Re: To 3rd man

On Feb 23, 12:29�pm, the 3rd Man <sir_de...@xxxxxxxxx> wrote:
On Feb 22, 8:16�pm, cowabungaba...@xxxxxxxxx wrote:



Soooo...are you saying then, that the ELISA is a perfectly valid test?

First, it is entirely a valid methodology as I understand it. It has
been validated. In THEORY it is fine. Perhaps done under the optimal
circumstances with the perfect laboratory using the perfect antigens
for capture it would be great (although not without its limitations,
better or worse as antibodies rise and fall, better in patients with
certain symptoms, affected by prior antibiotic use, depends on free
antibodies rather than all available antibodies being tied up in
immune complexes, no true seronegativity caused by dissociative b and
t cell response as per Volkman et al). So not perfect but no test it.
But fine IN THEORY:

Okay... but that is NOT what I understand you to have been saying. You
have been saying that it has a 50% sensitivity. Logically, it cannot
have a 50% sensitivity in one lab on one day and a 95% in another lab
on another day.

For one thing YES I have been saying it has 50% sensitivity as done in
commercial labs using commercial test kits. I was not talking about
the theory of the test or its puffed up validation figures (not
fraudulent, just optimized),.

And YES illogical though it may seem it has different sensitivities in
different labs.

Which apparently is the result of lab proficiency and lack thereof and
lab administration of the test. I'm not sure what other factors affect
sensitivity but apparently it does.

That is why there is both intra and inter lab variability. Also those
variabilities are due to non standardized tests (different test kits
used in different labs).

There are labs that apparently are more proficient in doing ELISAs, as
per that CDC paper which showed a range of sensitivities in different
labs.

If I understand NOW what you are saying...you are saying that the Lyme
ELISA may, in fact, have its claimed sensitivity ranges, if properly
applied.

Yes sure no doubt--although another factor that makes the validation
numbers different from the numbers in practice for sensitivities is
that the test is more or less sensitive for different groups of
patients exhibiting different symptoms--so sensitivity varies
depending on presentation for cardititis it is high, for neurological
symptons lower but still high, for EM rash lowest, for arthritis
relatively high though it varies amongst late early and overall.

Also timing of the test affects sensitivities, early in disease, late
in disease. Prior antibiotic use apparently affects sensitivity too.
Again all but the prior antibiotic use is detailed in that CDC paper
http://www.cdc.gov/ncidod/eid/vol2no2/craven.htm

When doing validation studies they optimize the patient sampling and
obviously lab proficiency. If you read the fine print the highest
proficiency claimed is not the same across all groups of patients.

See...my understanding is that the number assigned to its sensitivity/
specificity comes form the LAB...not from "real life". In "real life"
an individual lab, given that sensitivity, may have differing marks
for how well it correctly identifies samples. Accuracy, precision.

Which means that in real life the sensitivity is lower than in the
validation studies.

Also precision as defined in the CDC paper had a specific meaning:
"precision (frequency of obtaining the same result on duplicate
analysis of a specimen)"

and specificity meant: , "specificity (true negatives correctly
identified)"

and sensitivity meant: sensitivity (true positives correctly
identified), specificity (true negatives correctly identified),

Concordance meant: m"easure of concordance (agreement among
investigators) of results among the tests using the kappa
statistic." (whatever that is and I am NOT going there due to my math
stupidity).

So...what I have been asking you is how you are now coming up with a
number of 50% sensitivity?

As the test is done in real commercial labs using commercial test
kits. In the CDC paper of five labs the sensitivities in those labs
were variously 73, 79, 76, 49 and 40%. So two were LESS than 50%. In a
variety of studies sensitivity was 50% or less, in other greater. In
the lyme vaccine trials sensitivity was approximatelu 60%.

As I said, the 50% was to make the coin flip analogy easy (although
then I totally had a mind fart about the whole thing so think if we
had used the real numbers). The true sensitivity in the field on
ELISAs in Lyme across all patient groups considering all factors that
affect sensivitity is probably a little higher than 50% but not much.

As I said several times a variety of things affect sensitivity: lab
proficiency overall, specific specimen handling and administration of
the test, early vs late disease, symptomatic presentation, true
seronegativity (due to the dissociative b and t cell response as per
Volkman et al) and other factors known and unknown (I don't think
anyone has studied sensitivities being potentially affected by co
infections but for example that could be a currently unknown factor).

If I understand correctly...that makes no sense,

Here's how you just phrased it in response to SR:

I don't know why this is hard to understand?

Bart: "Problem remains: what about false negatives? How does a test
with 50%
specificity as used in clinical practice make sense as an initial
test"?

WHERE are you getting this number from? WHAT is the calculation? And
what does "as used in clinical practice" mean, in regard to
sensitivity?

As opposed to the validation numbers. A patient goes to a doctor, has
blood drawn, it is sent to a commercial lab which uses commercial test
kits. That's what I mean by "as used in clinical practice." As opposed
to the theoretical numbers of the test achieved in optimized
conditions in validation studies (I guess that isn't "theory" per se
but it isn't "as used in clinical practice" either--so let's say as
opposed to the real numbers achieved in optimized conditions using
optimized patient subsets in validtion studies.

YES...believe it or NOT I have read the cited papers...and many times
before you mentioned them. Thank you. If you go back to the bit about
the ELISA calculator...you will see that I caveated "lab error"...and
it was in specific reference to this issue...and this, or similar
studies of inter-lab discrepancy. I meant to say that assuming a
perfect world...

And I don't care about the perfect world (well I'd like to see a
perfect world and apparently all we all have to do is vote for barrack
and have hope and we'll have one???).

I care about the real world of patients and doctors where these tests
are used in clinical practice NOT validation studies.

Citing me to the thing doesn't tell me WHY you have determined a 50%
sensitivity.

Because that is the sensitivity achieved in many labs, in the lyme
vaccine trials and cited in many studies.

Seems to me that what you are really saying is that on any given day,
depending on the lab...you can get a bad result. THAT is substantially
different from saying the Lyme ELISA has a 50% sensitivity.

I'm saying that most patients who go to most doctors and have an ELISA
done, the sensitivity of the test is approximately 50%.

and that means that negative test results are no more meaningful than
a coin flip.

Positives have much more significance and are affected only by
specificity which has problems but is generally much better than
sensitivity in Lyme testing using the ELISA.

Are you attempting to average those results? You can't do that, I
don't think.

Why not? I'm saying that if 100 patients go to 100 doctors that the
average sensitivity of the ELISA is 50% approximately. Some labs are
going to be better others worse. Some patients with certain symptoms
will have a test that produces better or worse sensitivity varying by
their symptoms. Also affected by timing (early disease, late disease,
prior abx use, true seronegativity etc).

So I can do that and it is fair to do it. A fair general statement is
that the ELISA is not a reliable test if you have a negative result
due to poor sensitivity of the test. It is also fair to say that the
ELISA does not meet the standards for use as a screening test for the
two step method due to poor sensitivy as used in clinical practice.
Apparently both bakken et al and the College of American Pathologists
as I cite above say exactly that.

You can't lump all those results together and assign an overall number
to the ELISA in "real life".

Why not?

Such an average would NOT be an accurate indicator or predictor of a
correct result in any given situation, and probably would be HIGHLY
misleading.

It is only misleading in the instance where a particular lab has
consistently better results. If the doctor ordering the test and
patient receiving the results doesn't know and generally there is no
reason they would, then it is a fair statement to make about caution
in interpreting a negative result.

What would be HIGHLY misleading is to cite validation study figures
about the sensitivity of the ELISA leading doctors and or patients to
believe that a negative result is reliable.

But that isn't what you have been saying is it? �

If I have any recollection and forgiving the diversion where I was
focusing on only false negatives with the coin flip analogy for 50 or
more posts, yes that is what I have been saying.

I thought you were saying that it is only really 50% sensitive?

Not in theory. In reality in commercial labs using commercially
licensed and available test kits as they do. (and by the way whose
advertised sensitivities and specificities were achieved in optimal
circumstances amongst optimized patient populations so are not going
to be good in the real world of labs and patients anyway).

HUH? what does this mean? "Not in theory"?

Sorry. That is the best I can do. I'm not making this up.

Well...why don't you take a shot at telling me where the 50% number
comes from, then?

.

Quantcast