Back to the Top
Dear all,
For some analysis by HPLC, we use the zero sample (matrix sample
processed with internal standard) to make the calibration curve.
We include it for the regression calculation as a calibration
standard. We use linear regression with offset (y=ax+b), not forced
through zero (y=ax).
Is it correct?
In guidelines I've found "A calibration curve should consist of a
zero sample and 6 to 8 non-zero samples ..."
Does it mean that we should use the zero sample to do the regression?
With this method, we never need to use weighted least squares
regression. We always use "unweight" regression...
Another point is the selectivity of the method.
Could we accept interfering peaks, for example 10 or 20% of the LOQ?
As the first point, it's unclear in the guidelines.
Thanks.
Fabrice Guillet
Biopredic
Back to the Top
The following message was posted to: PharmPK
FDA guidelines indicate that the lowest calibration standard defines the
lower quantification limit of an HPLC assay. Use of a zero calibration
standard in HPLC assays is not recommended and not industry standard
practice. (Immunoassay may be different).
Thomas L. Tarnowski, Ph. D.
Project Team Leader
Group Leader / Dept. Head, Drug Metabolism and Pharmacokinetics
Roche Palo Alto
3431 Hillview Avenue, Palo Alto, CA 94304
* Phone: (650) 852-3182
* FAX: (650) 855-6428
* email tom.tarnowski.aaa.roche.com
Back to the Top
The following message was posted to: PharmPK
Dear Fabrice,
The use of the zero sample for drawing the calibration curve is not
advisable.
The use of a weighting factor is not necessary but a weighting factor
can help to reduce the variance of your results throughout the different
calibration levels.
Concerning interfering peak there are no well defined criteria but if
you have an interfering peak higher than 20 % of the LLOQ this peak will
influence your results because acceptance criteria at the LLOQ for
accuracy is 20 %. It's advisable to avoid interfering peak but if you
can't eliminate it by modifying extraction procedure or chromatographic
conditions don't work with peak higher than 10 % of the LLOQ.
Best regards.
Sebastien
Dr. Sebastien Block
Manager Bioanalysis
SGS India Pvt. Ltd.
Life Science Services
Ticel Biopark Ltd
Chennai, 600113, INDIA
Back to the Top
Dear Fabrice,
As cited from the FDA guidance: "A calibration curve should consist
of a blank sample (matrix sample processed without internal
standard), a zero sample (matrix sample processed with internal
standard), and six to eight non-zero samples covering the expected
range including LLOQ".
And later on in section F: Specific Recommendations for Method
Validation:
"The matrix -based standard curve should consist of a minimum of six
standard points, excluding blanks, ...."
OK? Is a "zero sample" not a special kind of "blank sample"? - hard
to guess ;-)
Blank sample: obviously not relevant for the calibration curve - but
you can yield information on selectivity issues from this sample.
Zero sample: Your decision - I see no problem in including zero
samples for regression calculations because in a proper method this
will not significantly alter the regression parameters (if you do not
force the curve through zero which is not recommended). In addition
you may get some information on selectivity (IS vs. analyte) from
this sample.
Nevertheless, as Tom stated correctly, due to the fact that the
calibration range is only defined between the LLOQ and the highest
calibration standard (ULOQ) it is not usually standard practice to
include zero samples. In addition - only for non-zero samples you can
derive any information on the batch acceptance criteria like
percentage deviations from the back calculated vs. nominal values,
accuracy and precision.
Weighted regression:
The use of weighting factors should be oriented the design of your
calibration: If you choose equidistant concentration steps between
your calibrators you should use no weighting - If you use a
"logarithmic" distribution with narrower steps at the lower
concentrations end of the calibration line you will better use 1/x or
1/x2 weighting.
Selectivity / Interfering peaks:
You may can get information from the FDA Guidance: Section C: see
definition of LLOQ:
"The analyte response at the LLOQ should be at least 5 times the
response compared to blank response" This can be interpreted as 20%
of an analyte signal at the LLOQ is suitable - and this value also
correlates to the LLOQ accuracy acceptance criteria of 80-120%.
Nevertheless, it will be wise to optimize your method, check it for
some systematic errors like carryover and choose some safety margin
from this value in routine work.
Besides that, with a method that is really on the edge you will run
into problems to "quantify" 20% of the LLOQ response in a
reproducible way.
In addition - but maybe not relevant for your method: interfering
peaks that affect the signal of the internal standard in an
irreproducible way will have serious impacts on quantification issues
because a bias on the IS does effect significantly the quantification
of all samples not only of samples at analyte concentrations near the
LLOQ. Therefore for the IS a "smaller" bias (not speaking in absolute
values) of for example 5 % of the IS peak area is recommended.
Best wishes
Sven
Dr. Sven Poetzsch
Merck KGaA
Central Analytical R & D . GLP - quantitative Bioanalytics
WL/ZD-A/ZFA/ZFA10
Location: A18 / 102b
Frankfurter Strasse 250
D 64293 Darmstadt
Phone: +49 (0)6151/72-5323
Fax: +49 (0)6151/72-94121
Email: sven.poetzsch.-a-.merck.de
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)