Friday, March 10, 2006

New Way to Explain Why an Epidemic Cannot be Inferred

We've all heard the argument posted at GenerationRescue's site which is often repeated by those who claim there's an autism epidemic:

The incidence of autism has increased from 1 in 10,000 in the 1970s to 1 in 150 today, an increase of over 6,000%. Many more children have been diagnosed with other neurodevelopmental disorders all considered to be on the same spectrum including Asperger's, ADHD/ADD, speech delay, and many other developmental delays and learning disabilities.


The refutation of this argument is well-known and straight forward: The prevalence of 1 in 150 includes Asperger's syndrome (no speech delay) whereas the definition of autism in the 1970s did not. Another important point is that these numbers represent prevalence, not incidence. In general, a formal way to refute it is to point out that the groups are not equivalent.

Pointing out that the groups are not equivalent means that the prevalence of traits in one group is not comparable to that of the other group. For example, in one group, prevalence of speech delay is much higher than in the other. Prevalence of mental retardation (according to Weschler) is probably much higher in the 1970s group.

In simpler terms, you can't compare the prevalence of apples to the prevalence of oranges, and then claim that the incidence of fruits has changed.

General application of this principle

I believe it's possible to make this argument from quarter to quarter, and from year to year. For example, the group of autistics in the CDDS in December 2002 is not equivalent to the group of autistics in the CDDS in December 2003. And this is demonstrably true. I'll list four groups below indicating trait proportions; namely, incidence of epilepsy and profound mental retardation (PMR) among autistics:

Group 1 - Dec. 2002: Epilepsy: 7.80%, PMR: 3.23%
Group 2 - Dec. 2003: Epilepsy: 7.42%, PMR: 2.86%
Group 3 - Dec. 2004: Epilepsy: 7.10%, PMR: 2.59%
Group 4 - Dec. 2005: Epilepsy: 6.73%, PMR: 2.35%

Not only is it crystal clear that these 4 groups are not equivalent, it would appear that autistics in the CDDS keep getting higher and higher functioning with every year that passes, with no end in sight. If this continues indefinitely, the CDDS will have to rename the 'Autism' category to 'NT'.

Conclusion

Claiming that the incidence of autism is increasing based on CDDS prevalence data suffers from a critical flaw. No such conclusion can be derived from the data, given that the group of autistics in one quarter is not equivalent to the group of autistics the next quarter.

5 comments:

  1. Joseph,

    Just a thought here; I have nothing to suggest that this is the case. Assuming that the incidence of ASD, in numerical terms, has stayed stable, could the increased diagnosis result from more cases being qualitatively more profound than was seen in the past? And if that is the case, could the qualitative difference be linked to environmental factors?

    ReplyDelete
  2. Assuming that the incidence of ASD, in numerical terms, has stayed stable, could the increased diagnosis result from more cases being qualitatively more profound than was seen in the past? And if that is the case, could the qualitative difference be linked to environmental factors?

    I think what you're saying is that autism has become milder with time, not so much that there's an 'expanding definition'?

    I've thought about that, and that would make sense with conditions that do not occur in a spectum with infinite levels of severity. Let me explain.

    Instead of just epilepsy or MR, suppose the equivalence marker is an objective measure of speech delay, repetitive behaviors, and social skills. Let's say this is a score in a test such as the AQ test (but accurate).

    I predict that the proportion of autistics who score above a threshold in this test would also go down from quarter to quarter. (Not far fetched, is it?)

    For this to happen, the 'extra' autistics in one quarter compared to the last quarter would have to score considerably lower than the previous average in this test. Right?

    If they do, that's exactly the definition of 'broadening criteria'. That is, the eligibility criteria of the CDDS would've expanded to include people who score fairly low in the autism scale.

    Now, markers such as epilepsy and MR are not as good, but it's not far fetched to suppose they work the same way, because there is significant correlation between those markers and autism.

    ReplyDelete
  3. BTW, this posting by itself does not show that environmental factors are not at play at all. But it does show that broadening criteria could easily be sufficient to account for the changing prevalence.

    ReplyDelete
  4. "I think what you're saying is that autism has become milder with time, not so much that there's an 'expanding definition'?"

    Not exactly. The question I'm trying to ask, rather imprecisely I'm afraid, is whether individual cases of autism might be relatively “mild” without the interaction of a environmental insult, and then show increased clinical indications associated with a diagnosis of autism as a result of the insult.

    The perception of an epidemic ⎯ and I am one who shares in that perception ⎯ is based not just on the statistics but also on casual (and unabashedly unscientific) observation. Had our children just shown “quirkiness,” it would never have occurred to us to seek any diagnosis. I wm wondering how much of the increase in the numbers is due to a possible shift in cases from one end of the spectrum to the other.

    Again, I am basing this conjecture on casual observation. If I understand your point, I believe you correctly point out a problem that makes it impossible to make any real sense of the numbers: the diagnostic merging of related but different points of the spectrum.

    ReplyDelete
  5. Not exactly. The question I'm trying to ask, rather imprecisely I'm afraid, is whether individual cases of autism might be relatively “mild” without the interaction of a environmental insult, and then show increased clinical indications associated with a diagnosis of autism as a result of the insult.


    Let's suppose that criteria is not broadened, but there is an environmental trigger that is more prevalent as time goes by. The trigger does not cause autism but just makes it more noticeable. What would happen, IMHO, is that the autistic population would grow, but it would be roughly equivalent from quarter to quarter.
    That is, the diagnostic threshold does not shift, so there's no reason for the population to become more or less autistic in average, even though it can grow in size. (There will always be people who are "almost autistic" who will cross the threshold with the trigger).

    The perception of an epidemic ⎯ and I am one who shares in that perception ⎯ is based not just on the statistics but also on casual (and unabashedly unscientific) observation. Had our children just shown “quirkiness,” it would never have occurred to us to seek any diagnosis. I wm wondering how much of the increase in the numbers is due to a possible shift in cases from one end of the spectrum to the other.


    It's hard to tell how previous generations would have seen our kids. I'd like to see video of my dad as a child, for example. All I know is that they worried he'd be mute and had some difficulties adjusting to school at first. I don't think he was ever taken to a psychiatrist. And there was certainly not a flood of information at the time indicating to the parents that they must panic.

    Again, I am basing this conjecture on casual observation. If I understand your point, I believe you correctly point out a problem that makes it impossible to make any real sense of the numbers: the diagnostic merging of related but different points of the spectrum.

    And it's not really about Aspergers. The shift is occurring throughout the spectrum.

    ReplyDelete