evidence favors rejection of a causal relationship–The evidence is strong and generally convincing, and suggests there is no causal relationship.

The results of this study support 14 specific adverse event relationships with vaccines ranked in the “convincingly supported” category (shown in the table below). In all but one case, this was based on strong mechanistic evidence along with epidemiologic evidence of either limited confidence of insufficient. This probably means that in many cases there is a real link but the occurrence of the event is rare.

The “favors acceptance” category four links between vaccine and adverse effect.

Convincingly supported links are:

Varicella Vaccine:

• Disseminated Oka VZV without other organ involvement (got chicken pox)
• Disseminated OK VZV with subsequent infection resulting in Pneumonia, Menningitis, or Hepatitis (got chickenbox, bad)
• Vaccine strain viral reactivation without other organ involvement
• Vaccine Strain viral reactivation with subsequent infection resulting in menningitis or encephalitis
• Anaphylaxis (a multi-system immune reaction, very variable but considered dangerous)

MMR Vaccine:

• Measles Inclusion Body encaphalitis
• Ferbile seizures
• Anaphylaxis

Influenza Vaccine:

• Anaphylaxis

Hepatitis A Vaccine

• Nothing Noted

Hepatitis B Vaccine:

• Anaphylaxi

s

HPV Vaccine:

• Nothing noted

DT-IT and aP containing vaccines:

• Anaphylaxis

Meningococcal Vaccine:

• Anaphylaxis

Injection-related events:

• Deltoid Bursitis (arm hurts)
• Syncope (fainting

)

Keep in mind that most of these effects are rare and many are minor. The main effect linked to these vaccines is immunity to a potentially deadly disease.

You are probably wondering about autism. I’ll give you the study’s results directly (keep in mind that this is an uncorrected proof):

The committee reviewed 22 studies to evaluate the risk of autism after the administration of MMR vaccine. Twelve studies…were not considered in the weight of epidemiologic evidence because they provided data from a passive surveillance system lacking an unvaccinated comparison population or an ecological comparison study lacking individual-level data. Five controlled studies … had very serious methodological limitations that precluded their inclusion in this assessment. The five remaining controlled studies contributed to the weight of epidemiologic evidence.

Taylor et al. (1999) conducted a self-controlled case series study in children with autistic disorders residing in the North East Thames region of the United Kingdom. The children were identified from computerized special needs or disability registers. A total of 498 children who were born from 1979 through 1998 and had an autism diagnosis before 16 years of age were included in the analysis. After reviewing the clinical records, the investigators confirmed that the autism diagnoses met the criteria of the International Classification of Diseases, 10th revision (ICD-10) in 82 percent of typical autism cases and 31 percent of atypical autism cases (the authors used the term core to describe typical autism, as noted in the methods). The self-controlled analysis investigated the risk of typical or atypical autism diagnosis among 357 cases during two postvaccination periods (12 or 24 months after vaccination). The reference period consisted of time from birth through August 1998, not including the postvaccination risk periods. The relative risk of autism diagnosis within 12 months of MMR vaccination was 0.94 (95% CI, 0.60-1.47) and within 24 months of MMR vaccination was 1.09 (95% CI, 0.79-1.52). The relative risk of autism diagnosis within 12 months and 24 months of vaccination with MMR or single-antigen measles with mumps and rubella was 0.80 (95% CI, 0.53-1.22) and 1.05 (95% CI,0.76-1.44), respectively. The authors noted the results were similar when the analyses were restricted to confirmed cases of typical or atypical autism. The authors concluded that MMR vaccination is not associated with autism.

Farrington et al. (2001) conducted a reanalysis of the study by Taylor et al. (1999). The two risk periods were changed to autism diagnosis within 59 months and any time after vaccination, and compared to a reference period that consisted of time from birth through 191 months of age or August 1998, whichever occurred first. The analysis was adjusted for both calendar year and age. The relative risk of autism diagnosis within 59 months of vaccination with MMR was 1.24 (95% CI, 0.67-2.27), and with MMR and any measles-containing vaccines was 0.96 (95% CI, 0.52-1.77). The relative risk of autism diagnosis any time after vaccination with MMR was 1.06 (95% CI, 0.49-2.30), and with MMR and any measles-containing vaccines was 2.03 (95% CI, 0.80-5.18). The authors concluded that there is no association between MMR or measles-containing vaccines and autism diagnosis any time after vaccination.

Madsen et al. (2002) 2 conducted a retrospective cohort study in children born in Denmark from January 1991 through December 1998. The children were enrolled from the Danish Civil Registration System, which stores personal identification information for all residents, and linked records to five other national registries. MMR vaccination data was obtained from the National Board of Health, autism diagnosis was derived from the Danish Psychiatric Central Register. The National Hospital Registry and Danish Medical Birth Registry provided birth weight and gestational age information, and data on socioeconomic status and mother’s education came from Statistics Denmark. Autism diagnoses were based on criteria from the ICD-10; the diagnostic codes were separated into cases of autistic disorder or other autistic-spectrum disorders. Children with congenital rubella or an inherited genetic condition (fragile X syndrome, Angelman’s syndrome, or tuberous sclerosis) were excluded from the analysis. A total of 537,303 children were included in the cohort, of which 316 had an autistic disorder diagnosis and 422 had an autistic-spectrum disorder diagnosis. Follow-up began at 1 year of age and continued through December 31, 1999, or the date of autism diagnosis, diagnosis of other associated conditions, emigration, or death. Children who were vaccinated with MMR
contributed 1,647,504 person-years of follow-up, and those not vaccinated contributed 482,360 person-years. Relative risks were calculated and adjusted for age, calendar period, sex, birth weight, gestation age, mother’s education, and socioeconomic status. The adjusted relative risk of autism diagnosis after MMR vaccination was 0.92 (95% CI, 0.68-1.24) and of other autistic spectrum disorders after MMR vaccination was 0.83 (95% CI, 0.65-1.07). The authors concluded that MMR vaccination is not associated with an increased risk of autistic disorder or other autistic-spectrum disorders.

Smeeth et al. (2004) conducted a case-control study in children (born between 1973 and 1999) enrolled in the General Practice Research Database (GPRD) from June 1987 through December 2001. The study included 991 cases with a recorded diagnosis of autism and 303 cases with other pervasive developmental disorder (PDD) diagnosis. A total of 4,469 controls were individually matched to cases on year of birth (within 1 year), sex, and general practice. The study excluded cases and controls that were not enrolled in the database for at least 12 months before the diagnosis or index date (date that control was same age as matched case at time of diagnosis). MMR vaccination data was abstracted from the GPRD records, and the case or control status was concealed during the assessment. The unadjusted odds ratio for autism diagnosis after MMR vaccination was 0.77 (95% CI, 0.60-0.98). After adjustment for the age at which participants joined the GPRD, the odds ratio was 0.88 (95% CI, 0.67-1.15). The authors concluded that MMR vaccination is not associated with an increased risk of autism.

Mrozek-Budzyn et al. (2010) conducted a case-control study in children identified in the general practitioner records in the Malopolska Province of Poland. The study included 96 cases and 192 matched controls. The cases were diagnosed with childhood or atypical autism by a child psychiatrist according to the ICD-10 criteria. Two controls were matched to each case on year of birth, gender, and physician’s practice. Vaccination histories and the date of autism diagnosis were extracted from the physician’s records. Date of onset of symptoms was derived from parental interview. If MMR or single-antigen measles vaccination preceded the onset of symptoms, cases were classified as vaccinated. Controls were considered vaccinated if they received an MMR or single-antigen measles vaccine before the age of symptom onset observed in the matched case. The analysis adjusted for mother’s age, medication during pregnancy, gestation time, perinatal injury, and 5-minute Apgar scale score. The adjusted odds ratio for autism diagnosis after MMR vaccination was 0.17 (95% CI, 0.06-0.52). The adjusted odds ratio for autism diagnosis after single-antigen measles or MMR vaccination was 0.28 (95% CI, 0.10- 0.76). The authors concluded that administration of MMR or single-antigen measles vaccine is not associated with an increased risk of autism in children.

So, from an epidemiological perspective, there is no link between MMR and autism. But what about the mechanistic studies? There were …

… four publications reporting autism developing after the administration of MMR vaccine. Three publications did not provide evidence beyond temporality, some too long (Frenkel et al., 1996; Spitzer et al., 2001; Wakefield et al., 1998). Long latencies between vaccine administration and development of behavioral symptoms make it impossible to rule out other possible causes In addition, the committee identified an editorial by Sharrard (2010) in which a temporal relationship between administration of a measles, mumps, and rubella vaccine and the development of autism was attributed to one patient reported in Verity et al. (2010). However, as reported in the original article and affirmed in a subsequent letter to the editor (Verity et al., 2011) the vaccinee did not develop autism, a fact that was misreported in the editorial by Sharrard. Two publications studied the association between MMR vaccination and autism with enteropathy (Hornig et al., 2008; Peltola et al., 1998). The authors reported a temporal relationship between vaccine administration and development of gastrointestinal disturbances but did not report autism after vaccination. The publications did not contribute to the weight of mechanistic evidence.

In the end, the committee was able to comfortably reject a connection between MMR vaccination and autism.

Source:

Stratton, Kathleen, Andrew Ford, Erin Rusch, and Ellen Wright Clayton,
Editors; Committee to Review Adverse Effects of Vaccines; Institute of
Medicine. 2011 (uncorrected proof). Adverse Effects of Vaccines: Evidence and Causality. The Ntional Academies Press. Washington DC

Spencer et al, working from various units of Cambridge University seem to have identified such a marker.

… we show, for the first time, that the neural response to facial expression of emotion differs between unaffected siblings and healthy controls with no family history of autism. Strikingly, the functional magnetic resonance imaging (fMRI) response to happy versus neutral faces was significantly reduced in unaffected siblings compared with controls within a number of brain areas implicated in empathy and face processing. The response in unaffected siblings did not differ significantly from the response in autism. Furthermore, investigation of the response to faces versus fixation crosses suggested that, within the context of this study, an atypical response specifically to happy faces, rather than to faces in general, accounts for the observed sibling versus controls difference and is a clear biomarker of familial risk.

Autism is associated with limited reactions to emotional states of others. This study seems to identify the neurological component of that phenomenon, and sees it in individuals who are not otherwise diagnosed as autistic. This could suggest that the neural phenotype seen in this fMRI study is the root cause of the condition of autism, though of course that would have to be confirmed somehow with developmental studies.

The study included 40 adolescents between 12 and 18 years of age who were diagnosed with autism spectrum disorder, specifically with either autism or Asperger syndrome, as well as 40 unaffected siblings of those affected and 40 controls who did not seem to have autism of any kind. They were then given a task of observing faces displaying different emotional states while their brains were being scanned. The results are rather stunning. The control group showed differences in brain response when comparing happy vs. neutral faces, while both the sibling and autism groups simply did not. When comparing fearful and neutral faces, the control group and sibling group showed similar responses, but the autism group showed no change.

The lack of brain response to facial stimuli in autistic subjects was expected and has been demonstrated in previous studies. What was not expected, apparently, is the similarity between non-autism spectrum disorder siblings and the autism-diagnosed individuals. Various methods were used to ensure that the observed pattern had to do with emotional aspects of the faces being observed rather than the overall ability to parse what was being seen.

Press reports of this study emphasize that finding a biomarker in siblings proves that autism is genetic (and not, for instance, caused by vaccines). This is in fact what attracted me to the paper to begin with. However, this is simply not demonstrated by the paper. Merely having a set of observations inside the head, which seems somehow more deeply biological, does not make the observation more about genes than when the same phenomenon was only observed outside the head!

Putting it another way, The Brain Is Not Made of DNA! (That’s a post by Neuroskeptic on the same paper.)

It is important, however, that autism may include a feature that exists in non-diagnosed individuals and apparently not in the population at large, though this should be confirmed with a larger sample. If siblings of autistic individuals show this neurological “endophenotype” then there must be people out there who have the endophenotype but no close relatives with autism. If so, then the reason that some individuals express diagnosable autism could be investigated. This study thus both broadens the potential expression of an important phenotype and provides an avenue for research in figuring out what it is that causes it, and may encourage a closer look at sibs with and without diagnosed autism.

Spencer, M., Holt, R., Chura, L., Suckling, J., Calder, A., Bullmore, E., & Baron-Cohen, S. (2011). A novel functional brain imaging endophenotype of autism: the neural response to facial expression of emotion Translational Psychiatry, 1 (7) DOI: 10.1038/tp.2011.18

The story is told by Jamie across two blogs: Autism One, Part One on Skepchick and How I Got Kicked Out of the AutismOne Con: Part 2 on Friendly Atheist. Ken Reibel gives his version of the events here.

The peer reviewed paper is available in the Open Access journal PLoS Genetics.

The sample included 2,832 individuals distributed among 912 families that had multiple autistic children. The control group consisted of 1,070 samples of disease-free children who presumably are not clustered from a smaller number of family groups.

Note the apparent imbalance in sample size. Actually, it is not as big of a difference as one might think, as the 2,832 individuals samples do not represent 2,832 cases because they come from a set of under 1,000 families. In effect, the sample size of autism-related individuals in 912.

Aside from the variable regions, the study also discovered two novel genetic alleles that may be related to autism, BZRAP1 and MDGA2, which are related to synaptic function and neurological development, respectively. It is important to note that suspect alleles of these genes were passed down in some, but not all, of the affected individuals in families.

From the author’s summary:

Autism spectrum disorders (ASDs) are common neurodevelopmental syndromes with a strong genetic component. ASDs are characterized by disturbances in social behavior, impaired verbal and nonverbal communication, as well as repetitive behaviors and/or a restricted range of interests. To identify genes likely to contribute to ASD etiology, we performed high density genotyping in 912 multiplex families from the Autism Genetics Resource Exchange (AGRE) collection and contrasted results to those obtained for 1,488 healthy controls. To enrich for variants most likely to interfere with gene function, we restricted our analyses to deletions and gains encompassing exons. Of the many genomic regions highlighted, 27 were seen to harbor rare variants in cases and not controls, both in the first phase of our analysis, and also in an independent replication cohort comprised of 859 cases and 1,051 controls. More work in a larger number of individuals will be required to determine which of the rare alleles highlighted here are indeed related to the ASDs and how they act to shape risk.

According to the research team’s leader, Hakon Hakonarson, “We focused on changes in the exons of DNA–protein-coding areas in which deletions or duplications are more likely to directly disrupt biological functions. We identified additional autism susceptibility genes, many of which belong to the neuronal cell adhesion molecule family involved in the development of brain circuitry in early childhood.”

This research, if confirmed, supports the notion that autism spectrum disorder is a heterogeneous group of disorders, with a heterogeneous set of causes some (or most?) of which may be genetic.

Bucan, M., Abrahams, B., Wang, K., Glessner, J., Herman, E., Sonnenblick, L., Alvarez Retuerto, A., Imielinski, M., Hadley, D., Bradfield, J., Kim, C., Gidaya, N., Lindquist, I., Hutman, T., Sigman, M., Kustanovich, V., Lajonchere, C., Singleton, A., Kim, J., Wassink, T., McMahon, W., Owley, T., Sweeney, J., Coon, H., Nurnberger, J., Li, M., Cantor, R., Minshew, N., Sutcliffe, J., Cook, E., Dawson, G., Buxbaum, J., Grant, S., Schellenberg, G., Geschwind, D., & Hakonarson, H. (2009). Genome-Wide Analyses of Exonic Copy Number Variants in a Family-Based Study Point to Novel Autism Susceptibility Genes PLoS Genetics, 5 (6) DOI: 10.1371/journal.pgen.1000536

…Andrew Wakefield manipulated patients’ data, which triggered fears that the MMR triple vaccine to protect against measles, mumps and rubella was linked to the condition.

The research [originally] claimed that the families of eight out of 12 children attending a routine clinic at the hospital had blamed MMR for their autism, and said that problems came on within days of the [vaccinatoi]. The team also claimed to have discovered a new inflammatory bowel disease underlying the children’s conditions.

However, our investigation, confirmed by evidence presented to the General Medical Council (GMC), reveals that: In most of the 12 cases, the children’s ailments as described in The Lancet were different from their hospital and GP records. Although the research paper claimed that problems came on within days of the jab, in only one case did medical records suggest this was true, and in many of the cases medical concerns had been raised before the children were vaccinated. Hospital pathologists, looking for inflammatory bowel disease, reported in the majority of cases that the gut was normal. This was then reviewed and the Lancet paper showed them as abnormal.

(Emphasis added)

The story is here.

See also:
A Quick Note to Huffington Post

Did the founder of the antivax movement fake autism-vaccine link?
Anti-vax study a case of scientific fraud?
The anti-vaccination movement—rotten to the core

This study is quickly becoming the focus of attention as the various factions with an interest in autism square off on assessing its validity. In the mean time, the study itself is rather modest in what it attempts and what it concludes.

Let’s have a look.

To date, there are three kinds of explanations given for this rise in Autism rate:

1) There is some artifact in the system such as changes in reporting or diagnosis, or changes in the definition of autism, that makes the rate of autism look like it is going up, but it is actually not going up;

2) There is a genetic change in populations, either because of local changes in the gene pool or immigration/migration causing this condition to increase in frequency;

3) There is an environmental cause of an actual increase in autism which is indicated in the increase in numbers of autism diagnoses.

The study by Hetz-Picciotto and Delwiche examines a number of different previously suggested explanations that fall into the first of these categories, and finds that these do have an impact on the apparent rise of reported autism, but according to these authors not enough to explain the entire phenomenon. The second hypothesis is also tested, or at least partly controlled for, by excluding immigrants (to California) from the study. While the authors are left concluding that explanations of the third category should be more closely investigated, they do not offer specific environmental explanations, and in fact conclude that “the extent to which the continued rise represents a true increase in the occurrence of autism remains unclear.

Taken at face value, this study seems to suggest an evening out of funding for autism research, which is allegedly highly biased towards genetic studies, to include more investigation into environmental causes. However, life is not so simple, and the following three considerations come to mind:

1) Previously, public interest groups had emerged which made a link between autism and vaccination. These groups asserted that specific chemicals in vaccines caused autism in some children, and that this accounted for the rise in autism that we see. These groups were largely funded and staffed by parents with autistic children who, understandably, wanted to find some entity to blame, but not so understandably, may have chosen a scape goat, with the cost of diminishing chances of finding a real explanation for autism (regardless of any rise in incidence). Various scientific studies seem to have debunked the vaccine link, the most important of which showed no decreases in autism prevalence after the most often blamed chemical found in some vaccines were removed from use.

2) It is possible that anti-vaccine denialist spokespeople, i.e, those who have all along asserted that the special interest groups mentioned above were wrong, have expanded their argument to include all or most environmental causes of the disorder, at the same time that those special interest groups have entrenched in an environmental cause camp. This makes it difficult to evaluate any research that simply tries to address the cause of the condition.

3) The research reported here was conducted by members of an institute funded by special interest groups that may have overlapped with those mentioned above. One has to consider the possibility of links between funding sources and researchers.

None of these three aspects of the problem should matter in the long run. The research reported here can to some extent be evaluated on its own terms, and attempts to replicate it or disprove it’s implications can and should (and likely will) be done by other research teams. But they do matter a great deal when it comes to discussing this issue. Recently, I posted a paragraph from a Scientific American article addressing this issue on this blog, without comment, and was instantly accused of being … I don’t know, some kind of anti-science denialist. Presumptions were made about what I was thinking, and not one person previously involved in this debate saw fit to be even remotely polite.

What does this mean? It means that if there are an anti- or non-scientific factions working hard in the wrong direction than necessary to nail down autism and related conditions, find out what causes them and work towards cure or relief, they will do better than they otherwise might in their efforts. Instant add-water-and-stir vitriol is stupid and counter productive. Go read the comments on my post and marvel at it all!

But what about the study at hand?

This is actually fairly complicated. This is an epidemiological study that looks at state databases showing autism diagnoses. The study notes and documents a remarkably high increase in autism rates, going from fewer than one per 10,000 children to close to 12 for the youngest cohort and 40 for the oldest cohort. For a broader context, other studies have shown rates of varying amounts, but close to (yet above) 10 per 10,000 for Autism, and a much larger value of near 50 or in some cases much more for the broader categories of diagnosis that include but are not limited to autism. The study consideres, but argues against, the idea that the shift seen in California is a change from a narrow diagnostic range to a broader one.

The study looks at the following possible causes of the increase: Younger ages of diagnosis, migration of autism-rich populations into the state, changes in diagnostic criteria, and inclusion of milder cases into the category as time goes by.

The study concludes that while any of these could account for some of the changes, the sum of these effects is insufficient to explain the data. In particular, this study suggests that a 56% increase can be explained by inclusion of milder cases, and a 12% increase can be explained by changes of age at diagnosis.

The problem is that this analysis does not (and can not) do what we really want, and that is to measure the actual number of children with the same exact criteria at two or more points in time. What we are left with instead is a difficult game of measuring rates. Rates are always tricky.

Fourty is 400 percent of 10, so a shift from 10 to 40 cases per 10,000 sounds like a large increase. But what if that ’10’ was off a bit, and it was actually five? What was 400 percent is now 800 percent. But what if that 10 was off in the other direction, and was actually a 15? Now we’re talking about 270%.

This is the problem with extended ratios. Small changes at one end or the other and especially changes at the ‘short’ end of the ratio, can make large changes in the key index variable (the percent increase).

Even more dramatic would be a comparison of ‘then’ vs. ‘now’ in which we assumed that some of the effects occurred early vs. later. That could get very complicated.

By the most extreme reasoning in one direction, the increase in new cases per 10,000 can be explained at one order of magnitude by diagnostic and reporting factors, but is occurring at one more level of magnitude higher. By the most extreme reasoning in the other direction, cumulative effects of diagnostic and reporting artifacts can easily include the increase.

There is another way to think about this. Take the data at face value. In particular as shown in this figure from the paper:

Is this a genetic change? No. Too fast. This is way sub-generational, and migration has been excluded from the analysis. Is this an environmental change? It looks like one would expect for certain environmental changes, but if it is, then what factor in the environment is so different spanning some 20 years. Such a thing should be obvious. This does resemble, in my view, a secular change in effect of the kind one might expect with differences in diagnosis and reporting, because such changes take years to take hold, but can have a large influence.

In other words, other than ruling out in situ evolution, I don’t think you can tell what causes this. The fact that a big chunk of this variation is probably explained by changes in reporting and diagnostic criteria suggests that more of the same sort of effect may be sufficiently explanatory. The fact that a careful look at reporting and diagnostic effects does not readily explain the level of magnitude of the change we see here suggests that more explanation is needed.

In the absence of a correlation between these data and a list of causal effects (which could then lead to some effective hypothesis testing) it important to keep an open mind about what causes autism. I can think of no reason that this study’s validity or lack thereof informs us in this regard. Those who wish to insist that no matter what there is no increase in autism rates are no less a failure at explaining autism as those who see a real increase in graphs like this one.

Meanwhile, the authors of this study and others are looking into the data further to test for environmental links. According to a press release from UC Davis:

Hertz-Picciotto and her colleagues at the M.I.N.D Institute are currently conducting two large studies aimed at discovering the causes of autism. Hertz-Picciotto is the principal investigator on the CHARGE (Childhood Autism Risk from Genetics and the Environment) and MARBLES (Markers of Autism Risk in Babies-Learning Early Signs) studies.

CHARGE is the largest epidemiologic study of reliably confirmed cases of autism to date, and the first major investigation of environmental factors and gene-environment interactions in the disorder. MARBLES is a prospective investigation that follows women who already have had one child with autism, beginning early in or even before a subsequent pregnancy, to search for early markers that predict autism in the younger sibling.

“We’re looking at the possible effects of metals, pesticides and infectious agents on neurodevelopment,” Hertz-Picciotto said. “If we’re going to stop the rise in autism in California, we need to keep these studies going and expand them to the extent possible.”

Irva Hertz-Picciotto, Lora Delwiche (2009). The Rise in Autism and the Role of Age at Diagnosis Epidemiology, 20 (1), 84-90 DOI: 10.1097/EDE.0b013e3181902d15