It is not clear to me whether Stone and Heptonstall have genuinely
not understood the point I have been trying to make, or whether they are
just going off on tangents to distract attention from the flaw I pointed
out in their original arguments. If the latter, I don't suppose anything I
say will make much difference, but if the former, then I'll try once more
to explain.
The thrust of their argument seems to be that reporting of autism in
the GPRD is incomplete. I have no idea whether that is true or not. But,
and here is the really important thing, incomplete reporting of autism
would make no difference to the interpretation of Smeeth at al's study.
If we assume that recording of autism is incomplete, the chances are
that there will be some children in the control group who are autistic.
The thing is, and again, this is important, there will be very few of
them, because autism is rare. Even if under-reporting of autism is so
atrocious that most cases of autism are unrecorded in the GPRD, there is
still no reason whatever to think that the control group of Smeeth et al's
study will have a higher prevalence of autism than the general population.
Even Heptonstall now seems to accept that fewer than 1% of the general
population are autistic. If 1% of the control group in Smeeth et al's
study were misclassified, it is hard to see how that would make much
difference.
As far as I can tell, Heptonstall's argument now seems to be based on
the idea that the control group of Smeeth et al's study could somehow have
a much higher prevalence than this. Why? What mechanism could possibly
account for over-representation of children with autism in the control
group?
Rapid Response:
Response to the two Johns
It is not clear to me whether Stone and Heptonstall have genuinely
not understood the point I have been trying to make, or whether they are
just going off on tangents to distract attention from the flaw I pointed
out in their original arguments. If the latter, I don't suppose anything I
say will make much difference, but if the former, then I'll try once more
to explain.
The thrust of their argument seems to be that reporting of autism in
the GPRD is incomplete. I have no idea whether that is true or not. But,
and here is the really important thing, incomplete reporting of autism
would make no difference to the interpretation of Smeeth at al's study.
If we assume that recording of autism is incomplete, the chances are
that there will be some children in the control group who are autistic.
The thing is, and again, this is important, there will be very few of
them, because autism is rare. Even if under-reporting of autism is so
atrocious that most cases of autism are unrecorded in the GPRD, there is
still no reason whatever to think that the control group of Smeeth et al's
study will have a higher prevalence of autism than the general population.
Even Heptonstall now seems to accept that fewer than 1% of the general
population are autistic. If 1% of the control group in Smeeth et al's
study were misclassified, it is hard to see how that would make much
difference.
As far as I can tell, Heptonstall's argument now seems to be based on
the idea that the control group of Smeeth et al's study could somehow have
a much higher prevalence than this. Why? What mechanism could possibly
account for over-representation of children with autism in the control
group?
Competing interests:
As stated previously
Competing interests: No competing interests