Back to the Top
I would appreciate very much your help concerning the FDA rules for
calibration curve approval.
In LC-MS/MS bioassays, when performing a calibration curve
calculation, If one makes a calibration curve with 8 standard
concentrations AND all of these in duplicate (16 "points"), is it
valid to exclude one of the replicas (anyone who deviates more than
15%, or 20% for the LLOQ) and still clame for the approval of the
calibration curve since at least one of the replicas of each standard
concentration is still valid?
(In this example we would have 8 points failed, but still 8 points
Ok, having all the standard concentrations Ok due to the remaining
Thanks in advance for your comments,
Back to the Top
You asked about whether it would be acceptable to delete one of each
pair of duplicate 2 replicates from a set of 8 standard
concentrations determined in duplicate.
The 2001 FDA guideline on Bioanalytical Method Validation (http://
www.fda.gov/cder/guidance/4252fnl.htm) expects at least 6-8 points in
a standard curve, depending on the range. Later in the guideline it
"Matrix-based standard calibration samples: 75%, or a minimum of
six standards, when back-calculated (including ULOQ) should fall
within 15%, except for LLOQ, when it should be 20% of the nominal
value. Values falling outside these limits can be discarded, provided
they do not change the established model. "
If I understand the case you cited, only 50% fall within the
percentages, so it does not meet the 75% requirement. If routinely
50% of the points fail the 15/20% requirement, then your method is
not very robust. Either you have difficulties in preparing
calibration standards, or there is variability in processing and
analyzing samples, or there is some other problem. Note also that
the guideline says that values "can" be discarded, but not "must."
We usually tried to retain values, even if they were outside the
15/20% range, to avoid discarding valid data. The QC samples may
help to do this. Your method and your SOPs should define how to go
about rejecting data from calibration standards.
On the other hand, if you are actually trying to set up the assay so
that you can select the "better" of the 2 replicates for each
concentration, then you are trying to force the data through a
preconceived model, you are discarding what is likely good data, and
you are not really treating calibration standards the way unknowns
In either case, I would expect questions from the FDA, as well as
from other colleagues, about why such a practice is justifiable.
Also, from a practical point of view, such a method is not very
(e.g., run time. prep time, solvent consumption, etc.).
I'd be curious about others' opinions.
Thomas L. Tarnowski, Ph.D.
Bioanalytics / Preclinical Science
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (firstname.lastname@example.org)