There has been a lot of discussion of the recent study reported in the
New England Journal of Medicine.   On the one hand too much has been
made of it because the results reported are not surprising in light of
what was already known.  If you look at Walsh's 1995 book on prostate
cancer, he already notes that a significant number of prostate cancers,
 I think it was 25 percent or so, arise in men with "normal" PSA.  The
NEJM report says that 15 percent of men in a certain placebo group of
another study who had PSAs below 4.0 showed prostate cancer on biopsy
after 7 years.  We shouldn't directly compare the 15 percent with the 25
percent since they are measuring somewhat different things, but still
the orders of magnitude are consistent.   Moreover, those like Welch at
Dartmouth are jumping to conclusions when they imply that the recent
study shows PSA testing is useless.   Note also that all researchers
have vested interests in their various hypotheses.  Welch has been a
consistent critic of early detection and has attempted to show with
various populations studies that he is right.   Researchers like William
Catlona, who pioneedred the use of PSA testing, probably have a bias in
favor of showing that PSA testing is useful.  But this is all worked out
in the ordinary course of publications and analysis of all the research.
  Moreover, in scientific articles, researchers have to adhere to
certain methodological guidelinse which tend to make them more
objective.  Welch  sounding off to a reporter about someone else's
research, doesn't merit much attention.  If he has something to say
worth paying attention to, he should submit it to a scientific journal
where referees can examine his arguments carefully.  It should be noted
that Catalona restricts himself mostly to scientific forums and doesn't
carry the debate to the public arena.  Maybe he should answer those
critics who do so.

One the other hand, the paper does raise some interesting points, which
should be explored further.   The purpose of any diagnostic test for
early cancer should be to select a population at greater risk which can
be tested further.   I don't think any serious student of the subject
can maintain that PSA is useless for that.   Men with PSA greater than
4.0 ng/ml are in factor more likely to have prostate cancer, and the
higher the PSA, the more likely that is.   The question is exactly how
much more likely than men in the general population.  the paper gives
some information about that, but I don't think it settles it.  In
addition, it would be important to know what additional information,
such as PSA velocity or free PSA tests might add and thereby increase
the selectivity.

A related question is the following.   The most critics can conclude
from studies of this kind is that you would detect as many prostate
cancers simply by randomly doing biopsies in the same numbers on all men
over a certain age.  Note that the absolute number of men in the US who
die each year of prostate cancer has been declining during the era of
PSA testing.  I find it extremely unlikely that this would occur only if
men had biopsies after abnormal DRE exams or even after explicit
symptoms of prostate cancer.  The usual argument is improved treatment,
but we know that the basic primary treatments aren't of any use if the
cancer has progressed, and the likelihood of progression is higher if
one waits for abnormal DRE results.   (Note in this context that some of
the critics of early diagnosis also argue against routine DRE.)  So it
is possible that the major effect of PSA testing has been that a lot of
men have had early biopsies which has led to detection of more early
cancers.  I don't find this too plausible, but it would be something for
a bright biostatistician to investigate.