Re: what methods to use?


"Richard Ulrich" <Rich.Ulrich@xxxxxxxxxxx> wrote in message
news:cfq733d31oltgc5k33vcsfkagpa6j93pjd@xxxxxxxxxx
On 28 Apr 2007 12:08:16 -0700, franco <franco@xxxxxxxx> wrote:

This is a naive question, not really about specific statistical issues
but more on what procedure to use to investigate something, for me to
learn.
Suppose we have a very large population exposed to a certain product
and suppose that the product can cause a certain specific reaction in
a subpopulation of individuals who are genetically predetermined to
have an interaction with the product (it's a hypothesis we need to
test). Let's say we have indivdual data on various reactions from
those who have been exposed (those reactions may or may not be caused
by the exposure but they happen to occur at various time-points
following the exposure). We have no extra data on the population. We
don't know what is the specific reaction we are looking for (the issue
is to identify it), it could be rare or common. Can we detect such
interactions by using statistical methods? And what would the best
methods be? Thanks.

When folks survey for side effects to vaccinations, etc.,
the universe of side effects is first tabulated for anything
suspicious. In futher investigation, a narrow window in time
for the effect is one thing that is particularly suspicious.
Any kind of specificity helps, like the previous rareness of
the reaction.

I'm thinking of Swine flu vaccine, for instance, and Guillane-Barr
disease. Those (rare) diagnoses were several weeks later.

Also, thalidimide was an extremely popular tranquilizer in Europe,
and it happened to have a small window of bad effect of several days
in the age of the fetus, when the limb-buds were developing.

Thalidomide, in case anyone wants to do a search


There can be a lot of false positives in what gets noted for
side effects for popular long-term prescriptions -- where the time
window may be hard to define -- if the only penalty is a
tiny note on the drug-info *** that says, for example, "Stop
taking if it causes nausea."

Nick
UK


.