Back to the Top
Dear all,
1. Please clarifies whether, Statistical Outliers in the Quality
control samples cannot be taken for calculating Accuracy and
Precision. (During inter and intra day results reporting in
bioanalytical method validation). As per USFDA guidelines identify
the statistical outlier in the bioanalytical report, but it not
applied in the calculation. Same to be applied for study sample
analysis?
2. Whether the Inter day and Intra day Accuracy and Precision tables
can be reported with and without QC outliers in the bioanalytical
report?
Thanks and regards,
GS
g.Sundar
Back to the Top
The following message was posted to: PharmPK
I believe they (regs) suggest you report both with and without
Ed O'Connor, PhD
Technical Director, Immunoanalytical
Tandem Laboratories
115 Silvia Street
West Trenton, New Jersey
609-228-0243
Back to the Top
The following message was posted to: PharmPK
Dear Sundar,
I believe that one can only eliminate a QC or calibrant result from
summary
statistics if one has a definite reason for removing the data point from
consideration. If a method produces occasional random results, there
may be
any number of excellent analytical reasons for the outlier BUT you
may not
have a method suitable for reliable analysis of clinical samples.
For instance, if a technician or robot ads too much or too little of an
internal standard to a sample, the ratio of analyte to internal standard
will not reflect the intent of the method (i.e., precise and accurate
addition of an identical amount of internal standard to each sample).
It may
be acceptable, in the case of a mis-spike, to eliminate a QC from
summary
statistics, but the result would be misleading - it may be that the
techniques used in your laboratory will randomly result in poor
quantitation
of clinical samples (and the ensuing gnashing of teach in the
pharmacokineticist's office). This also raises the question of to what
criteria must internal standard peak height or area measurements
comply so
that a mis-spike might be identified.
Additionally, there is much evidence that some transient materials in
some
clinical samples (as well as QC and calibration standard solutions) can
enhance or suppress ionization of the analyte, thus producing an
artifactual
result for an individual sample or standard. If this is occurring in
your
QC/calibrant standards, it is not sufficient to trade in blank matrix
until
one finds a non-interfering matrix lot. It is much more important to
improve
the separation and/or detection method so that the artifacts do not
surprise
the analyst - or the kineticists.
In summary, it is not whether a QC/calibrant CAN be ignored for the
purposes
of reporting summary validation or in-study statistics, but whether the
analyst has a sufficiently robust method to conduct the clinical sample
analyses.
It should always be the analytical science that informs the
statistics, not
what the statistics can do to make the method seem better than it is.
Best wishes,
Ian M. Davis, M.S.
Back to the Top
I agree with Ed Conner's post that the regs suggest you report the
results as recorded as well as modified (to eliminate the outlier).
And justify your stats for regarding the datum as an outlier.
Worse yet, regardless of how you assess a datum as a "statistical
outlier", my last reading on the subject suggested that if you don't
like the data, do the whole study over, or at least repeat the
measurement in question.
David Farrier
/\ /\
SummitPK.com /\ / \ /\ / \
/ / / /\ / \
==
David S. Farrier, Ph.D. Phone: 970-249-1389
Summit Research Services Fax: 970-249-1360
68911 Open Field Dr. Email: DFarrier.at.SummitPK.com
Montrose, CO 81401 Web: http://www.SummitPK.com
Back to the Top
The following message was posted to: PharmPK
David Farrier wrote "if you don't like the data, do the whole study
over, or at least repeat the measurement in question."
Just repeating the measurement in question is guaranteed (on average)
to bias the results because of regression to the mean. If you repeat
one measurement you need to repeat and use them all to avoid this bias.
Nick
--
Nick Holford, Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New
Zealand
email:n.holford.-at-.auckland.ac.nz tel:+64(9)373-7599x86730 fax:373-7556
http://www.health.auckland.ac.nz/pharmacology/staff/nholford/
Back to the Top
Dear Friends
I have a query regarding Statistical outlier..in BA/BE studies
1.If my data has n=10 (example) how many times the outlier can be
applied, is it indefinite or there is a limit. Does it depend on no
of data samples, ex. in long term stability if one qc out of six (the
batch is acceptable) is showing erratic behaviour , Do we need to
report it as an outlier, do i need to repeat the outlier test for
remaining 5 qc and henceforth continue.
2.if in a study (pivotol study n=40) all my batches are passing the
QC acceptance criteria (67% and 50%), do we still need to use the
outlier test.
3. are there any limitations on using outlier as per any pharmacoepia
(USP etc..), compendium, can anyone pls provide me a reference.
Pls reply its kinda urgent
Thanks in advance for your help
--
Bharat
Back to the Top
The following message was posted to: PharmPK
I believe it can be applied down to the minimum number of samples for
each test- I believe for Dixon's and Grubbs it is n=3. Once you reject
an outlier, you recalculate and retest for the next outlier.
Why apply this to QCs? The higher ruling is that for three levels of
QC, in duplicate sets, 4 of 6 QC must pass with at least half of the QC
surviving at each level. For stability samples, where you measure 3
samples at each point, 2/3 must pass. If you measure 6, 4 must pass. If
you use a different number of samples say 10, then at least 7 must pass
(~2/3).
It sounds like your stability QC are passing. You are already rejecting
the data because it failed your acceptance why look at an outlier test?
Ed O'Connor, PhD
Technical Director, Immunoanalytical
Tandem Laboratories
115 Silvia Street
West Trenton, New Jersey
609-228-0243
Back to the Top
Dear Bharat,
you can apply the grubbs test for zoutliers for no more than 2
samples with a set of 6 nos as a set. In Bioanalytical application we
usually have 6 QC in each set and 67% of each should pass the set.
Therefore the Grubbs test becomes invalid for application because its
applicaation to the 3rd will imply that 50% are within the limit and
50% tested for outliers.
For more number of samples like more than 6, you can apply the same
test and the rest will rely upon your SOP guidelines. for the Grubbs
test, have a look at http://www.itl.nist.gov/div898/handbook/eda/
section3/eda35.htm.
Inviting opinion of experts,
Regards,
Santosh Tata
Bioanalytical Division, BEC,
Apotex Inc, Bangalore
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 "Statistical outlier" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)