Ron,

I think your general point is well taken - that it's impossible to
get meaningful statistical comparisons if the standard for
recurrence is not the same.

However in this particular case there is a problem with the
PSA cut-off values.  Imagine a treatment in which everyone
winds up with a PSA of 5.0, but no one ever experiences
symptoms or dies of the disease.  Wouldn't that be a
successful treatment?  Might it not be better than some
other treatment in which 90% of men have undetectable
PSA but 10% develop symptoms and die?

Radiation may lead to a life that is free of actual cancer
progression (no metastasis, no symptoms, no death from
PCa) without lowering the PSA to below 0.2.  If we used
PSA > 0.2 as the standard for failure, some radiation patients
would be declared as failed without their actually ever
developing metastases, symptoms, or death from the
disease.

All of us had PSA values greater than .2 before we had
cancer.  The PSA value itself is merely a useful leading
indicator.  It is not, itself, except in the case of post-RP
patients, proof of cancer.

I don't know if it's true or not, but the radiation oncologists
generally claim it is not necessary to get a PSA below .2 to
have a "cure" in the above senses.  So declaring everyone
with a PSA > .2 as failed treatment would understate the
success of radiation.

PSA testing happens to be the cheapest and easiest way
to follow up on treatment, so it is the one most heavily used.
It allows the researchers to get some feedback about the
treatment without waiting 10 years or more to see how many
people die of the disease.  It also has some ability to
abstract away from the age of the patients.  Since more
relatively young men get surgery and more relatively old
men get radiation, equal outcomes of treatment would probably
mean that more surgery patients die of PCa simply because
more radiation patients die of other diseases of old age
even if their PCa is not successfully controlled.  We'd need
age adjustments, and possibly other adjustments, that would
increase the required sample sizes and make comparison
more expensive and difficult.

So the measurement problem is a hard one.

One measure that I've seen that meets the requirement for
being cheap (though not as cheap) and easy, while still
compensating for the problems of age and PSA outcome
treatment differentials, is to look at rising PSA as the test.
I think that one of the radiation oncology groups uses 3
successive rises of PSA over some period (it might be
at least 3 months between tests, I don't remember) as
an indication of treatment failure rather than some
specific cutoff value.

That's not as cheap as the ASTRO test because it requires
more tests, more follow up, and more record keeping.
But it might still be very practical.

That sounds to me like a test that should work for both
RP and RT patients.  However I don't know what studies
use that test, and what other problems it might have.

   Alan