Method Validation Revisited: Its Role in Traceability of Results

Wolfhard WegscheiderWolfhard Wegscheider

Institute for General and Analytical Chemistry Institute for General and Analytical Chemistry University of Leoben, AUSTRIAUniversity of Leoben, AUSTRIA

wegschei@unileoben.ac.atwegschei@unileoben.ac.at

Standard view of the relation between uncertainty and

validation

MeasurementMeasurement

uncertaintyuncertaintyMeasurementMeasurement

uncertaintyuncertainty

otherother

effectseffects

biasbiasprecisionprecision

from SLR Ellison 1999from SLR Ellison 1999

Traditional roles of method validation

• Establish performance characteristics– Linearity– Limits of detection/determination– Precision: repeatability, intermediate

reproducibility– Effect of concomitants

• Present data for approval of method• Produce control limits for everyday

operation

What shouldshould be the role of validation ?

As analytical chemistry is about making decisions

• validation should support the decision making process

results

limit

Traditional role of traceability

• Provide a firm and identical base of units worldwide

• Supply this base in a manner stable in time

• Underpin the hierarchy of measurements/procedures/laboratories

• .....

Requirements of customer

Convert to uncertainty

Apportion to the different steps

Express as figures of merit

Specify validation plan

Validate

Compute figures of merit

Estimate factual uncertainty

Accept/reject contract

Do analyses

Report to customerReport results

Validation starts out from the customer´s needs

Two key components to traceability:

stu

pru

22

prstxuuu

Uncertainty of a standard:

• Identity• Purity• Preparation

Influence of uncertainty of standards in calibration

• NO uncertainty from standard

• With uncertainty from standard and dilution

M. Rösslein, EMPA2001

Two (extreme) ways to define the measurand / analyte

A) Careful and complete description of circumstances:– (exactly) what species– What (kind of)

samples/concomitants– Which environmental conditions

(p, T)

B) Measure as specified and give the analyte a name

A general model of measurement and validation in chemistry

ijkx

ijkijkx valuetrue

resultxijk

......

....

huge and unknown

)()(ijkijk

fxu ijkijk

x ´ for N results:

somewhat smaller,but unknown

A general model of measurement and validation in chemistry

After validation:

ijkkjiijkx

small and accessiblefrom precision data

identified effects on result

Two extreme ways to define the measurand /analyte: MODEL 1

A) Measure as specified and give the analyte a name:

ijkkjiijkx

Accessible throughinterlab comparisons

The „new“ analyte)()(

ijkijkfxu

Two extreme ways to define the measurand /analyte: MODEL 2

ijkkjiijkx

Very small partof uncertaintyThe „old“ analyte well defined

B) Careful and complete description of circumstances:

),,,()(ijkkjiijk

fxu

The Measurement of pHIUPAC Working Party on pH

• Notational definition:

• Operational definition:

0lglg

m

mapH HH

H

AgAgClClSbufferHPt ,2

Operational Definition of pH

• Debye-Hückel formalism with Bates-Guggenheim convention

• (only) 5 primary buffers• 3 < pH < 10• I < 0.1 mol/l• Aqueous solution

unce

rtai

nty

(a.u

.)un

cert

aint

y (a

.u.)

a b c

bottom-upbottom-up

real+ udef

avoidavoid+ u+ udef def + u+ ucorrcorr

Uncertainty of standard methods:experience in practice

Redesigning validation studies:

what has been missing so far ?

• Establish the quality of the standards

• Complete coverage of scope OR clear redefinition of the analyte

• Anchored in the requirements of the customer

Traceability of results and reference values is a central issue in modern laboratory operation. It is not an end in itself, but serves the purpose of achieving a reliable result.

Traceability of results can only be claimed if results are accompanied by an uncertainty statement based on traceability of all references, chemical and physical, as well as on procedural contributions to uncertainty.

A result must be "fit for purpose", thus estimation of measurement uncertainty from uncertainties of references and procedures is added value for laboratories and simple when guidelines are followed.