Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site terak.UUCP
Path: utzoo!decvax!decwrl!Glacier!oliveb!hplabs!hao!noao!terak!doug
From: doug@terak.UUCP (Doug Pardee)
Newsgroups: net.med
Subject: Re: Risk factors
Message-ID: <822@terak.UUCP>
Date: Fri, 25-Oct-85 14:49:21 EST
Article-I.D.: terak.822
Posted: Fri Oct 25 14:49:21 1985
Date-Received: Tue, 29-Oct-85 15:11:41 EST
References: <1939@aecom.UUCP> <796@terak.UUCP> <1955@aecom.UUCP>
Distribution: na
Organization: Calcomp Display Products Division, Scottsdale, AZ, USA
Lines: 36

[It's hard to argue with someone who knows what he's talking about]

> For instance, in american males, heart disease is related to smoking both by
> the number of years smoked and number of packs (it can be expressed as pack-
> years.) So quitting smoking does help.

But, but, but...

The conclusion does not *necessarily* follow.  I do accept the
conclusion, but I don't believe that it has been properly proven.  It
is the method of "proof" that I'm questioning.

As a counter-theory, try this:

The kind of person who smokes is often a "Type A" person.
The kind of person who has heart disease is often a "Type A" person.

If we express the current theory as
  Type A --> smoking --> heart disease
then we can express an alternate theory as
  Type A --> other factor --> heart disease
  Type A --> smoking

Under this alternate theory, people who quit being Type A quit smoking
*and* reduce the risk of heart disease.  This gives the effect which has
been observed in population studies.  But if a person quits smoking
in order to avoid heart disease *when he otherwise would continue* and
remains a Type A person, there would be no effect.

It is *very* difficult to set up a properly controlled experiment when
the variable to be introduced is human behavior.  It is equally
difficult to draw reliable conclusions from population studies when the
independent variable is human behavior.  Human behavior is too often
not (maybe never is?) an "independent" variable.
-- 
Doug Pardee -- CalComp -- {calcom1,savax,seismo,decvax,ihnp4}!terak!doug