Back to the Top
Dear group members,
I have a query on Calibration curve acceptance criteria.
We have 11 points in the calibration curve starting from STD 1 to STD
11, and one blank and blank IS (blank sample with IS).
In one of the subject sample analyis run, all the standards (STD 1 to
STD 11) are with in the specifications for accuracy, but, both blank
and blank IS are not meeting the acceptance criteria (less 20 % of
the LOQ(STD 1)) against the STD 1, so we have removed the STD 1 from
the regression and calculated the interference against the STD 2,
making STD 2 as the new LOQ, unfortunately still both blank and
blank IS are not meeting the acceptance criteria against the STD 2,
at this stage we have gone further and calculated the interference
against the STD 3,making STD 3 as new LOQ and against the STD 3 both
the blank and blank IS are meeting the acceptance criteria,
Shall we accept the CC curve (truncated curve) even though 2 points
are removed form the regression, by taking in to consideration that
75 % of the CC points are included in the regression?
One more problem is, we have 3 QCs with each batch [QC1(low), QC2
(medium)and QC3(high)], after making the STD 3 as new LOQ (truncated
curve) the low level QC (QC1) is falling below the STD 3,
Shall we accept the subject samples batch even though low level QC is
falling out of the truncated CC range by repeating the unknown
samples, which are below the STD 3.
Please provide your valuable inputs.
Regards
Debbie.
Back to the Top
Dear Debbie,
In my opinion, you need to go for a scientific evaluation of the
situation before finally accepting or rejecting your batch. What is
becoming most important here is that, if you still accept the batch
based on truncated CC, you may invite questions about the possibility
of study sample contamination during the processing and analysis of
that particular batch. One more point to be noted is that the
acceptable QC results don't necessarily ensure the validity of study
sample results particularly in such a scenerio. One more question
which is coming into my mind is till what extent you would have
continued removing your standards from bottom because you can still
remove one more point i.e., STD 3 (some may argue that 75% of 11 is
8.25 and may say that minimum 9 standards are to be within accuracy
acceptance criteria to accept the batch but some other may still
accept a 8 point calibration considering 0.25 as insignificant)?
The rationale/practice of going for a lower or higher end truncated
CC or both end truncated CC is actually in a situation where the
lowest or highest standard fails to meet the accuracy acceptance
criteria due to sample processing error such as loss of sample,
improper addition of IS but not limited to and/or poor
chromatography. Your situation is reverse and it is apparent that
this is a prominent contamination issue. Therefore it would be really
difficult for you to justify (at least for this particular batch of
analysis) whether the data generated for the study samples are
reliable. There may be an impact of this contamination on the results
of study samples as well.
I would like to add few more points here and I invite the group to
have their say. Whenever, we accept truncated CC, we need to ensure
that minimum three QC levels are bracketed by the CC standards. In
order to have this, it is recommended to place the LQC and HQC
concentrations in relation to the standard curve range as LQC above
two lowest standards (STD 1, STD 2 followed by LQC concentration)
while the HQC below two highest standards (HQC concentration followed
by STD 10 and STD 11, taking your range). I would recommend you to
have two blank matrix samples and two blank IS samples in the batch
instead of one each. You can also incorporate duplicate processing
and analyses of the lowest and highest standard. But before doing all
this you need to have the acceptance criteria written in the SOP in
case you have blank, blank IS, lowest and highest standard all in
duplicate. Another point that is recommended, it is better not to
accept a batch if two consecutive standards fail to meet accuracy
acceptance criteria as the resulted change in calibration curve slope
and intercept values will have impact more on the QC level lying
closer to the removed standards, more so in the weighted regression.
So, to conclude, it is more defendable and scientifically justifiable
if you are accepting truncated CC in cases where there is an issue
with the established lowest or highest standard and not with the
blanks and/or blank IS ( i.e., Zero Standard). Moreover, for any
analytical batch to be accepted, minimum three levels of QCs viz,
LQC, MQC and HQC are required to be present as it is the very
criteria of constructing a batch and these are key constituents of
any analytical batch applied to do bioanalysis.
Hope it helps.
Regards
Neel Kamal Mohan
Back to the Top
Dear Debbie,
In my opinion, you should not truncate the calibration curve for
passing the batch. This clearly indicates the production of bias
which will not be accepted by any regulatory. The reassay of the
subject is the correct way and if the sample volume is enough then
you should go for it.
Dr. Mandar Mote
Head Bio-analytical
CRO
Cadila Pharmaceuticals Ltd.
Ahmedabad.
Back to the Top
The following message was posted to: PharmPK
Dear Debbie,
I think you should not truncate cc (especially remove lower points)
just for passing blank and blank IS or just for passing a batch.
I hope you would have done validation of method using your standard 1
as LOQ. If you remove standard 1 and 2 during subject sample
analysis, obviously LOQ changes (as you are claiming), which should
not be done. The CC range and CC points are fixed, which means they
should be constant during entire process of validation and subject
sample analysis. If you are reading blank and blank IS from
particular LOQ during validation, I think it should be same LOQ
during subject sample analysis also. There are some points to be noted:
1. If you are reading blank and blank IS from another point (which is
above original LOQ) the question of contamination raises.
2. As new LOQ is different (higher than original LOQ), some of
subject samples may fall BLQ (below LOQ), which may not be the case
on the other day.
3. Even though you have truncated CC, it should be always ensured
that all QC's are set to be within CC range. And obviously in this
case, your QC's are also not acceptable as you are changing LOQ.
Additionally, you can not accept subject samples batch as there is no
QC (low).
So, I would recommend you to go for the reanalysis of samples rather
than changing LOQ as it is an indication of bias.
I agree with Mr. Neel Kamal Mohan. It is always good to have
duplicate samples of lower, higher standards, blank and blank IS.
I feel its better not to accept a batch if two consecutive lower/
higher standards are as there is a possibility of change in CC slope
and intercept. If you are rejecting two consecutive CC points in the
middle range, I think there would not be more impact on slope and
intercept. So, you can reject two, but not three in this scenario.
Regards
Sivacharan K.
Back to the Top
The following message was posted to: PharmPK
Hi,
It looks to me that your case indicates that you
either had a contamination or carry over, or an
interference peak in your plasma lot used.
In my opnion, it is not neccessary to truncate your
std curve since every points on the std curve meet
your criteria.
Xiaodong
Back to the Top
Dear Debbie,
First of all, have you checked if your blank is contaminated? It may
very well be, that you have no reason to truncate your curve but,
rather, to prepare new blanks. As indication to that might be the raw
data of std.1 and 2: are their responses as expected from a linear
curve? If you back calculate the concentration of std 1 and 2
according to the calibration curve (even the truncated one, by
extrapolation), do you get results close enough to the expected value
of those standards? If so, is likely that they are above LOQ, and
truncation is not necessary. Also, you can use your lowest QC as a
test: if the full calibration curve "translates" QC1 signal
correctly, then the curve is all right, and the problem lies with
your blanks.
If you did truncate your calibration curve because the problem does
not lie with the blanks, then obviously the low QC might be now below
your lowest standard. You can certainly accept all the test results
which are above std 3, provided that the results of the QC's of the
relevant concentrations are within acceptable range. Test results
below your new LOQ will have to be retested, or reported as "not
detected", stating the LOQ of the run.
I hope I have been of some help. Good luck and best regards,
Michal Rotenberg, PhD
Lab. Of Clinical Toxicology & Pharmacology
Sheba Medical Centre
Israel
Michal.Rotenberg.at.sheba.health.gov.il
Back to the Top
The following message was posted to: PharmPK
If you can no longer reach your LLOQ, you should revalidate the method
at an LLOQ you can reach, or modify the method using sample clean up
that will allow you to reach the LLOQ you had initially. In either case
you will need to re-validate. In the former case you will need to
prepare new QCs, in the latter case should work.
Removing two consecutive standards is generally not done. And the
confusion and retesting needed after resetting the LLOQ is really not
work the effort. Your assay is telling you that you have selected a
LLOQ that cannot be reached under usual assay conditions.
Resetting the LLOQ on the fly is more than a little suspect. Consider
the effort and discipline required to establish an LLOQ usually 3
accuracy and precision runs of six samples each for the LLOQ as well as
checking the LLOQ through several stability measures.
--
Ed O'Connor, Ph.D.
Laboratory Director
Matrix BioAnalytical Laboratories
25 Science Park at Yale
New Haven, CT 06511
Web: www.matrixbioanalytical.com
Email: eoconnor.-at-.matrixbioanalytical.com
Want to post a follow-up message on this topic?
If this link does not work with your browser send a follow-up message to PharmPK@boomer.org with "Calibration curve acceptance criteria" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)