The first official sounding of the death knell for the ‘hidden horde’ hypothesis of autism (it must be vaccines because there aren’t 1 in 160/150/100 adult autistics) was recently sounded (as reported by Anthony) when an NHS study announced that the rate of autism amongst adults was 1% – exactly the same as for kids.
Fully aware of the implications for their beloved anti-vaccine hypothesis John Stone and Polly Tommey write/star in a recent post on that repository of all things bullshitty – Age of Autism.
NHS Autism Report suggests the increase in autism in recent years was all down to an oversight
Says John Stone, trying to reassure his readership theres still life in the terminal old dog of the hidden horde hypothesis. And why does he say this?
“Are we really able to believe all that we hear on such important subjects, or is there a stronger hand with adifferent agenda behind it?”
Ahhhh, of course! When a piece of science/news story doesn’t support your point of view then it _must_ be the work of….oh, whoever – Illuminati? Out of control government? GAVI? Take your pick.
But surely Stone and Tommey have better reasons than that…? Don’t they?
Well no, not really. They didn’t like the perfectly valid Autism Diagnostic Observation Schedule that was used. Although they don’t refer to it by name, instead they cherry pick questions to try and demean the validity of the test.
The report was based on adults living in households throughout England and the assessments were done in two stages. Stage One asked individuals to rate how well they agreed with 20 statements about their likes, dislikes and abilities – such as whether they preferred going to ‘libraries’ or ‘parties’.
…If the full range of ability had been included in the NHS report, alongside the verbally fluent, high functioning adults living at home, there would surely be far more than 1 in 100. So, either there are more adults than children with an ASD and autism is on the decline(!?) or there is something wrong with the report.
A quick glance at many mainstream autism orgs claim on prevalence would show Stone that yes, there are more autistic adults than children. This report could indeed be interpreted to show a decline in autism.
People will claim that thats counter intuitive but it really isn’t. As autism has become more recognised, better diagnosed and there are more centers worldwide for its early detection, child numbers have not risen but have become more apparent. This is true across the whole ‘spectrum’.
Taking the NAS numbers on prevalence They say that at a rate of 1 in 100 children gives a total of around 133,500 kids with an ASD. This leaves 366,500 adults. A rate of 0.8%.
So the estimate prior to the new report was 0.2% out. But Stone was right, there are now (and always were) more ASD adults than kids. Thats because there are more adults in the general population than kids. In terms of whether the _rate_ of autism has decreased there’s no research to compare this to but an estimate of 0.8% is pretty damn close to 1%. In real terms the NAS were only out by 733 adults. So I would say that this report reflects reality – with no doubt regional variance and other factors, the adult rate is (and always was) around 1%.
But the stoopid continues apace:
One thing is clear, however, the report has no relevance to children withautism like my son Billy and the many thousands like him.
Quoth Polly Toomey who somehow missed the point that this was a study of adults and thus was never going to have much relevance to her children.
We’re then treated to the science of Carol ‘try me shithead‘ Stott who personally received over £100,000 of legal aid money to prop up the dead MMR hypothesis. She claims (after further digs at ADOS) that a ‘further detailed critique’ will be appearing in that robust science journal AutismFile….owned and operated by one Polly Tommey.
So there we have it. This new study, which whilst far from 100% perfect is also far from the dead dog portrayed by the purveyors of science at The Autism File and Age of Autism. They think theres a conspiracy to ‘big it up’ and refuse to see the statistical truth. ASD hasn’t ever been in epidemic, theres been a stable rate for a long time. These dealers in anti-vaccinationism will just have to deal with that.
Left Brain Right Brain 
“and still think, well, it’s still possible autistic adults practically don’t exist, that’s incredibly closed-minded.”
Joseph, you are changing the question again. I am not saying that adults with autism don’t exist. Even if the prevalence in adults is *only* 0.2 percent that is still a significant group.
I am questioning a specific prevalence level that you feel is appropriate and I am questioning the assertion that the prevalence in adults is the same as in children.
So far you have failed to provide any direct evidence to support your ideas.
Thank you for the pointer to the discussion by Interverbal, I will have to take a look at it later.
@MJ: Again, why would I have to do that, when I’m not the one proposing any hypothesis at all?
I don’t have to demonstrate that the prevalence of ASD is some specific exact value. Prevalence of ASD, based on current methods and criteria, is approximately known (it’s not precisely 1% but something like that.) There’s no compelling/sound reason to think it was significantly different in the past, or that it varies a lot with age. Anyone who proposes the hypothesis that it was different in the past has an obligation to present evidence.
What the evidence I presented suggests, and what the NHS study tells us, is that the interpretation of data by those who propose the epidemic hypothesis is evidently mistaken, so the hypothesis is unfounded to being with, on top of the fact that it lacks real evidence.
Socrates that is an interesting quote you dragged up from St Michelle there, and something that would be a point of agreement between the both of us. I do not consider autism to be a genetic phenotype any more than Englishness is, but I still consider that there is a neurological substrate of varying and multiple etiologies that is essentially a naturally occuring but distinct neurological and cognitive configuration. That is as far as I will go so far as defining what I consider Autism to be beyond social construction and cultural identity.
What I do not consider it to be is what the NHS and any other prevalence surveys think they are measuring, they are at least a dimension short of that and so there figures will never be right whether they come up with 1 in 1000 or 1 in 10.
Why oh why do we find ourselves agreeing with the very people who have stereotyped and misrepresented us historically?
Let me explain it another way. The epidemic hypothesis is based mostly or entirely on the 2 following arguments:
1) There are increases in special education counts, CDC phone survey counts, etc.
2) If you compare an autism prevalence study from 2009 with one from the 1960s to the 1980s, you see a big discrepancy in the results.
Argument #1 is not compelling or sound because you can’t expect special education counts to be equivalent to a whole-population screening and assessment of autism. How many kids are actually recognized as being autistic probably depends on time-dependent variables such as awareness and so forth. It’s well known that autism is often missed.
Argument #2 is not compelling or sound because the studies look at different constructs (autism proper vs. autism + PDD-NOS + Asperger’s), they use different criteria, different ascertainment methodology and tools, and the researchers probably have very different subjective views of what autism looks like.
Despite this, there might still be a testable hypothesis there. How do you test it? A common suggestion is to estimate the whole-population prevalence of ASD in adults using modern methods.
That particular test has had its first major failure.
Another suggestion might be to try to replicate an old study using the very methods and criteria of the study. Logistically this might be more difficult, and it hasn’t been done.
But the hypothesis has had other low-profile failures. For example, another test of the hypothesis might be to see if ASD among intellectually disabled adults is much less common than among intellectually disabled children. This test has failed in my view a couple times.
An implicit assumption of the hypothesis is that autism was not missed terribly often in the past. This assumption can be tested as well, and this test has also failed, on multiple occasions.
I think there are a few more low-profile failures like that. I’d suggest this hypothesis has failed enough times already.
This, to me, is the part that most directly relates to the recent report on adults.
The vaccines-cause-autism groups are showing themselves to be trapped in a mindset and, really, advocating a position rather than searching for the truth.
Assume that vaccines cause autism in some cases. If you want to find the real answer, you work with all the data available.
There may be a true increase in autism rates. The data as of now can’t say one way or another. We do know there are major factors contributing to the increases observed. You need to weed those false signals out or you will never get a real answer.
These organizations doesn’t appear to care about what the answer may be. They are shooting themselves in the foot.
In the process of doing so, they are also lessening the chances that autistic adults (and their own children when they grow up) will get adequate supports. This isn’t because anyone is listening to them, but because they could and should be advocating for better supports. It is the lack of effort rather than their misguided efforts that count.
Do you agree/disagree with people or with the opinions of people, Larry?
That rather depends whether there actions are consistently all of a piece with there opinions.
Trouble is I can’t extradite them to stand trial as there are no prisons in Regione Caecorum 🙂
Thing is notwithstanding the unrepentant ones who do share some guilt as our oppressors it is never too late to change ones opinions.
This is escaping my point though, which is essentially why some individuals amongst any oppressed and devalued class buy into and accept the authority of those who have placed them in such a position, it’s the Stockholm syndrome basically.
How though can we be an honest movement when on the one hand we make various claims, but on the other go running to those old authorities for validation when there ideas have outrun there course.
Never mind the workers seizing the means of production, I want to see us joining the psychiatric survivors movement and siezing the commanding heights of ideation, the production of research and academic discourse, my hidden agenda. (whoops not hidden anymore is it)
Today a young girl died after a cervical cancer vaccination. This is a first in this country although there have been other fatalities in the US. How many will die from the swine flu jab? Can we really guarantee that the MMR jab is not implicated in the rise in autism?
Will Wakefield be justified? Are Stone and Tommey right to be concerned? Are the big pharmas reckless, irresponsible and greedy for even bigger profits at the expense of the public’s health?
Let’s see where we all are in a year’s time. There are still many twists in this story.
Fewer than will die of swine flu idiot!
No I don’t want to be a fatality in either case, nor do you, but life is ultimately a fatal disease, and it is a lottery who lives and who dies, this is reality and no amount of whining will stop the grim reaper, the US constitution may stand for life liberty and the pursuit of happiness but I see no promise of any of that being eternal.
Joseph, you said –
“I don’t have to demonstrate that the prevalence of ASD is some specific exact value. Prevalence of ASD, based on current methods and criteria, is approximately known (it’s not precisely 1% but something like that.) ”
I am saying that you have failed to demonstrate that there support for the idea that the prevalence in adults is anywhere close to 1 percent. None of the literature you cited estimates the prevalence in the general population of adults.
Therefore, you have failed to show support for the idea that one percent of adults have a form of autism.
“Let me explain it another way. The epidemic hypothesis is based mostly or entirely on the 2 following arguments:”
Joseph, I know you want this to be about the hypothesis that there is an epidemic of autism, but that isn’t want I was talking about. And the prevalence in adults is really a separate issue.
Actual all the literature is flawed by that very same charge, that it has failed historically to address the notion that autism is anything other than (as Lenin might have said) an infantile disorder.
We are all victims of history, because it was presumed in the days when autism was construed as an infantile form of schizophrenia that it was something that would not endure into adulthood. Essentially ‘science’ (as Joseph deifies this fairly recent fashion in human affairs) has been very blinkered and remains so.
Anyway lets drop the “epidemic” terminology, it isn’t accurate, if you want to dispute anything dispute whether the prevalence is increasing, and even if it is, that would not constitute an epidemic proper.
Laurentius Rex you said –
“MJ this is precisely where you are wrong. Autism is only classified as a medical disorder because of the historical consequences of how it came to be in our present lexicon.”
And here is where we will have to agree to disagree. Autism is clearly a medical disorder and not a historical consequence of language use.
“According to Autism Speaks we don’t even have the right to life let alone liberty and the pursuit of happiness.”
I have heard this claim many times but have yet to see any evidence for it.
Nonsense. The prevalence in children is approximately 1% at, say, age 7. Since diagnostic stability is good after that age, and life expectancy is near normal, adult prevalence should be about the same.
Oh, you’re asking if adult prevalence is about 1% for current adults? Any reason why it would not be the same? I fail to see it. In fact, a recent NHS survey confirms the prevalence for adults living currently is about 1%. Further, screening tools for adults behave similar to screening tools for children (plus all the other data, etc.)
Larry, I think you’re confusing the limitations of scientists, who are only human after all, with the scientific endeavor itself.
In fact, not all scientists have been like that. Lorna Wing and colleagues studied autism in adults in 1982. She later basically said that, even at the time, she was already thinking there’s an ‘autism spectrum’ with a prevalence near 1% of the population.
Research can be influenced, and scientists have biases, but sooner or later I believe the method produces real knowledge.
No, I think they care, Sullivan, except I don’t think they are intellectually curious about the answer. They simply want it to be their pre-ordained answer at all costs (in a literal sense.)
Joseph,
One observation: MJ appears to have difficulty understanding the difference between “evidence” or “support” and “absolute proof”.
The NHS survey is evidence that the prevalence of autism amongst adults is somewhere near 1%. It is even somewhat good evidence.
Now, it is possible that somehow they managed to put all 19 autistic adults into the phase-2 part of the study. The odds are quite low, but it is possible. This would have to be either through random chance, or their screening methods were amazingly good and they are misleading us about how they applied them.
She also gets really close to an important point and then shies away again:
There were 19 adult autsitics identified out of the phase-2 group of 618. That’s about 3%. But, again, it is possible that the 19 were all that were in the entire phase-1 group (highly unlikely, but possible). For the sake of argument, that gives a prevalence of 19 out of 7353, or 25 out of 10,000.
Since the counter argument is that the prevalence for this group should be 3.3 per 10,000 (as espoused by people citing an old ND–North Dakota–study), even the 25/10,000 is a big difference.
Like, a factor of 7.
It isn’t a question of whether there were factors like expansion of the criteria or diagnositic substitution that caused part or all of the increase in autism “rate”. The question is what are all the factors and how much do they contribute.
BTW, I had almost forgotten I wrote a post right here at LB/RB in May, 2008, where I basically criticized the same sort of evidence I listed in this discussion, thusly:
I argued that it was necessary to carry out more specific research aimed at countering a harmful hypothesis, assuming I was correct in my evaluation.
I made 4 suggestions. My second suggestion was a prevalence study of autism in adults.
My first suggestion was replicating Lotter (1967). I’ve changed my mind about this. What I would do instead is pick one of the better studies from the time period that can be feasibly replicated – one that doesn’t rely too much in, say, what teachers might think when you ask them to report kids who are “odd.”
My third suggestion was: Determine if large regional differences in administrative prevalence are real. I still think this is a good idea.
My forth suggestion was: Explain the increase in administrative counts with a mathematical model. It may or may not be doable. I actually have some data related to this that I’ve never discussed.
“One observation: MJ appears to have difficulty understanding the difference between “evidence” or “support” and “absolute proof”.”
No, I don’t have difficulty understanding that concept. Do you understand the difference between circumstantial and direct evidence?
I can make a circumstantial case that autism is caused in part by micronutrient deficiency, but I doubt that you would believe it.
Your favorite boogie men, the “anti-vaxers”, can make a very strong circumstantial case that vaccines can cause autism under the right conditions, yet you certainly don’t believe that.
The truly hilarious thing here is that you have almost the same religious devotion to the idea that the rate of autism is not increasing as do the “anti-vaxers” to the idea that autism is caused by vaccines. Yet, the evidence available does not support either theory.
The reason for your insistence is that all of your theories about autism would fall apart if there were in fact an increasing number of children with autism or if there were not as many adults as children. If it were increasing that would mean that it isn’t genetic (in the classical sense) and is likely being caused by some external factor.
Yet, like the “anti-vaxers”, you are so invested in your pet theories that you cannot take a step back and look at the evidence for what it is. I believe the proper term for this is “closed minded”.
So here is a an interesting though experiment for you, take a step back and look what would happen to your beliefs if there were in fact an epidemic of autism.
“The NHS survey is evidence that the prevalence of autism amongst adults is somewhere near 1%. It is even somewhat good evidence.”
I disagree with this for the reasons that I stated numerous times above.
“Now, it is possible that somehow they managed to put all 19 autistic adults into the phase-2 part of the study”
You really don’t get the concept of subgroups, do you? Let me try this one more time. Unless you know how the sub groups in phase 1 and phase 2 were constructed, and unless their properties are known the extrapolation back to the main population is an unknown.
“She also gets really close”
The proper pronoun in this case is “he”.
“There were 19 adult autsitics identified out of the phase-2 group of 618. That’s about 3%.”
Again with the misuse of percentages. I can pick any size group out a population and make the percent anything that I want by selecting the subset properly. If I have a group of 1,000 out of which 3 have autism I can make that seem to be 3 percent by selecting the proper sub group of 100.
“Since the counter argument is that the prevalence for this group should be 3.3 per 10,000 (as espoused by people citing an old ND—North Dakota—study), even the 25/10,000 is a big difference.”
You do know that the criteria changed, right? And then there was something about better recognition, I am almost positive that I have heard you say it before.
By the way, speaking of someone in the third person instead of directly addressing them is a classic passive-aggressive behavior.
“Nonsense. The prevalence in children is approximately 1% at, say, age 7. Since diagnostic stability is good after that age, and life expectancy is near normal, adult prevalence should be about the same.”
Good, produce direct evidence to back up your assertion. If this is the case then there should be no trouble producing results that show the prevalence of autism is 1 percent in adults.
Not possible, statistically, Sullivan, unless the NHS is lying to us with a straight face (maybe that’s what MJ thinks but doesn’t want to say.) Let me bore you with some more math…
Given the results, it’s obvious that autistics from phase 1 had ~3 times the probability of being selected compared to someone in the general population (hence the adjustment from 3% prevalence.) The average selection probability for phase 2 was 618 divided by 7461, or 0.083. The probability for autistics should have been 0.083 x 3 = 0.249 (in average.)
What are the odds that a specific group of 19 persons (19 out of 19) would be picked, assuming the probability that picking each one is 0.249. That’s 0.249^19 = 0.0000000000034. Not possible.
Whereas if there are 72 autistics in phase 1, and each has a probability of 0.249 of being picked, how many would you expect to pick? 17.9, give or take.
“that’s 0.249^19 = 0.0000000000034. Not possible.”
Strictly speaking, it is possible, just incredibly improbable. We are dealing with people who take offense at equating “incredibly improbable” with “not possible”.
You can’t prove a negative. But, you can show that it is so improbable that it is quite rational to say, “not possible”. Some don’t want to except that 3.5×10-12 is “not possible”.
I was having a bit of fun at those who are attacking this study on extremely weak grounds.
“it’s obvious that autistics from phase 1 had ~3 times the probability of being selected compared to someone in the general population”
Actually they had a much, much higher chance than that. I finally got a chance to reread the survey.
Look at the criteria used to select the set for phase 2 (section 3.2.6 forward). They administered four screening tests, one for Psychosis, ASD, Borderline personality disorder, and Antisocial personality disorder. The higher that you scored on any of these tests the more likely you were to be selected for phase 2, up to a probability of 1 (as in, 100 percent).
For example, the ASD test had scores from 0 to 12+, a score of 12+ would guarantee that you would be selected for the second phase, while a score of 10 only had a 25 percent chance.
So the design is even worse than I had thought – the second phase that was given the ADOS was made up on those MOST likely to have autism out of the entire population. Therefore I would expect that the rate found in this group would be much,much higher than the general population.
It would be like giving the MCHAT to a group of children, selecting the group that had the highest score, and then using that subset to estimate the prevalence for all children.
You are just figuring this out? Joseph pointed this out a long while back.
They got a raw prevalence of 3% and then claimed a prevalence of 1%.
Didn’t you read the study at all before you started commenting on it?
@ Robert: what country are you in? I can’t find any record (via google) of a recent death due to Gardasil. Of the deaths reported in the USA, I don’t believe any of them have been linked directly to the vaccine. (One death was a car accident, for heaven’s sake, which the father claimed she was involved in due to the vaccine). Remember, please that the VAERS database is open to anyone to report in, no matter how accurate their data is.
Either Kev or Sullivan reported that a vaccine turned them into the Hulk and the database accepted that report!
@MJ: I’m a little confused with your problem with the report, and maybe I’m not understanding what you are saying (the caffeine hasn’t kicked in yet).
I am not a researcher, but it seems to me if I want to find out the prevalence of anything, I would test a certain subset of the population, then out of that subset, determine who has the condition I am looking for. It isn’t financially possible for me to test the entire population, and I can’t test all of the subset completely so I need to use a screening test on on the first subset then a diagnostic test on the second.
@ Sullivan: do I have the idea as to how the study was created?
So what you’re saying, MJ, is that they just lied, or they made a big mistake that no one noticed, like not being able to tell that 1 divided by 0.1 is way more than 3.
Here’s the problem. Leaving aside the other questionnaires, which are probably not very good for screening, you’re basically assuming that the AQ-20 has excellent screening properties at a cut-off of 12+. But you don’t know that. The AQ-50 is known to have excellent screening properties at a cut-off of 26+. The corresponding AQ-20 cut-off should be 10.4+. However, it comes to reason that the AQ-20 is much less reliable than the AQ-50, so maybe the ideal cut-off there is 8+ or 9+. Perhaps there’s no ideal cut-off in the AQ-20 at all.
The average AQ-20 score of autistics in this survey was probably between 10 and 11. This seems a little low, if you try to match that with norming of the AQ-50. Again, the AQ-20 is not the AQ-50, plus, the way autistics are selected for AQ norming studies is very different to the way autistics were found in the NHS survey.
Hello friends –
The biggest problem with this type of discussion seems to be that the discussion points seem to be:
1) There is an epidemic and vaccines are the culprit.
2) There is no evidence of an epidemic.
It is the abandonment of critical thinking. What if we are observing an actual rise in prevalance that has nothing to do with vaccines?
This morning this paper hit pubmed.
Risk of Autism and Increasing Maternal and Paternal Age in a Large North American Population
Its results are largely in line with a slew of similar papers; having older parents means a greater likelyhood of having a child with autism. Other papers are available for more subtle cognitive measurements, or other disorders like bipoloar, or schizophrenia. There are a variety of mechanisms that could explain this, (I believe) many of them involve accumulated polymorphisms in progenitor reproductive cells.
It is relatively undeniable that the average age of a mother and father has increased in the past few decades in developed countries. It is a big, big change from how humans have been having babies for the past couple of tens of thousands of years, and how mammals have been doing things for tens of millions of years.
Does this paper from the UK really provide us with sufficient evidence to believe that all of the papers above found exactly the same risk associations, and all were wrong in exactly the same way?
I’m not sold on the innocence of vaccination in regards to our apparent explosion of a variety of autoimmune or other disorders, but there are plenty of ways we can have an actual increase in prevalance without vaccines being part of the problem.
– pD
What you call critical thinking appears indistinguishable to credulity, in my view, pD. Plus I have said the implications of this are much more broad than the anti-vax movement and the MMR vaccine, even though that’s the only thing the few media who have covered this seem to be interested in.
Plausible, but it seems it would have very limited impact in prevalence. We can probably do some math to see.
The NHS study does find a slight downward trend with age, but it’s not statistically significant, so technically there’s no need to even consider it. It’s interesting, though, so let’s see (table B2)…
[Age Group] [Prevalence]
16-44 1.1%
45-74 0.9%
75+ 0.8% (someone tell John Best)
Could it be because of increasing father’s age? Maybe. But a much simpler explanation is that autistics have a slightly lower life expectancy. Autistics tend to be male, single, some having epilepsy, etc.
@ Joseph – I seem to remember that an INCREASE in autism was shown with increasing parental age, from a study done in Israel. Or is my memory faulty?
@Dawn: Yes. The new CalDDS study shows the same thing, except it finds that both maternal and paternal age are factors. The Israeli study, as I recall, claimed that mother’s age was not a factor, after controlling for father’s age.
BTW, pD’s link does not take you to the right paper, which is here.
I realize now the table I posted might have been confusing. It’s not related to the study pD linked. It’s from the NHS adult prevalence study. Older adults presumably had younger parents at birth.
Hi Joseph –
Thanks for fixing my link. I had that fix waiting in the wings.
Heh.
But we do have evidence for advancing parental age affects likelyhood of having a child on the spectrum. I only posted a couple of papers on this in my previous post, there are many others.
Advanced parental age at birth is associated with poorer social functioning in adolescent males: shedding light on a core symptom of schizophrenia and autism
Maternal and paternal age and risk of autism spectrum disorders
Parental age and risk of schizophrenia: a case-control study
I doubt anyone would deny that the average of motherhood has undergone a gradual increase in the past decades. ( ? ) Do we really have any reason to think that all of these studies are inaccurate other than simply accusing me of coming to conclusions with little evidence?
In any case, the problem lies in that if we are to really believe that there has been no increase in autism, as suggested by this paper; then none of our newly found external environmental forces our mothers and infants are subjected to are having any effect. Increased stress, increased age, increased exposure to endocrine disrupters, pesticides, and sythentic chemicals, all have been having no effect.
The problem with believing this, no matter how comforting, is that at a clinical level, we know that exposure to many of these things can cause physiological or behavioral manifestations similar to what we observe in autism. Asking people if they prefer to read fiction or non fiction as a means to determine if they might be on the spectrum is, indeed, a credulous way to accomplish this.
It isn’t that the vaccine hypothesis needs to be wrong to support a true increase; it is, in fact, that a veritable spectrum of environmental exposures are all having no effect, despite mountains of applied evidence to the contrary, and our evidence for this is a handful of studies that involve miniscule numbers of individuals that answered questions like : “Do you prefer reading fiction or non fiction” as a means to determine placement on the spectrum. That is my definition of credulity.
– pD
@pD: It might be no effect. The weight of the evidence is not yet convincing for any of those risk factors, even though some might be plausible. Alternatively, they might have an unnoticeable effect, one that can’t really be seen in prevalence figures.
The remaining problem with the parental age factor, which is indeed fairly well replicated, is one specific confound that has not been looked into yet: Whether autistic characteristics in the parents might correlate with age of procreation. It would probably be expensive to do, though.
Is that a line from AoA about the NHS study? A commenter said the same thing over at RI. The AQ-20 indeed has a question about whether you enjoy reading fiction. If you answer ‘no’, that’s one point toward ‘more autistic,’ contrary to what people seem to be thinking. I answer ‘slightly disagree’ there, incidentally, as I enjoy reading non-fiction a lot more.
It seems a little odd, but they ran the questions through statistical analysis in pre-trials. I’m assuming the question performed reasonably, so they kept it. I doubt the researchers are complete jackasses or anything like that.
It’s very naive (credulous perhaps 🙂 to suppose that every question of a screening instrument has to have a precise correlation with what the tool measures. It’s sufficient that it trends a little.
Sullivan you said –
“You are just figuring this out? Joseph pointed this out a long while back”
Really, where? Because all I found was him saying that phase 2 selection was semi-random.
“Didn’t you read the study at all before you started commenting on it?”
I did, but apparently you did not read my comment before commenting on it. I believe I said “I finally got a chance to reread the survey.” Reread implies reading for a second (or third) time.
Joseph, you said –
“you’re basically assuming that the AQ-20 has excellent screening properties at a cut-off of 12+. But you don’t know that.”
Actually that is part of what I have been saying from the start – we do not know how well this screening test works. Therefore we don’t have any good way of going from the subset back to the original population. Go back and read some of my earlier comments.
“The AQ-50 is known to have excellent screening properties at a cut-off of 26+. The corresponding AQ-20 cut-off should be 10.4+”
You think you can apply a linear transformation on a score simply on the basis of the number of questions? Even though the wording of the actual questions changed? It does not work that way.
“The average AQ-20 score of autistics in this survey was probably between 10 and 11”
And how did you arrive at this number?
Dawn, you said –
“I’m a little confused with your problem with the report, and maybe I’m not understanding what you are saying (the caffeine hasn’t kicked in yet).”
Basically there are two problems.
First, we have no idea how good of a job it does at selecting people with autism. If it is very good that would mean that it would find almost everyone with autism – if it were very bad it would basically miss everyone with autism. Either way, if the test selects a group of 10 out of 100 and we then use a second level and find out there are 3 in the 10, does that mean that we expect the 3 to be entire population of people with autism (screen very good), or that there are likely others who did not make it into the initial 10 and as a result there are some others (screen so-so), or that the 10 selected were basically random which would mean that 3 percent of the 100 would have autism.
This is the first time this test has been used so we have no data on how well it works.
Second, the group of 600 for phase 2 was composed of what the researchers thought were the people most likely to have autism. This group accounts for 8 percent of the initial population which means that, if the screen was even somewhat decent, the majority (if not almost all) of the people with autism should be in this group. Yet, after setting up the groups this way the researchers assert that most of the people with autism ( ~ 70 percent) were in fact not in this group but were instead of the set not selected. I think that is unlikely given how the groups were constructed.
I wrote about this elsewhere, so if you wanted more details click on my name on this comment and you will find it.
And I explained this was practically irrelevant. We all know the AQ-20 is an untested tool (at least untested to the extent that we can read about the testing in other papers), probably unreliable. We all know the other questionnaires are probably untested, probably unreliable.
That’s what’s interesting about the methodology. It’s designed to work with practically untested, made-up ‘screening’ methods that could very well miss a large portion of the population you want to find.
The ADOS assessment applied to the 600+ semi-randomly selected people is the part that really matters in so far as finding ASD.
They obviously didn’t think they had any screening method that would not miss any autistics of any kind. So they came up with that.
I grant that the methodology could be challenged on the grounds of its being novel (if in fact it is.) I do find it pretty clever, though.
26% ‘screening’ would be very bad if they had actually been using proven screening tools. Apparently they had fairly poor tools, and they wanted to limit the work in phase 1. Furthermore, they couldn’t have known what an appropriate cut-off should be. I mean, they might have had an idea from testing of diagnosed autistics who volunteer for AQ testing, but they couldn’t have known how general-population, probably undiagnosed autistics would score in it. Evidently, most of them must have scored lower than 12, unless the results are a fabrication.
If autistics simply scored 10 in average, that drops the ‘screening’ rate to 25%, and we have no way to judge that this is ‘unlikely.’ The paper results tell us this is pretty much what happened. Anything else is just speculation that we’re being defrauded by the UK’s NHS.
Hello MJ and Joseph –
I am enjoying reading your discussion, though it does remind me of the areas I am very weak in as opposed to just modestly weak.
I wish that the type of nuance being described here were translatable to the media at large, or even other blogs were this might be discussed. It is very useful contextual information compared to the take away message from the headlines. This seems to be a common theme.
Thanks.
– pD
Joseph, you said –
“That’s what’s interesting about the methodology. It’s designed to work with practically untested, made-up ‘screening’ methods that could very well miss a large portion of the population you want to find.”
No, there is no designing around a selection process that skews your selections in an unknown manner. It would have been far better if the selections for phase 2 had been random – the results could have been generalized in an appropriate manner back to the original population.
“The ADOS assessment applied to the 600+ semi-randomly selected people is the part that really matters in so far as finding ASD.”
You have also said that you believe that people with autism will have an average score that is considered high. How can you reconcile calling a test “semi-random” while saying at the same time that the group you are looking for will score, on average, in a range that is targeted in the selection of the sets? Wouldn’t that imply that the test isn’t semi-random?
“They obviously didn’t think they had any screening method that would not miss any autistics of any kind.”
No, they state quite clearly that they felt the original AQ would be too long (time wise) to use, otherwise they would have used it. If they had then we would have a better idea how many people with autism would left in the general population after the phase 2 screening.
“they wanted to limit the work in phase 1”
There screening in phase 1 only eliminated about a third of the population. It is only when the ranked the people who were eligible for phase 2 by their scores that they limited down to 800. So if they goal was to limit work it didn’t go a good job.
“Furthermore, they couldn’t have known what an appropriate cut-off should be”
Which is why they used probabilities instead of a cutoff.
“but they couldn’t have known how general-population, probably undiagnosed autistics would score in it.”
Which is why a known screening tool should have been used.
Do you have any reason to think that people with undiagnosed autism would score differently from diagnosed ones?
“Evidently, most of them must have scored lower than 12, unless the results are a fabrication.”
I find it odd that the survey does not include a distribution of the scores on the screening test nor any sort of analysis of how well the screen identified people with autism. This is particularly strange in light of the fact that the entire results rests on the extrapolation of the small result back to the general population. Maybe it is in there and I missed it.
“If autistics simply scored 10 in average, that drops the ‘screening’ rate to 25%”
How do you arrive at that number? From the probabilities published having a 10 would give you a 25 percent change of being included while an 11 gives you a 61 and anything 12 or over is 100 percent. But, the screening rate (I assume you are referring to selectivity and sensitivity) in total would not be 25 percent. In general, you would expect a lower cutoff would yield find a higher the percentage of the total population that you are looking for would be included. This is clearly at the expense of more false positives, but the point is that you will get most, if not all, of the population you are looking for in your set.
And this appears to be exactly what was done with this test – they lowered the threshold so that the phase 2 set was as large as possible and then selected the most likely to test with the ADOS. Since the phase 2 set that was tested (8 percent) was larger than any possible set of people with autism that means that most of the people with autism should have been in this set. Yet the researchers tell us that 70 percent weren’t in this set.
The problem here is that the screening test is a major unknown – but it is in no form any form of semi-random selection. It was designed with a very clear intent and goal – to identify as many people with autism as possible.
“Anything else is just speculation that we’re being defrauded by the UK’s NHS.”
I find it interesting that while I am suggesting that the methods employed by the researchers might have been flawed and thus the conclusions questionable you use the word “defrauded”. I am not implying that, so are you saying that you think there may have been fraud involved?
You’re entirely mistaken about this, MJ. The methodology uses stratified probabilities, and it doesn’t matter how the probabilities are distributed or anything like that. I explain how the method works the best I can in this comment. It’s quite simple and sound.
It’s clear that it wouldn’t matter methodologically if the ‘screening’ method involved asking people questions like ‘is your sister overweight?’ It would work exactly the same way, except a poor screening method would have an impact on statistical power, no doubt.
Yes. I don’t know exactly how they recruit autistic adults for AQ studies, but I would imagine they are mostly ones diagnosed with Asperger’s, without intellectual disability, who already know some things about ASD. Autistics in the general population might have more intellectual disability, might be less able to fill out questionnaires as the testers intend, are probably unaware of ASD, and might be reluctant to be labeled as having psychiatric issues.
I believe you are thinking that. Why? I don’t think it’s possible that knowing only 19 out of 72 autistics were successfully screened, they would be able to reconcile that with the fact that their screening methods were great. If they were great, they would immediately know their estimate is in error, unless they are trying to deceive. Therefore, either their screening methods turned out to be poor (which they might not have expected a priori), or they are trying to deceive. You apparently believe their screening methods were not that poor.
“The methodology uses stratified probabilities, and it doesn’t matter how the probabilities are distributed or anything like that.”
That still does not get around the unknowns of the screening test. You cannot use the results of the screen to apply the results back to the general population, it simply doesn’t work. You can’t say that X people with score Y had autism therefore I would expect that people with score Y would have autism because we don’t know that this is an actual relation. We only know that in this sample people with score Y had autism, we have no clue if that relation holds for the rest of the population.
So it doesn’t matter how many groups you slice your population into, it doesn’t work.
“It’s clear that it wouldn’t matter methodologically if the ‘screening’ method involved asking people questions like ‘is your sister overweight?’”
You are confusing a random question with ones specifically designed to identify autism. They don’t ask if your sister is overweight they asked things like “I prefer to do things the same way over and over again”.
“I don’t think it’s possible that knowing only 19 out of 72 autistics were successfully screened”
You are putting the cart in front on the horse. They don’t know that there were 72 people with autism in the group, they estimated that there were. How did they estimate? By applying what they found in the sub group back to the main group which required knowledge of how well their screen worked. And how do they know how well their screen worked? Because it didn’t find extra 53 people with autism, a figure which they arrived by by assuming their screen didn’t work…. and around and around we go.
“You apparently believe their screening methods were not that poor”
I don’t know how well the methods worked, which is the point really.
The results of the extrapolation to 1 percent cannot be taken at face value, they are just too weak.
Did you try solving the simple exercise I proposed in the other thread? I honestly can’t do much more than that to convince you that the ‘extrapolation’ math is actually simple and sound.
A summarized one: If 10 of the autistics identified with the ADOS in phase 2 had all been assigned a selection probability of 0.25, how many autistics (known and unknown) would we expect were assigned a probability of 0.25 in phase 1?
Joseph, I responded to your simple exercise in the other thread. My concerns are still valid, you have a chicken and an egg and one of then has to come first.
The Big Pharma shills are thick on this page. If you have to resort to name-calling, wild derogatory associations, unsubstantiated general criticisms, crass lies and relying heavily on the proven liar/shill Deer… these strategies don’t reflect well on the strength of any of your “arguments”.
When these behaviors and flip attitude are heaped on the serious subject of what is affecting approximately 1% of vaccinated children the shills go beyond ludicrous to heinous.
Stop the inuendo, lies, intolerance and start dealing in facts. Consider the plight of millions affected directly by autism and the millions more that we could have prevented if vaccines and a low-percentage pre-disposition/susceptibility are to blame.
Michael, did you read what you wrote?
First you used the Pharma Shill Gambit, followed by saying:
Using the Pharma Shill gambit is name-calling, wild derogatory associations, unsubstantiated general criticisms!
If you have any real evidence that vaccines cause autism, please feel free to share. We would be very grateful if you would show us the real research that a vaccine that has been used in the USA since 1971, the MMR, starting to cause an increase in autism over a decade before it was introduced in the UK.
Chris,
you beat me to it.
Want to be this was another drive-by comment (i.e. Michael will never be seen again?)
Michael, the bloggers and commenters on this site are largely parents of autstics. Some are autistic themselves (I am not). We certainly consider the “plight” of those affected by autism.