The Shape Of An Elephant

17 May

Remember this old Bhuddist parable?

Five blind men of Savatthi are all describing an elephant. The problem is that one grabs the tail, the other a leg, the other the side, the other an ear and the fifth, the tusk. Each, remaining blindfolded, seeks to articulate the attributes of an elephant. The one who grabbed the tail insisted that the elephant was like a rope. The one who grabbed the leg was as certain that an elephant was not like a rope, but a tree. The one who was feeling the side of the elephant was convinced that an elephant was like a mud baked wall. The fourth blind man, feeling the ear, was shocked that the others could not understand that the elephant was like a banana leaf. The fifth denounced them all as he held to the tusk, insisting that an elephant was most like a brandished sword.

Every time I hear talk about the ‘autism epidemic’ I remember this parable.

The only way we can _definitively_ establish if thiomersal (or any other vaccine ingredient) causes autism is to take a hundred kids and do a double blind study involving injecting them with either an applicable amount of thiomersal containing vaccines or a control over an established time period.

Obviously, thats never going to happen. Firstly there are the obvious ethics of such a thing – with the prevailing beliefs about what autism is, no parent is going to risk ‘causing’ autism. Secondly there is the more practical reason that there aren’t really any thiomersal containing vaccines left in the West anymore – hence Burbacher’s need to get vaccines and then _add_ thiomersal to them. I suppose our ficticious study could do that but nobody really knows what confounders there may be in such an action. Burbacher certainly didn’t control for them.

So, what else can we do to try and establish if thiomersal (or whatever) can cause autism?

We can examine the symptoms of mercury poisoning (in the case of thiomersal) and see if there seems to be a relationship with autism. This is in essence what the Bernard et al paper tried to do. They concluded there _was_ a link but a closer examination of the paper shows that there is _not one_ common symptom between the diagnstic symptoms of mercury poisoning and the DSM(IV) diagnostic criteria for autism. This fact usually results in two counter-claims. Firstly that the DSM(IV) is not ‘up to the job’ of reflecting the current state of knowledge about autism. Secondly, that autism is such a novel form of mercury poisoning that autism is totally different from all other forms of mercury poisoning.

The first objection is essentially a call to retro-fit the DSM(IV) to fit one persons own beliefs about autism and thiomersal. This is pointless. The DSM criteria (which _are_ periodically adjusted) reflect the symptoms it requires to fulfil a diagnosis of ASD. This means the symptoms are common to _all_ autistic people. People have quoted gut issues, constipation and various other issues to me as ‘evidence’ of the damage resulting in autism, that thiomersal can do. Trouble is, none of the things that get quoted at me are common to all autistic people. These things may be comorbidities. If people have found ways to treat debilitating comorbidities then more power to them I say. I do exactly the same every time I administer a puff of a ventolin inhaler to my daughter. People then go on to say, well, maybe we should start sub-dividing autism into different ‘types’. However, we have no idea what prevalence these ‘sub-types’ might have. As far as we know they might only exist in statistically insignificant numbers that wouldn’t justify a sub-type categorisation. One of the biggest comorbidities is epilepsy. Should we create a sub-category of ‘epileptic autism’. Why? The underlying autism would be just the same. No – this is the very reason why secondary conditions are called comorbidities and not subtypes.

The second theory – that autism is so unique it doesn’t resemble any other form of mercury poisoning – is very hard to take seriously. Anorexia appears to be common across all types of mercury poisoning (its mentioned in Mad Hatters Disease, Pinks Disease and typical mercury poisoning) – why would it skip autism? Occams Razor applies here. The simplest explanation is one which requires no mangling/disappearing/ignoring of known facts – autism doesn’t really resemble mercury poisoning.

So whats next? Epidemiology. We’re left with looking at the numbers.

The ‘autism epidemic’ is central to the thiomersal hypothesis. The argument goes that as thiomersal useage increased both temporaly (vaccines were administered in shorter time frames) and in amount (maximum body burden in the US was 187.5 ug of Hg) that the number of autism diagnosis increased.

The main problem with the epidemic idea is that this chain of events is _far_ from established. The reason is mainly the quality of the underlying data.

There are three main US sources for prevalence data – the Dept of Education, VAERS and CDDS.

Many autism advocacy groups use the data collected by the US Department of Education (USDE) to show a rapidly increasing prevalence of autism. Closer examination of these data to follow each birth-year cohort reveals anomalies within the USDE data on autism. The USDE data show not only a rise in overall autism prevalence with time but also a significant and nearly linear rise in autism prevalence within a birth-year cohort as it ages, with significant numbers of new cases as late as 17 years of age. In addition, an unexpected reduction in the rise of autism prevalence occurs in most cohorts at 12 years of age, the age when most children would be entering middle school. These anomalies point to internal problems in the USDE data that make them *unsuitable for tracking autism prevalence*.


This is a shame but Jim Laidler is absolutely correct that we must use good, accurate data – USDE data clearly isn’t.

VAERS has massive problems. It allows anyone to enter any data at any time. I recently demonstrated this when I, a UK citizen, managed to submit a VAERS entry stating that a vaccine had turned my daughter into Wonder Woman. Clearly, this is not an acceptable source.

CDDS is the most contraversial. Rick Rollens has toally misintrpretted the data time after time. CDDS themselves state that their data should not be used for tracking autism prevalence. However, if it is insisted that we _do_ use CDDS data then we need to be clear about its use. Rick Rollens lumped all stats from all age groups together – quite obviously this results in meaningless data. As David Kirby conceeded:

…total cases among 3-5 year olds, not changes in the rate of increase is the right measure.

When one does isolate this cohort things are very different. In this cohort, nnot only are autism cases still rising, in the last quarter, the increase in the rate of increase is climbing.In other words, when one uses the correct group of cases to examine, data that David Kirby has referred to as ‘the Gold Standard’ for testing prevalence, shows that autism cases are still rising despite his statement in the New York Times in *2005* that:

Because autism is usually diagnosed sometime between a child’s third and fourth birthdays and thimerosal was largely removed from childhood vaccines in 2001, the incidence of autism should fall this year.

However, despite all this, we need to rememember that CDDS disclaimers appply to our interpretation of the data as well as Rollen’s or Kirby’s. However, ours are more accurate and at least are preformed on the right section of the data.

Make sure to read Joseph’s first comment in this thread which addresses another failing of this data I forgot to address.

So there are very large problems with the epidemiology as well. This is vexing and means, as Paul Shattuck recently concluded, that the true growth of autism cannot be realistically determined. So we’re left with the opinions and research of experts – people who study autism. What do they say?

Almost to a man they say that the idea of an epidemic is questionable. They state there may well have been a rise in _numbers_ but not necessarily a rise in _prevalence_. The distinction is important.

What they say is that improved tests and more recognition adds up to more diagnosis. This is simple common sense. If you know what you’re looking for, you’ll find more of it than you would if you _didn’t_ know what you were looking for.

What do we know that might support this opinion? Here are a few ideas from my neck of the woods.

In 2004, an ‘autism audit‘ was performed in Scotland. One of the questions the audit asked was how accurate they thought the prevalence rate estimates were for their area. 45% of authorities who responded made a point of noting that they felt diagnosis for adults was very underrepresented. For example, Perth and Kinross council stated

Figures for adults reflect the national findings that the numbers known to services/diagnosed represent a significant underestimate of those individuals likely to be affected. For example day centre managers locally consider a number of people to be on the spectrum who have had no formal diagnosis.

Also, in a New Scientist piece last year, the findings of the University of Nottingham were reported. The team reexamined data from the 1970’s which resulted in five diagnosis. Using modern diagnostic criteria, the team found 56 cases, a ten-fold increase.

Lastly, earlier this year, Health Minister Liam Byrne reported figures that demonstrated autism diagnoses for children have nearly doubled in 8 years from 3100 to 6170. Meanwhile adult diagnoses have nearly tripled in the same period from 1120 to 3000.

That seems to pretty firmly establish the idea of widescale underdiagnosis. What about misdiagnosis? From its very start as a categorised diagnosis, autism has been misdiagnosed. Kanner mentioned several of his patients were diagnosed with schizophrenia. However, as Shattuck _also_ concluded, its not possible to ascertain to what degree diagnostic substitution in the past has resulted in more cases now we know better.

A fascinating news article caught my eye this morning and led to this post. It seems that even _after_ we factor in more availability of diagnosis and better tests for it, most doctors still don’t screen for autism because a lot don’t know how.

The study of 255 Maryland and Delaware pediatricians found that 209 (82 percent) said they regularly screen their patients for general developmental delays, but only 20 (8 percent) of them said they regularly screen for ASD. Of the pediatricians who said they do not routinely screen for ASD, 62 percent said they didn’t do it because they weren’t familiar with the screening tools.


Remember that this is in a time when awareness and screening tools are better than they’ve ever been – if they’re like this now, just imagine how bad they must’ve been 10, 15 or 20 years ago.

Do you see a rope? A tree? A wall? A leaf? Maybe a sword? Or do you put these things together and percieve an elephant?

102 Responses to “The Shape Of An Elephant”

  1. mcguffin May 28, 2006 at 18:24 #

  2. Chris June 7, 2006 at 08:13 #

    There are several problems with the DSM autism set.
    (1) it changes; III-R isn’t the same as IV.
    Theoretically, this change comes about because of clinical observations; but this necessarily involves clinical observations of people who don’t have the existing criteria but are still autistic. Which means that there is a concept of autism out there that’s independent of and prior to the criteria.
    (2) Criteria, items – the chinese menu approach still involves there being 64,000 (III-R) or 12,500 (IV) possible presentations, which makes generalisations almost impossible;
    (3) the absence of norms to plot deviations against.

    The purpose of DSM isn’t to sort a group into one of two camps; it’s to label people we already know are deviant.

Comments are closed.

%d bloggers like this: