Dear Anders, Andy and others,
Through this exchange has reminded me of an apparent conflict in approaches to
laboratory results, QC and quality standards that has interested me for a while.
1. The Biological Variation (BV) / Critical Difference (CD) paradigm. In this
approach the theoretical background includes as an input or "output" CVa
(analytical CV). This is used as an input for Critical Difference calculations and
an output by way pf providing quality speficiations for CVa from within subject
biological variation (CVi). In this circumstance the CVa is generally taken from
QC material from an analyser in routine use conditions.
The BV provides an objective quality standard and assay performance is the
measured CV.
2. The Westgard Critical Error paradigm. The Westard approach identifies (amongst
other things) the amount of error that may occur with certain levels of
probability for known values of CVa and the QC system in use. Thus using standard
QC processes and westgard rules there can be a shift of over 2.5 SD which will
remain undetected over 10% of the time. This approach allows assessment of the
likelihood that an assay will meet pre-determined quality specifications given the
CVa and the QC processes.
This is related to the concept of assay capability (SD as a fraction of allowable
error).
While the process is very robust in terms of possible result variation, the the
missing part appears to be the link between the size of the critical error and
clinical decision making.
Some thoughts:
In an ideal world the reconcilliation of these two approaches would be that the
CVa should be sufficiently low that even if the assay drifted the maximal
undetected limit, the result would still not affect clinical decision making. For
CD decisions, this means that the CVa would ideally be less than approx 1/3rd of
the chosen standard (eg "Desirable" CVa < 0.5 x CVi) using Frasers approach. Ie
CVa standard would be approx 0.16 x CVi. This would make it very difficult to meet
quality specificiations for most analytes.
Of course because an assay might vary that much without being detected does not
mean that it will. We have many assays which almost never shift by the maximal
undetectable amount and even fewer that do so without being detected (this is a
90% probability of the shift being detected). For stable assays the BV paradigm
using CVa measured using standard QC seems appropriate. For very unstable assays
the combination of the critical error and BV approach may be a better paradigm.
Alternatively a better way may be to measure CVa with different material than that
with which we control the assay (ie no excluded data) to provide a better measure
of CVa.
It would be nice to be able to align these two concepts of required laboratory
precision.
Regards,
Graham
Graham Jones
Staff Specialist in Chemical Pathology
St Vincent's Hospital, Sydney
Ph: (02) 8382-9160
Fax: (02) 8382-2489
>>> Andy Minett <[log in to unmask]> 25/01/2010 1:34 am >>>
Dear Mr Kallner,
Thank you for your reply!
I think my previous response was probably badly worded and maybe didn't get
across the notion is was intended to. I have a tendency to type as I think,
which I expect comes across a little confounding.
My intention was to examine this statement in your initial letter:
"This is the quantity that should accompany each result, which is
impractical in the
clinical setting but should be available for those interested and used as
limits for the IQC."
I was hoping to suggest that it is necessary to include a second component
in the calculation of IQC limits, which is the ability of the IQC effort to
detect error. Westgard explains it better than I can, and terms this
component "critical error" (see here http://www.westgard.com/lesson5.htm ).
Without this, we assume that the analysers are completely error free (beyond
stable variance).
As such, I use the calculation:
Total Error = CV.(Critical Error + Z) + Bias
Otherwise, I concur completely with the rest of the letter, and am grateful
for the links/references. Do you have links to the work of Callum Fraser and
Per Hyltoft Petersen I could look at?
Thank you for your responses so far, I enjoy discussing this type of
mathematics/statistics...although I find it to be quite a lonely corner of
my work!
Best regards,
Andy Minett
------ACB discussion List Information--------
This is an open discussion list for the academic and clinical community working
in clinical biochemistry.
Please note, archived messages are public and can be viewed via the internet.
Views expressed are those of the individual and they are responsible for all
message content.
ACB Web Site
http://www.acb.org.uk
Green Laboratories Work
http://www.laboratorymedicine.nhs.uk
List Archives
http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html
List Instructions (How to leave etc.)
http://www.jiscmail.ac.uk/
**********************************************************************
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please notify
the system manager.
This footnote also confirms that this email message has been virus
scanned and although no viruses were detected by the system, St Vincents &
Mater Health Sydney accepts no liability for any consequential damage
resulting from email containing any computer viruses.
**********************************************************************
------ACB discussion List Information--------
This is an open discussion list for the academic and clinical community working in clinical biochemistry.
Please note, archived messages are public and can be viewed via the internet. Views expressed are those of the individual and they are responsible for all message content.
ACB Web Site
http://www.acb.org.uk
Green Laboratories Work
http://www.laboratorymedicine.nhs.uk
List Archives
http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html
List Instructions (How to leave etc.)
http://www.jiscmail.ac.uk/
|