www.pbs.org/newshour/index/prostate/html

includes a panel of luminaries including Dr. Patrick Walsh himself.

Any comments?

Very interesting. It does as good a job as you can expect on such a
complex subject in a relatively limited amount of time. But I have got
to expect this from public radio. It is much superior to other news
sources in this regard.

I was also glad that Walsh was given adequate opportunity to address
most of the crucial issues. As you might expect I found he made a
convincing case. I was less than impressed by the interview with
Gelmann. He made the usual arguments but he also sloughed over some
very important details that for a man who might have prostate cancer are
very crucial. He kept emphasizing that in the majority of cases,
prostate cancer progresses slowly and and watchful waiting, perhaps
followed by hormone therapy can deal with it. Men like that may face
adverse consequences from treatment. But using his own statistics,
that argument clearly doesn’t apply to men under 65 who might be
expected to live another 20 years. It is for such men that PSA testing
provides the greatest benefit. I get impatient with people who make
these areguments because they oversimplify and act as if all cases of
prostate cancer are the same. Gelmann also seems to suggest that
treatment rarely does any good, and I think there is little evidence for
that and lots of evidence, some of which Walsh quotes, that it often
does quite a lot of good.

Such estimates are based on assumptions. One would have to see how he
did the calculation in order to see how seriously to take such a
projection. I think about 180,000 men are diagnosed yearly with
prostate cancer. If you assume that every one of these men has a RP and
that the death rate due to the surgery is 1 percent, you come up with
1800 deaths. I suppose you can get this up to 5,000 by playing with the
increased number of men who would be diagnosed and with throwing in some
deaths from the biopsy. But in fact done by a qualified urologist,
biopsy is extremely safe, and the death rate due to surgery, again when
done by qualified surgeons, is more like 0.2 percent than 1 percent.
Also, most of the men who die as a result of the surgery have some
condition which puts them in danger anyway. You would have to try to
estimate how many of these men would die anyway within a year or so of
the surgery. Finally, one would have to compare this with the number of
lives that would be saved.

Gelmann and those who agree with him seem to want it both ways. They
argue correctly that there is no absolutely unassailable evidence that
aggressive treatment for prostate cancer reduces overall mortality. So
they question estimates of the number of lives saved by treatment. But
then they engage in highly questionable statistical arguments, which are
even less supported by data, about the risks of treatment, and although
they qualify them by ifs there is a clear implication that they are firm
statistics.

For that reason, he is almost certainly an MD, but it is remotely
possible he is a PhD in an allied field. It doesn’t matter. He is
qualified to make statements of this kind, and I doubt if he has any
motive except pushing his particular point of view. But I also am
suspicious of such statements without seeing the basis for the
calculation. It is certainly true that some number of men die as a
result of the train of events which start with a PSA test. Although I
doubt it, it is possible that under some scenarios if every man over 50
in the US were tested, the death toll could reach 5,000 per year. For
example, some of them might even be killed in auto accidents on their
way to be treated. But what he is guilty of is making such a statement
out of context. A relevant question would be how many of those 5000
would die anyway within a brief period of time. More important, how
many men won’t die as a result of testing? Gelmann and others like him
don’t believe that screening and the treatment which follows it
generally makes much difference in how long men live afterwards. I
think they are wrong, but this is an old controversy and it won’t be
settled any time soon.

There is one other point that often is ignored. The risks associated
with prostate cancer are highly dependent on details. It is quite
possible that overall there is no reduction in mortality following
aggressive treatment (although I also doubt that). But any given man
has to worry about the risks to him, and that will depend on his
individual circumstances, and what he considers important. For example,
I think a very strong case can be made for testing men between 50 and 70
and following up with aggressive treatment in most cases when cancer is
diagnosed. For men over 70, it would depend on the man’s health and
other factors. In my case, I was diagnosed with a Gleason 7 tumor at
age 67. It would be hard for Gelmann to make the argument that I didn’t
benefit from PSA testing and aggressive treatment to cure the cancer.
He might argue that other men might have lost more in total than what I
gained, so from a population point of view there was a net loss. But
what is relevant to me when deciding on whether or not I should be
tested is the potential gain to me vs the potential loss to me, not to
men in general. What I wanted to try to avoid was advanced prostate
cancer. I considered that much more of a threat than the threats
associated with treatment. Had I had a Gleason 5 cancer, it is possible
that the risks of treatment would have outweighed the risks of the
cancer. But how would I know about my choices without PSA testing and
followup where indicated? A doctor might reasonably ask whether he is
endangering more patients than he is helping by encouraging PSA testing.

Testing may lead to treating cases that don’t need treating and some
of those men may die and all of them will suffer some side effects,
perhaps only temporary, of treatment. But from the man’s point of view
it really looks different, as I’ve noted above.

Statistical decision theory involves subtleties which sometimes even
experienced epidemiologists miss. Related but different questions may
have different answers.