-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sun, 05 Jun 2005 14:12:08 -0700
Andrew Lentvorski <[EMAIL PROTECTED]> wrote:

> David J. Looney wrote:
> 
> > Studies from which you can draw
> > conclusions don't necessarily have to be randomized, blinded, and
> > controlled, but case studies need carefully chosen controls in order
> > to be meaningful.
> 
> Not necessarily.
> 
> Statistics is often used to tease out the independent variables in 
> "messy" data.
> 
> Much of the original AIDS work and identification was done in this
> way.

Statistical anomalies are only anomalous with respect to some norm. 
Data mining is most likely to work only if the data is not biased. The
"control" in this case was historical. These very rare diseases
(Kaposi's sarcoma and Pneumocystis carinii pneumonia) were simply never
seen in young men without leukemia, etc., in the U.S. before.  It
was immediately obvious to most that something new and different
was happening. That it was infectious, and a virus, and sexually
transmitted were all good hypotheses. But to prove that HIV was
associated with AIDS, there had to be cases and controls. 

> It is less reliable, but sometimes you need an answer *now*.  Waiting 
> was unacceptable.  People began forming hypotheses and adjusting 
> behavior long before people knew exactly what was going on.

True, waiting was unacceptable.  Part of the response to any emergency
is to be  to determine what can be done to ameliorate risk, based on
inadequate data and a large number of potential causes.  In
many communities of men who have sex with men, there was a high
risk of sexually transmitted infection, a high frequency of
havning multiple partners, and also a higher prevalence of use
of some drugs than the general community (a control), which didn't seem
to be getting these unusual infections. Certainly it was not
unreasonable to say, "Hey, this may be infectious, maybe we ought to
close the bath houses until we get more data and find out for sure ! 
And maybe we ought to stop using drugs.  And we should should
be using condoms anyway."

But an important part "not waiting" was to approach the problem in a
controlled fashion. Many of those initial hypotheses were dead wrong
(e.g. amyl nitrite, immune suppression due to exposure to
histocompatibility antigens from multiple partners). Some people acted
on anecdotal data by limiting themselves to just a few sexual partners,
or discontinuing use of poppers while continuing to have unprotected
sex.  This was unwise.

To test the hypotheses, controlled data was needed.

> Often,
> conclusions  drawn on fairly thin evidence become embedded as facts in
> the medical  community.

I agree. They are only exorcised by studies which use controlled
data.  BTW, in what enlightened community does this not occur ?

David Looney

- -- 
Life is short, and the Art long; the occasion fleeting; experience
fallacious, and judgment difficult. - Hippocrates
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)

iD8DBQFCo31vZnDUcKSydkURAn1iAJ4pRs9ztZnUux+1uoJyuZpOhU4/9wCcCac/
o98iMyp1HcWfr6mRrOoI6gg=
=3twO
-----END PGP SIGNATURE-----


-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to