GO
BACK TO: DIAGNOSIS: PSA: GO TO YANANOW
HOME PAGE
ULTRA
SENSITVE PSA
Detection
limits of the ultra-sensitive PSA assays is a long and complicated subject.
One
of the key issues is what is termed "background noise" . This refers to the fact
that PSA is NOT prostate specific despite it being named as Prostate Specific
Antigen. Other glands in the body can and do produce this protein and when measuring
the minuscule quantities in ultra-sensitive assays after radical prostatectomy,
the PSA from other organs can occasionally be confused with PSA generated from
prostate cancer or any recurrence of the disease. Again it is important to bear
in mind that PSA is NOT prostate cancer specific.
A
good deal of the questionable value of ultra-sensitive assays can be traced back
to the mid-1990's around the time Diagnostic Products began marketing their Immulite
3rd generation assay when the competition between assay manufacturers resulted
in misleading sales pitches.
In 1996 Dr. Thomas Stamey of Stanford wrote
an interesting essay for the medical journal "Clinical Chemistry" about the resulting
confusions regarding assay sensitivity titled "Lower limits of detection, biological
detection limits, functional sensitivity, or residual cancer detection limit?
Sensitivity reports on prostate- specific antigen assays mislead clinicians."
(Clin Chem. 1996 Jun;42 (6 Pt 1):853-7) Here are some quotes from the article:
"Unfortunately,
commercialism has led assay manufacturers to tout their differences in LLD (lower
limit of detection) among assays to the point where laboratorians believe that
the LLD has clinical significance and subsequently relay these values to the clinicians
on the laboratory report...'
"Many assays use bovine serum albumin (BSA)
as their matrix, although other animal sera are also used. However, BSA is a highly
purified simple protein that bears no resemblance to the complexities of human
serum. Intra-assay variation with such simple proteins might be elegantly low,
but this would be misleading for the clinician and potentially harmful for the
patient..."
"In a research lab setting, greater pains are taken
to assure consistent quality of reagents and to fine tune calibration of the instruments,
and they can accurately measure PSA down to a level of 0.01 when using a better
assay such as the Immulite. One hears of PSA assays that can measure levels down
to around .002, but this kind of accuracy is not possible in a medium as complicated
as human blood. This level of accuracy is possible only when PSA is suspended
in a simpler medium such as bovine serum albumin."
There is also
the question of accuracy in measuring volumes. In an exchange on a Mailing List,
one member said: " I was ' under 0.1' and asked for an ultra-sensitive test
to see just how low it was. The 1st test with the more sensitive scale was 0.009,
and is now 0.003. I find it a miracle that lab workers can even measure down to
a trillionth of a gram per ml."
The response, from a man whose career
was in measurement at Princeton, Argonne National Labs, and Fermilab, said in
part: "Nothing on earth can be measured to this precision. Well, there are
some obscure things (besides time) that only physicists care about. But think
about it. To measure the PSA, one must take a volume sample. The above ml. Now
how accurately can you do that? Think about measuring gasoline or milk. It is
hard to do it to 1%. ……ignore any small changes. You just can't measure them."
The
other issue is the calibration. Here, there are issues like what standard was
used, how long ago the instrument was calibrated, and what the environmental sensitivity
of the instrument is. For example, if the instrument was calibrated at one temperature
and is used at another temperature, then all the measurements can be off by some
fixed amount (such errors are called systematic errors). To prove to yourself
that instruments are not perfect, look at the thermometers being sold in a store.
Some will show, e.g., 71 deg F, some 73, some 72, etc - or check out the precise
time on a digital watch with some of your friends. Most watches show the time
down to one tenth or even one hundredth of a second - and all claim accuracy.
But it is extremely unlikely that five or six watches will match time that precisely.
Assuming
that a laboratory's instrument is calibrated and used properly (a big assumption),
one way to get an idea of the "accuracy" of the reported result is to, say, divide
a large blood sample into 100 equal parts and submit each as if it were a different
sample. The results may come back as 0.002, 0.001, 0.002, 0.003, 0.005, 0.001,
etc. The average of the 100 measurements may be 0.002, but, for any given sample,
the result may be substantialy different - not because the sample was different
but because of measurement noise (or random errors). Generally, it is not practical
for commercial labs to calibrate their instruments and cope with differences of
batches of reagents well enough to squeeze maximum possible accuracy out of the
ultra-sensitive assays, so they will draw an arbitrary line, and report nothing
lower than about 0.03. Different labs may choose slightly different numbers, even
though they may be using the same assay.
A laboratory should report the
results as a range with a midpoint - for example between 0.001 and 0.003 with
a mid-point of 0.002 or as a value plus an uncertainty, e.g., 0.002 plus or minus
0.001 (written as +/- 0.001). What range and uncertainty is reported depends on
the "confidence interval." If the uncertainty is the standard deviation of the
measurements, then that means that 65% of the measurements will be within that
uncertainty (and of course 35% will be outside). If it is six times the standard
deviation, then 99% of measurements will be within the uncertainty (i.e., one
percent of the time, the measurement will still be in error by more than the uncertainty
reported.) although the range will be greater. I am not aware of any laboratory
that reports both the results and the uncertainty in the results. Maybe the labs
are trying to protect us from the complexities of measurement science.
So,
if a PSA reading goes from 0.002 to 0.003, the "increase" may NOT be due to any
real increase. It could be due to just measurement noise.
Analytical labs
switch assays often. This can play havoc with patients (and their doctors) if
they are not totally aware of these changes and the potential implication of these
changes. Ignoring human errors (of which there are many) the actual design of
the assays and their calibration standards are fundamentally important. The reason
for that is related to the antibodies used in the design. Some assays use monoclonal
antibodies and some use polyclonal ones. These might detect the different forms
of immuno-reactive PSA in different ways based on the composition of the calibration
standard.
The important thing is to try to get all tests done at the same
lab, using the same assay Although there has been some discussion of standardization
in the industry and a protocol - the so-called Stanford Protocol - was agreed
for total PSA, the highly sensitive assays are not standardized relative to each
other. This means that simply switching assays could result in a surprising jump
even with no true change. An universal calibrator would surely reduce the inter
assay variability while the intra variability depends more on sample handling
and the physiological variation of PSA in patients.
It has become harder
and harder to stick with the original assay because of Labs and health insurance
changes. If a significant change is noticed, it is worthwhile to do a baseline
comparison using the old and new assays on the same blood sample.
This
piece was written after an extensive discussion on the PPML
Mailing List and thanks are due to the wise men who participated and shared
their knowledge.
