| George Conklin 2004-10-04, 2:21 am |
|
"Leonard Evens" <len@math.northwestern.edu> wrote in message
news:hsidnRweUbTcpsHcRVn-hA@comcast.com...
> George Conklin wrote:
far.[vbcol=seagreen]
>
> Actually not all the study results are what you call negative. The
> Swdish study recently reported in the NEJM found a statistically
> significant lower death rate from prostate cancer in a group treated by
> RP than that in a group treated by watchful waiting followed by hormone
> therapy where needed. I'm sure you will also point out that it showed
> no statistically significant difference in overall death rates. But
> the follow up period was only an average of about six years. So the
> information from this study is mixed. It is also of limited use for
> comparison with US populations for a variety of reasons.
>
Life is the issue, not what you die of. Remember in that study it was
hard to judge what people really died of and it was a committee assignment.
> As far as the studies in the US, since they are not yet complete, one
> can't draw any conclusions from them. However, the followup period is
> about 12 years. It is then reasonable to suggest that whatever result
> they come up with, you would only be safe in using that result if you
> expect only to live about 12 years.
You seem to extrapolate your own fears here. Why?
One has to be careful about
> generalizing any study to a population which differs significantly from
> the study population. You may say I am trying to prepare in advance
> for results showing no effect of treatment, but I think I am just
> pointing out the obvious. Your argument is that if no effect shows up
> in time T than surely nothing much is going to change after that. I am
> always suspicious of people who are sure about things they don't
> actually know. I argue that this conclusion follows neither logically
> or statistically. Moreover, there have been several prostate cancer
> studies which have been shown to be flawed because the chose too short a
> time line. That is, there was no significant effect during the chosen
> time horizon but later research on the same study population showed a
> significant effect with a longer time horizon. So not only can this
> sort of thing happen, but in the real world it has happened.
So if all studies you have a flawed, then why do you insist on dramatic
'cures' which have never been proven? If the results were going to be even
a little bit important, the results would show up long before 15 years.
|