When does statistics work? - new drugs

[Paul Hollingworth <PH@paulholl.demon.co.uk>]

David Kerridge <dfkerridge@mac.com> writes
>Yet in other cases, such as testing a new drug, we can, with care, get
>results which reliably predict whether the drug will work in future.

I have to take issue with David on this point.

Careful tests and analyses may predict, with estimated risk of error,
what main effects a drug should have on a typical person. I do not
accept such studies reliably prove that a drug 'works'. What I am taking
issue with here is not the use of statistics in clinical trials, but the
widely held assumption that drugs 'work' from the viewpoint of health. I
don't believe drug companies are motivated purely by 'health'.

In the year 2001 scientists are still discovering hitherto unknown
effects (both 'good' and 'bad') from use of the 'simple' drug aspirin. 
Human physiology is so complex and conditions of medicine use so varied
that these effects can NOT be predicted from a trial no matter how
carefully it is carried out. Someone has to decide if a result is "good
enough". This is not the job of the statistician.

"The final decision on whether a proposed frame covers enough of the
universe belongs to the client"
        WED: Sample Design in Business Research

Regards
Paul H
----------------------------------------------------------------
Paul Hollingworth                       4GM Consulting Ltd.
email: PH@4GM.com                       http://www.4GM.com
phone: +44(0) 1423 322225               fax: +44(0) 1423 322205 
----------------------------------------------------------------