I am as much angered by the abuse of statistics as the next man,
probably more so, but I can't go along with the conclusions here.  I am
no apologist for the pharmaceutical industry and believe that it is
guilty of some serious abuses, for example the creation of new syndromes
in order to sell drugs to 'cure' them, or using hard-sell techniques to
distort doctors clinical judgement, but don't agree that they have been
correctly identified here.

In the analysis, you often refer to the normal distribution as being an
inappropriate model, but this seems to be a matter of faith, I don't see
any reasons given, any explanation of what distribution does apply, or
how the different model affects the validity of the formula for (eg)
confidence interval.

Of the conclusions, of course for a new medicine one hopes that the
researchers already have pretty good evidence for the usefulness of
their product before going to double-blind trials, but no way can I
agree that medicines or techniques should be approved for mass treatment
-without- such a trial.  It has often happened that apparently effective
medications have problems which only show up when applied to a wider
range of the population.  A drug that might be very effective in a
handful of volunteers can have serious problems when applied to the
elderly or children, or when used in combination with other therapies,
or it might show up a small risk of very serious side effects.

Giving a patient an additional treatment for a very small but
statistically significant benefit or chance of benefit may seem
pointless, but really it is a matter of risk analysis, if there are no
drawbacks (including cost, inconvenience etc.) then why not get the
benefit?

Often these small-benefit analyses do not refer to new medications but
to changes in method. For example a recent trial compared giving
chemotherapy every three weeks to giving a smaller dose weekly.  This is
not additional treatment but an either-or situation.  Not only does this
sort of research optimise the efficacy of the treatment, but it also
points out any trend, showing researchers where further benefits might
be obtained.

The same principle applies to drug research. If a drug gives a small but
significant benefit then it is a potential area for research.  For
example, Clodronate is a bisphosphonate which helps bone regrowth.  It
was not very effective and of pretty limited use itself, but over the
last decade or so it pointed the way to a series of newer drugs like
Zometa which are hundreds of times more effective

I feel sure your concern over risk factors is mostly a matter of
semantics.  I don't believe that in identifying higher education as a
risk factor for breast cancer the ACS was suggesting that it should be
avoided, although I accept that I have often seen such statements
mis-interpreted by journalists.  Surely what they were saying was that
women with higher education appear to have a higher risk of breast
cancer, for whatever reason, and so should perhaps take more care over
checks etc.

Likewise the apparent benefits of (eg) drinking red wine should not be
simply ignored.  The fact that some groups live longer or avoid certain
diseases of course does not imply that -changing- your lifestyle to
match certain identifiable attributes of theirs, such as the wine
drinking, will have any benefit, but it does indicate there is a
potential health benefit to be had somewhere in their lifestyle or heredity.

While you may feel that with cervical cancer risk at 1 in 250 it is
small enough that doubling it can be ignored, but you still look both
ways before crossing the road even though your risk of being road-kill
is even smaller.  You are ignoring the possibility of prolonging
something like half a million women's lives in the US alone.

Finally how do you -know- that milk has nothing to do with cervical
cancer.  (Maybe it's not the milk itself but stuff the cow was fed that
gets into it?)  When I was at school my maths teacher insisted that
"obviously" is not proof, and if the only reason you can offer is that
the result is obvious, then it is probably wrong.