Gert van der Hoek
In Memoriam - Editor, Senior Moderator
From ProMED
Q FEVER - NETHERLANDS (14): PCR TEST RELIABILITY
***********************************************
Date: Wed 17 Feb 2010
Source: Annex to an update letter from the Ministers of Agriculture
and of Health to the Dutch Parliament on the Q-fever situation, dated
16 Feb 2010 [trans. by communicator, edited]
http://www.minlnv.nl/pls/portal/url/page/minlnv/actueel/voorlooppagina?p_file_id=51083
[The above Ministers' letter included, as an annex, the following
report, sent on 7 Feb 2010 from the Central Veterinary Institute,
University of Wageningen, to the Dutch CVO Dr Christianne Bruschke,
addressing the PCR test in use for the detection of infected dairy
goat and sheep farms in the Netherlands. - Mod.AS]
Doubts about the validity of the PCR tests for Q-fever have been
expressed in the media and by various persons acting as experts.
As mentioned earlier in several notices both by the broad experts'
group as well as by the CVI, a negative PCR result does not guarantee
that the animal in question or, in case of bulk milk testing, the farm
in question are not infected/contaminated. This aspect has been
elaborately described: there could be times when the animals shed
more, less or no bacteria. Repeatedly testing samples from farms with
continuously negative results will increase the reliability of the
decision to define the status of a farm as not contaminated. For this
reason, bulk milk testing is now performed with an interval of 2 weeks.
Recently, there have also been doubts about the reliability of
positive test results, also because of the different outcome of the
results from the "same" sample tested at various laboratories.
Doubting a positive result, however, is not justified. I will try to
clarify the apparent discrepancies of the results at the various
laboratories or within one and the same laboratory.
There are various PCR tests available to demonstrate the DNA of the
bacterium _Coxiella burnettii_ in a sample. The sensitivity of such a
test depends, among other things, upon the chosen "target DNA." For
_C. burnettii_, some unique DNA particles are described. Some of these
particles occur only once in a bacterium. Others occur more than once
in one bacterium. The number of times that such a DNA particle shows
in _C. burnettii_ can vary per strain. Our CVI test used to detect
positive farms aims to detect the so called IS1111 gene, several
copies of which are known to be present in _C. burnettii_; their
number varies in different strains.
As described in the validation report, this PCR test is 100-fold more
sensitive than a PCR test based on a DNA particle that appears only
once in one bacterium. This means that differences between results
from various laboratories could occur: One laboratory can test more
sensitively than the other depending on the choice of the target DNA
particle.
Even if the various laboratories use the same target DNA sample,
different results can be generated by small differences of test
procedures. Even within the same laboratory, various random tests from
one sample could give different results. This can happen when the
amount of bacteria in the sample is close to the detection level of
the test. A positive result incontrovertibly demonstrates the presence
of _C. burnettii_ DNA and, therefore, the presence of the bacterium. A
positive result is, therefore, absolutely reliable.
A negative result gives 2 possibilities: either there is no DNA in the
sample and, therefore, _C. burnettii_ is absent; or, there is a chance
that the DNA is insufficient in the sample and, therefore, does not
show.
To illustrate the above: suppose I have received a sample of one liter
of milk. As you know, but this might not be clear to laymen, only part
of the submitted sample is used for the final test (in other words, a
single test of a sample is only a single random test from the sample).
Suppose I use one ml in my test with a detection limit of one
bacterium per ml. If, for instance, there were one million bacteria in
that one liter of milk, thus 1000 bacteria per ml, any laboratory will
test this sample to be positive, and various random tests within the
same laboratory will all give a positive test result. On the other
hand, if the amount of bacteria in that liter of milk is lower, e.g.
1000, meaning one bacterium per ml, this will be on the detection
limit of my test, and it could, therefore, lead to a one-time negative
random test. That applies even more in case of an even lower amount of
bacteria (e.g. with 10 random tests from one sample: 7 random tests
with negative results and 3 with a positive result). Despite the 7
negative scores, the final outcome of the said test is a positive milk
sample.
Apparently, there is some confusion as well about the specificity of
the test using the data from the CVI-validity report.
To clarify this: the test has an analytical specificity of 100
percent: if the test is positive, DNA of the _Coxiella burnettii_
bacterium will show in all cases. There are no false-positive results.
Apparently, some expert has indicated that based on data from our
report, the predictive value of a positive test is 84 percent; on
account of a comparison of our results in the validation report made
between the PCR-test and an ELISA for detection of antibodies. We are
asked for such data to accredit the tests. In technical terms, it has
to do with a comparison of 2 different tests when no "golden standard"
is available and relative diagnostic sensitivities are bilaterally
calculated. Such data, however, are irrelevant for the predictive
value of the PCR test in order to show bacterial DNA and, therefore,
cannot be used to estimate the reliability of a positive PCR result,
as apparently has been done.
F.G. van Zijderveld
Head of Bacteriology and TSEs Division
Acting Director CVI of Wageningen UR
[The above explanatory note is timely and rather convincing. One point
deserves to be stressed and is valid for PCR tests in general: in view
of the high sensitivity of PCR tests, with the potential to detect the
target DNA and thus produce a positive result even when the number of
bacteria is indeed minimal, the samples should be collected,
transported and handled throughout with maximal biosecurity and
caution, avoiding any chance of their incidental cross-contamination.
- Mod.AS]
http://www.promedmail.org/pls/otn/f..._BACK_PAGE,F2400_P1001_PUB_MAIL_ID:1000,81417
Q FEVER - NETHERLANDS (14): PCR TEST RELIABILITY
***********************************************
Date: Wed 17 Feb 2010
Source: Annex to an update letter from the Ministers of Agriculture
and of Health to the Dutch Parliament on the Q-fever situation, dated
16 Feb 2010 [trans. by communicator, edited]
http://www.minlnv.nl/pls/portal/url/page/minlnv/actueel/voorlooppagina?p_file_id=51083
[The above Ministers' letter included, as an annex, the following
report, sent on 7 Feb 2010 from the Central Veterinary Institute,
University of Wageningen, to the Dutch CVO Dr Christianne Bruschke,
addressing the PCR test in use for the detection of infected dairy
goat and sheep farms in the Netherlands. - Mod.AS]
Doubts about the validity of the PCR tests for Q-fever have been
expressed in the media and by various persons acting as experts.
As mentioned earlier in several notices both by the broad experts'
group as well as by the CVI, a negative PCR result does not guarantee
that the animal in question or, in case of bulk milk testing, the farm
in question are not infected/contaminated. This aspect has been
elaborately described: there could be times when the animals shed
more, less or no bacteria. Repeatedly testing samples from farms with
continuously negative results will increase the reliability of the
decision to define the status of a farm as not contaminated. For this
reason, bulk milk testing is now performed with an interval of 2 weeks.
Recently, there have also been doubts about the reliability of
positive test results, also because of the different outcome of the
results from the "same" sample tested at various laboratories.
Doubting a positive result, however, is not justified. I will try to
clarify the apparent discrepancies of the results at the various
laboratories or within one and the same laboratory.
There are various PCR tests available to demonstrate the DNA of the
bacterium _Coxiella burnettii_ in a sample. The sensitivity of such a
test depends, among other things, upon the chosen "target DNA." For
_C. burnettii_, some unique DNA particles are described. Some of these
particles occur only once in a bacterium. Others occur more than once
in one bacterium. The number of times that such a DNA particle shows
in _C. burnettii_ can vary per strain. Our CVI test used to detect
positive farms aims to detect the so called IS1111 gene, several
copies of which are known to be present in _C. burnettii_; their
number varies in different strains.
As described in the validation report, this PCR test is 100-fold more
sensitive than a PCR test based on a DNA particle that appears only
once in one bacterium. This means that differences between results
from various laboratories could occur: One laboratory can test more
sensitively than the other depending on the choice of the target DNA
particle.
Even if the various laboratories use the same target DNA sample,
different results can be generated by small differences of test
procedures. Even within the same laboratory, various random tests from
one sample could give different results. This can happen when the
amount of bacteria in the sample is close to the detection level of
the test. A positive result incontrovertibly demonstrates the presence
of _C. burnettii_ DNA and, therefore, the presence of the bacterium. A
positive result is, therefore, absolutely reliable.
A negative result gives 2 possibilities: either there is no DNA in the
sample and, therefore, _C. burnettii_ is absent; or, there is a chance
that the DNA is insufficient in the sample and, therefore, does not
show.
To illustrate the above: suppose I have received a sample of one liter
of milk. As you know, but this might not be clear to laymen, only part
of the submitted sample is used for the final test (in other words, a
single test of a sample is only a single random test from the sample).
Suppose I use one ml in my test with a detection limit of one
bacterium per ml. If, for instance, there were one million bacteria in
that one liter of milk, thus 1000 bacteria per ml, any laboratory will
test this sample to be positive, and various random tests within the
same laboratory will all give a positive test result. On the other
hand, if the amount of bacteria in that liter of milk is lower, e.g.
1000, meaning one bacterium per ml, this will be on the detection
limit of my test, and it could, therefore, lead to a one-time negative
random test. That applies even more in case of an even lower amount of
bacteria (e.g. with 10 random tests from one sample: 7 random tests
with negative results and 3 with a positive result). Despite the 7
negative scores, the final outcome of the said test is a positive milk
sample.
Apparently, there is some confusion as well about the specificity of
the test using the data from the CVI-validity report.
To clarify this: the test has an analytical specificity of 100
percent: if the test is positive, DNA of the _Coxiella burnettii_
bacterium will show in all cases. There are no false-positive results.
Apparently, some expert has indicated that based on data from our
report, the predictive value of a positive test is 84 percent; on
account of a comparison of our results in the validation report made
between the PCR-test and an ELISA for detection of antibodies. We are
asked for such data to accredit the tests. In technical terms, it has
to do with a comparison of 2 different tests when no "golden standard"
is available and relative diagnostic sensitivities are bilaterally
calculated. Such data, however, are irrelevant for the predictive
value of the PCR test in order to show bacterial DNA and, therefore,
cannot be used to estimate the reliability of a positive PCR result,
as apparently has been done.
F.G. van Zijderveld
Head of Bacteriology and TSEs Division
Acting Director CVI of Wageningen UR
[The above explanatory note is timely and rather convincing. One point
deserves to be stressed and is valid for PCR tests in general: in view
of the high sensitivity of PCR tests, with the potential to detect the
target DNA and thus produce a positive result even when the number of
bacteria is indeed minimal, the samples should be collected,
transported and handled throughout with maximal biosecurity and
caution, avoiding any chance of their incidental cross-contamination.
- Mod.AS]
http://www.promedmail.org/pls/otn/f..._BACK_PAGE,F2400_P1001_PUB_MAIL_ID:1000,81417