🚀 go-pugleaf

On Tue, 04 Apr 2006 20:41:05 -0700, The Real Bev
<bashley101+usenet@gmail.com> wrote in message
<g%GYf.293$rd2.59@fe05.lga>

>Andy wrote:
>
>> The problem with all studies that say that activity X causes a 240% (or
>> whatever percent) increase in the risk of cancer (or some other health
>> consequence) is that this  is really useless information.  If the
>> underlying risk is 1 in 100,000 a 240% increase means that the modified
>> risk is only 2.4 in 100,000, which, for all practical purposes, are the
>> same odds.  On the other hand, if the underlying odds 1 out of 5, then
>> a 240% increase is really significant.
>
>Nobody ever teaches doctors about statistics.

A lot of research facilities use medical statisticians to handle this,
rather than trying to use cookbook statistics.

One real problem that comes up is ethical.  There is always a clause
in the protocol for medical experiments which allows the doctor to
yank a patient from the study if he believes it is in the best
interests of the patient to do so.  This is a requirement of the
ethics committee, but can really screw up the statistics, as the
sample set could then become skewed (no longer random, etc.)

There may also be medically unnecessary tests involved, especially in
dose-related studies, sometimes dangerous tests.  For example, liver
biopsies performed to determine the effect of various doses of a given
drug have in treatment of Hepatitis C.