"Paul Johnson" mentioned a study about

Oh, yeah ... like I'd WANT another one. After all, Walsh says it gives more
grief for less benefit than any other organ in the body.

I.P.

As the article states, statisticians don't like studies of this type,
called observational studies.  That is because there could be
uncontrolled factors which could explain the results.  For example, it
could be that the men who were treated by WW were more likely to have
other serious medical problems.   They much prefer randomized studies
where patients are assigned at the beginning to two or more groups, each
treated by a different protocol.  Measures are taken to make sure that
men in each group don't differ substantially in other ways from men in
the other groups.

Indirectly, this study shows why the statisticians favored randomized
method may deliver misleading results.   In large studies of this kind,
including the current ones,  it is very difficult to control for
treatment effectiveness, so they don't do it.   The usual rationale for
this goes something like this.  The best surgeons may do better, but in
the real world, patients are going to be treated by a surgeons with a
range of skills, so the study approximates what actually is relevant for
the population as a whole.   But this ignores the fact that treatment
effectivenss will generally improve over time as more doctors train at
the best centers and the needed skills diffuse into the general medical
population.  Studying what may be true for one limited period of time
and then assuming it will remain true for the indefinite future is
faulty reasoning.  Another problem with randomized studies for diseases
like prostate cancer, which may rquire long study times, is that while
the study is in progress, factors which weren't known to be relevant at
the beginning of the study are discovered to be important while it is
going on.   But modifying a study while in process violates the
statistical protocols.