First I agree with Johnathan that it's imperative
to understand the meaning of the words one uses.
This should not be brushed aside as mere
"semantics", and nothing to do with the real,
practical, world in which we live and work;
furthermore it's important that agreement is
reached regarding the concepts that a word
represents.
In particular, it's essential that the scientific
definitions of terms should accord as closely as
possible with their common use in a language
(generally English); otherwise confusion
inevitably occurs. Also that the definition
should be logically unimpeachable.
Anders Kallner refers in this context to the
recommendations of the new edition of VIM.
My (Italian) wife has a simple (albeit
statistically challengeable) test for deciding
whether an Italian cookbook is worthy of
consideration: i.e. its description of the
preparation of the well known Milanese dish "ossi
bucchi". (Elizabeth David's Italian Cooking is
one of the few that survives this test.)
My own simple test of the trust one can place in
a glossary of scientific terms used the assay
field is its definition of the term
"sensitivity". (This is, après tous, one of the
basic terms we all use to describe assay
performance; moreover one of the common aims of
the analyst is to maximize an assay's sensitivity
and precision. )
Note that, in general terms, an assay system or
instrument is viewed as more "sensitive" if it
enables measurement or detection of a smaller
amount or quantity of that which the system is
designed to measure. Countless examples of such
usage are to be found, for example, in articles
published in Scientific American.
So what does VIM say about sensitivity ?
4.12
sensitivity of a measuring system
sensitivity
quotient of the change in the indication of a
measuring system and the corresponding
change in the value of the quantity being measured
NOTES
1 The sensitivity can depend on the value of the quantity being measured.
2 The change considered in the value of the
quantity being measured must be large compared
with
the resolution of the measuring system.
Stated mathematically this somewhat opaque and
poorly presented definition (which quotient?) can
be re-written as dR/dD where dR is the change in
the response caused by a change in dose (ie the
quantity being measured) of dD; i.e the slope of
the dose-response curve . (Clearly dR/dD can
change if the curve is not linear).
Years ago I pointed out to René Dybkaer that this
definition is untenable. He totally rejected
this charge. His position was that one can
define a word in any way one likes. (In a sense
this is, of course, true, but not if the outcome
is nonsense.) So I'm not in the least surprised
that this definition continues to be promulgated
in the latest edition of VIM.
So why is the definition nonsensical?
Members of this discussion group may remember my
posting a picture of two balances (again
attached) and asking which of the two balances
shown was the more sensitive Regettably Dr
Dybkaer was not amongst the >100 members who
replied, so I can only wonder what his reply
would have been. Of those that replied,
approximately 47% opted for balance A, a roughly
equal number opted for balance B and the
remainder were 'don't knows'.
One may think members of the ACB have reached a
pretty pass if they don't know which of two
balances is the more sensitive. But they are to
be forgiven. The American Chemical Society has
defined a balance's sensitivity (on different
occasions) as A. the movement of the pointer on
the balance scale. and B. the rotation of the
balance beam, when a weight is placed on the pan.
In other words, a definition that essentially
accords with the VIM definition.
It's perhaps not surprising that Mettler
subsequently published and circulated a long
protest explaining why it considered the ACS
definition(s) to be "unworkable" (the copy of
which I've unfortunately since lost).
But ACB members may not be interested in the
sensitivity of balances, so we should perhaps
move closer to their field by considering the
sensitivity of two RIAs. The second (attached)
diagram shows the response curves plotted using
the same system, but using Hi and Lo antibody
concentrations, plotting the (same data) data in
terms of Free/Bound (F/B) labelled antigen (1) or
B/F (2).
Someone plotting the data in terms of F/B guided
by VIM will conclude that sensitivity is improved
using the Lo amount of antibody; the converse if
he/she plots exactly the same data in the form
of B/F.
Which is absurd.
And which represents another example of proving
the falsity of a premiss by showing the logical
consequence to be absurd.
Except that we are not proving the falsity of a
premiss, but the falsity of the definition of an
important term by VIM, and which its authors
refuse to retract.
This is again not a synthetic and academic
example (or a case of my being a pedant) with no
practical relevance. On countless occasions I've
witnessed immunoassay practitioners reaching
totally incorrect conclusions in consequence of
not understanding the concepts underlying the
term "sensitivity".
And the fallacious concepts embodied in the VIM
definition also formed the basis of Berson and
Yalow's theoretical work on RIA design, which
purportedly showed that an RIA system should rely
on the use of an antibody concentration that
binds 33% of the labelled antigen to maximize
assay sensitivity.
So I would suggest that one carefully examines
the definitions of terms advocated in VIM.
Roger Ekins
>
>Taking a break from preparing for our CPA EQA inspection!!
>
>
>Definitions of 'analyte' on the web
>===================================
>
># the substance which a laboratory test aims to
>detect. In cholesterol testing, for example, the
>analyte is cholesterol. In genetic testing, the
>analyte could be, for example, a specific allele
>or genetic mutation.
>www.cs.uu.nl/people/ronnie/local/genome/a.html
>
># (an·a·lyte) (an¢[schwa]-l[imacr]t) a substance undergoing analysis.
>www.mercksource.com/pp/us/cns/cns_hl_dorlands.jspzQzpgzEzzSzppdocszSzuszSzcommonzSzdorlandszSzdorlandzSzdmd_a_34zPzhtm
>
># A substance measured in the laboratory. A
>chemical for which a sample (such as water, air,
>or blood) is tested in a laboratory. For
>example, if the analyte is mercury, the
>laboratory test will determine the amount of
>mercury in the sample (ATSDR 2003).
>www.racteam.com/LANLRisk/Glossary.htm
>
># Identifies the analyte that was measured. See
>Analyte Categories With Associated Analytes for
>association between analyte category and analyte.
>www.cropcomposition.org/databaseorg/Glossary%20of%20Terms.htm
>
># The substance measured by a laboratory test;
>for instance, a specific mutation or allele.
>www.cdc.gov/genomics/gtesting/ACCE/FBR/CF/CFGlossary2.htm
>
># Substance to be measured.
>ojjdp.ncjrs.org/PUBS/drugid/glossary.html
>
># A name given to describe a substance which is
>being produced in an analytical procedure.
>www.pestmanagement.co.uk/lib/glossary/glossary_a.shtml
>
># A chemical being tested for in a laboratory test.
>www.rrmsc.com/resources/glossary.htm
>
># An Analyte is the substance or chemical
>constituent that is undergoing analysis. The
>substance being measured in an analytical
>procedure.
>en.wikipedia.org/wiki/Analyte
>
>======================
>
>The common theme seems to be 'a substance that
>is measured/analysed in a laboratory'.
>I fear we may get into more arguments if we come
>to define 'substance', 'measure' or 'laboratory!
>
>I can't make any statements about the physical
>world because this may just be a construct of my
>consciousness, which may itself just be an
>emergent property of brain complexity!
>
>======================
>
>To be constructive though, is this all about
>naming things? We use the scientific process to
>refine our knowledge of the components of the
>natural world and in clinical biochemistry we
>concentrate on the components of biological
>systems, especially those that help us
>understand health and disease. We observe,
>separate out, analyse, purify, characterise
>things and give them names. By doing so we
>actually bring those things into existence in
>our 'universe of discourse' so that we can
>refine techniques to study them and ultimately
>measure them for the good of patients. Some
>things are physical entities which we can
>separate out and name and others are attributes
>(I hesitate to call them properties!) because we
>don’t know enough about what generates what we
>observe.
>
>Some things are easy to name and are truly
>unique - elements, simple compounds etc. Others
>are more difficult - proteins etc - which are
>heterogeneous and exist in multiple forms in
>health and disease. In respect of the latter we
>end up with names of things that imply
>uniqueness (to us and our customers) but the
>reality is different. JCTLM classify 'analytes'
>as type 1 (the easy ones as above) and type 2
>(the difficult ill-defined ones)
>
>The ordinary man in the street would naturally
>assume that if we say we are measuring analyte X
>in his blood, then analyte X must exist and we
>must know exactly what analyte X is otherwise we
>can't possibly measure it. I try to understand
>what Roger and Anders are saying but I could not
>possibly explain their detailed metrological
>philosophy to a patient (or a doctor).
>Enjoyable as this discussion has been we cannot
>escape the fact that we have to agree amongst
>ourselves what we do and then communicate this
>to a lay public. So in Mike's real world of
>'consensus reality' where we have to demonstrate
>and prove to our customers and paymasters that
>what we do adds value to the clinical event, we
>must be sure in our own minds what we mean by
>these words.
>
>As a starting point for discussion, might we
>agree on the following working ‘common sense’
>concepts for the 'things' that we list in our
>test repertoire:
>
>Type 1 component - an entity which is uniquely
>named, chemically defined, exists in the same
>form wherever it is encountered, where a
>standard exists which can calibrate measurement
>procedures, and where results can be expressed
>in method-independent SI units of mass or molar
>amount - eg Cortisol
>
>Type 2 component - a group of closely related
>entities which share structure and function and
>thus may be named under a single heading, which
>can be purified and characterised to some
>extent, but without certain knowledge that these
>entities are the same in the standard material
>as are in the patient, and where results are
>expressed in arbitrary units which are
>method-dependent - eg TSH
>
>Attribute - an observed characteristic of a
>sample which can be quantitated in arbitrary
>unitage in a defined system - eg ‘ischaemia
>modified albumin’ (serum cobalt binding)
>
>These are all analytes (probably) but we have to
>think about them in different ways when we
>discuss and assess comparability, trueness,
>uncertainty, standardisation and traceability
>and when we communicate to our users on their
>meaning and significance.
>
>My assertions:
>
>Type 1 component - absolutely no option but to
>achieve common standardisation and harmonised
>results using established reference measurement
>systems with true metrological traceability
>
>Type 2 component - the way forward for many of
>these is exemplified in the approach taken by
>the the ISOBM TD-7 Workshop on hCG and related
>molecules : Berger et al. Towards user-oriented
>standardization of pregnancy and tumor
>diagnosis: assignment of epitopes to the
>three-dimensional structure of diagnostically
>and commercially relevant monoclonal antibodies
>directed against human chorionic gonadotropin
>and derivatives. Tumour Biol. 2002
>Jan-Feb;23(1):1-38.
>
>Attribute - more hard scientific work to sort
>out what is going on before we make claims about
>what is being measured and how useful it is.
>
>Sorry if this is repeating myself and maybe stating the obvious!
>
>Hope this is helpful
>
>J
>
>Back to the Quality Management System now!
>
>
>Dr Jonathan Middle
>Deputy Director, UK NEQAS Birmingham
>0121 414 7300, fax 0121 414 1179
>-----------------------------------------------
>All opinions expressed in this email are mine
>alone and are not necessarily representative of
>the views of the UK NEQAS organisation, UK NEQAS
>Birmingham (Wolfson EQA Laboratory), University
>Hospital Birmingham NHS Foundation Trust or
>University of Birmingham.
>
>
>
>-----Original Message-----
>From: Clinical biochemistry discussion list on behalf of Mike Toop
>Sent: Wed 22/11/2006 19:58
>To: [log in to unmask]
>Subject: Re: Quantities and properties
>
>Pedantry is a great spectator sport!
>
>Here is my (brief ) 5 pence worth
>
>What do we mean by analyte and how does that definition relate to the
>physical world?
>
>
>
>Mike
>
>
>
>------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
>List Archives
>http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html
>List Instructions (How to leave etc.)
>http://www.jiscmail.ac.uk/
--
Regards
Roger
Molecular Endocrinology
University College London Medical School
London W1N 8AA
Fax +44 20 7580 2737
Phone +44 20 7679 9410
The information in this e-mail is confidential and may be legally
privileged. It is intended for the exclusive attention of the addressee
stated above and should not be copied or
disclosed to any other without the express
agreement of the sender.
------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
List Archives
http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html
List Instructions (How to leave etc.)
http://www.jiscmail.ac.uk/
