Back to the Top
The following message was posted to: PharmPK
Dear PK and TK Scientists,
I am sure this is not a new question to this forum. I have seen a lot of
discussions regarding this issue back to 1998. But as far as I know, it
seems like there has not been a commonly agreed method to handle the
BLOQ data(below the limitation of quantitation). Could anyone please let
me know if FDA has comw up with any guidelines on how to deal with the
BLOQs in PK and TK studies? Any papers, methods of assigning values to
BLOQ you would suggest?
Thanks.
Linghui
Back to the Top
Linghui:
FDA guideline does mentioned LOQ issue:
The lowest standard on the calibration curve should be accepted as the
limit of quantification if the following conditions are met:
A The analyte response at the LLOQ should be at least 5 times the
response compared to blank response.
B Analyte peak (response) should be identifiable, discrete, and
reproducible with a precision of 20% and accuracy of 80-120%.
Your LOQ should be met these conditions. Any values lower than your LOQ
should be treated as "zero" for your PK report.
Bang Qian Xu, Ph.D.
Apotex Inc.
Back to the Top
The following message was posted to: PharmPK
Dear Linghui:
I do understand that. However, can you trust your BLOQ values? That's
why one uses "zero" in practical PK reports. None of the regulatory
agents will accept your BLOQ values. If your values are only for
research purposes rather than a PK report used for FDA submission then
you are okay.
Regards,
Bang Qian Xu, Ph.D.
Apotex Inc.
[Seems like missing data would be more appropriate. That is, a 'dash' to
indicate no value - db]
Back to the Top
The following message was posted to: PharmPK
I do agree to statements A and B of Bang Qian Xu but I disagree that
values lower than the limit of quantitation should be treated as zero
for the report. The reason for this is that BLOQ means that this value
is somewhere between the limit of quantitation and zero. There are often
cases where it makes a lot of sense to define the first value being
below the limit of quantitation as half of BLOQ for PK or TK
calculations like AUCs. However, PK-programs might not accept BLOQ but
only discrete values so that "zero" or "half of the limit of
quantitation" is often taken for calculations. In any case those values
have to be reported as BLOQ and it has to be indicated how they were
treated for PK-calculations. Any other opinions from the pharma
community?
Regards,
Bernhard
Bernhard J. Ladstetter, Ph.D.
Head of the Institute of Pharmacokinetics and Metabolism Merck KGaA
Am Feld 32
D-85567 Grafing
Germany
Back to the Top
The following message was posted to: PharmPK
At 20:33 -0600 02-02-21, Bangqian Xu wrote:
>
>I do understand that. However, can you trust your BLOQ values? That's
why one uses "zero" in practical PK reports.
My response would be that I "trust" the BLOQ measured values as
being closer to the true value (in general) than a forced zero
>None of the regulatory agents will accept your BLOQ values.
Then its time to "educate" the ostriches, er, regulatory agents
>[Seems like missing data would be more appropriate. That is, a 'dash'
to indicate no value - db]
No, No, No, a thousand times No, please.
[Why not - continued below - db]
Derrick J. (Rick) Bates | mailto:DJ.Rick.Bates.aaa.PNL.gov
Statistical and Quantitative Sciences | http://www.PNL.gov/Statistics
2400 Stevens Drive, K5-12
Pac-NW National Lab (PNNL), Battelle
P.O. Box 999, Richland, WA 99352 US | ALL POSSIBLE DISCLAIMERS
[Why not? If your data is below limits of quantitation it shouldn't be
used. Looking at an individual subject (column of data) stop reporting
numbers (time or Cp) once the LOQ is reached. You shouldn't average
across the table...I thought we had stopped doing that years ago. If you
include 'a' number in a modeling data set what do you tell the software
about the weight. Here's a data point but ignore it! If you are doing
'non-compartmental' analysis your last few data points are going to be
the same (i.e. LOQ or LOQ/2 or 0) so the terminal slope will be zero.
Could make for some interesting estimates of AUC(last to infinity).
If you can't read a number greater than LOQ it should be thrown out or
you should develop a better assay method!!
If you are reporting a table of results and some of the subjects provide
concentrations below LOQ a '-' or NA or something is better than 0 or a
blank (later filled in as 0)
- db]
Back to the Top
The following message was posted to: PharmPK
1. If you do follow the regulatory guidelines and work compliance
with GLP, you will not use your inaccurate BLOQ values in your PK
report. 2. Using "zero" or NA will have different AUCs in your PK
report. 3. If you don't want to waste your (or your company's)time and
money, you'd better
to follow guidelines and work compliance with GLP.
Ban Qian Xu, Ph.D.
Apotex Inc.
Back to the Top
The following message was posted to: PharmPK
No you cannot interpolate between LLOQ and zero. An you cannot
extrapolate therefore do not use.
Back to the Top
The following message was posted to: PharmPK
please excuse the lack of uppercase; i broke my hand.
regulatory bodies certainly won't accept a number other than zero if it
is bloq. if i'm not mistaken, what fda is most concerned with is
model-independent ("non-compartmental") pk parameters. my question
therefore is, need one validate the nonlinear, i.e. curve fitting, parts
of winnonlin?
Back to the Top
The following message was posted to: PharmPK
WOneil.at.qltinc.com wrote:
>
>regulatory bodies certainly won't accept a number other than zero if it
is bloq.
Please be specific. Exactly which regulatory authorities insist on this
kind of stupidity?
>if i'm not mistaken, what fda is most concerned with is
model-independent ("non-compartmental") pk parameters.
People I know at FDA and others interested in PK analysis do NOT
consider non-compartmental approaches to be 'model independent'. Their
use for bioequivalence analysis assumes AUC can be used to predict
clearance/F. This assumption is dependent on a specific model with first
order elimination from the central compartment. Extrapolation to
infinity also requires a specific model.
--
Nick Holford, Divn Pharmacology & Clinical Pharmacology University of
Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand
email:n.holford.at.auckland.ac.nz
http://www.health.auckland.ac.nz/pharmacology/staff/nholford/
Back to the Top
The following message was posted to: PharmPK
Here are some practical considerations in dealing with BLOQ
measurements.
First, to use zero seems to me to be illogical. If you could reliably
demonstrate zero molecules of some substance in your sample, then your
method would have no lower limit of quantitation, and no BLOQ values.
Most common PK analyses cannot do this because they measure not
molecules themselves but some
surrogate quantity like fluorescence.
Second, as Prof. Holford mentions, even noncompartmental methods involve
fitting a model. Generally, that model is a single exponential fit to
the terminal phase. For fitting single exponentials, it is common for
software to take the log of the concentration data and fit a straight
line to that. Zero concentrations are problematic as their log is minus
infinity. Indeed any kind
of logarithmically based analysis is apt to have problems when
attempting to display or fit concentrations of zero. I consider this yet
another reason not to use zero as a BLOQ value.
Someday software and users of software will realize, I hope, that
experimental
data is at best semiquantitative. Sprinkled throughout the quantitative
measurements are missing data (oops, I dropped that vial), null or not
applicable data, and, as here, out of range data. We need to handle
these cases
better. Representing BLOQ values as zero or removing them altogether
from the data set seem inappropriate to me.
Paul Laub
P.S. I wonder what Olympic athletes and those who test their blood and
urine would have to say about this issue.
Back to the Top
Which Regulatory bodies? Well, the FDA (USA), the TPD (Canada), the EMEA
(Europe).
Noncompartmental methods in BE? Well Cmax and AUCt for the FDA, Canada
bends a bit by adding the AUCinf as does Europe. These Regulatory bodies
really are serious about getting valid data.
Edmond Edwards,
Manager, PK & Statistics
Biovail Contract Research
Toronto, Ontario
Back to the Top
The following message was posted to: PharmPK
Please give a citation. I did a search of FDA regulatory guidances and
couldn't find it. Also I used to work in international drug regulations
and don't know of any countries off the top of my head. Finally, please
differentiate between NDAs and ANDAs as the requirements could be
different.
Ron Kavanagh
FDA
(The views expressed are my own)
Back to the Top
The following message was posted to: PharmPK
Dear,
Science and practice not always go hand in hand. Scientifically it is
probably the best to apply population PK modeling and estimate a
concentration for the BLOQ value. Practically this is often not feasible
since there is simply not the time and resources to do this for every
study. Within the industry most studies are Phase I and a routine
analysis is done by data-analysts, not the popPK specialist. Therefore,
simple and easy rules are applied to assure consistency and
transparancy. These rules must be written down in SOP's. According to
our SOP we replace the BLOQ value by 0 for the calculation of
descriptive statistics. If more than half of the number of individual
values is BLOQ, the statistics are set to NQ. I know from two other
pharmaceutical companies that also apply these rules. Drugs have been
filed with these rules thus apparantly these rules are acceptable. These
rules will lead to some bias, but in most cases this bias is negligible.
The conclusions of the study will generally not change. One should
realize that if a concentration is BLOQ most of the drug has already
been eliminated. The part of the curve corresponding with the BLOQ
measurement generally adds very little to the overal AUC. Often 90% of
the AUC has already been characterized.
Best regards,
Kees Bol, PhD
Director Clinical Pharmacokinetics
Kinesis Consultants
Smederijstraat 2
4814 DB Breda
The Netherlands
Back to the Top
The following message was posted to: PharmPK
This discussion is going a long time. Perhaps an amateur could sum it
up: 1. Use the below LOQ results - quite unjustifiable, these
'detections are not quantifiable.
2. Set these results to 0. This may please people who take the view that
the unquantifiable is not there, but we know it is not right. Assume a
result above but near the LOQ. The next result is shortly after but
below LOQ. It is obviously wrong to set this to 0. The error in terms of
AUC may not be great but there will be an underestimation. Why do it?
Apart from possible bureucratic demand.
3. Extrapolate - this is probably the most reasonable where the rate of
decrement is close to log-linear and will yield the result likely to be
closest to real.Where the beta tail is not defined the AUC is pretty
well unknown anyway and perhaps the best thing to do is another study.
4. Go POP - this will probably yield a more reliable beta anyway. Now I
await being shot down in flames by the pros. Best regards, at least I am
learning a great deal from this group. Tom Torda
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)