You are on page 1of 14

Journal of Chromatography A, 1158 (2007) 3346

Review

Principles of analytical calibration/quantification for the


separation sciences
Luis Cuadros-Rodrguez , M. Gracia Bagur-Gonzalez, Mercedes Sanchez-Vinas,
Antonio Gonzalez-Casado, Antonio M. Gomez-Saez
Department of Analytical Chemistry, University of Granada, E-18071 Granada, Spain
Available online 16 March 2007

Abstract
Calibration is an operation whose main objective is to know the metrological status of a measurement system. Nevertheless, in analytical sciences,
calibration has special connotations since it is the basis to do the quantification of the amount of one or more components (analytes) in a sample, or
to obtain the value of one or more analytical parameters related with that quantity. Regarding this subject, the aim of analytical calibration is to find
an empiric relationship, called measurement function, which permits subsequently to calculate the values of the amount (x-variable) of a substance
in a sample, from the measured values on it of an analytical signal (y-variable). In this paper, the metrological bases of analytical calibration and
quantification are established and, the different work schemes and calibration methodologies, which can be applied depending on the characteristic
of the sample (analyte + matrix) to analyse, are distinguished and discussed. Likewise, the different terms and related names are clarified. A special
attention has been paid to those analytical methods which use separation techniques, in relation with its effect on calibration operations and later
analytical quantification.
2007 Elsevier B.V. All rights reserved.
Keywords: Chemical measurement processes; Metrological and analytical calibration; Analytical quantification; Calibration schemes; Calibration methodologies

Contents
1.
2.
3.

4.

5.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Metrological fundamentals: measurement function for analytical calibration/quantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Schemes for analytical calibration/quantification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1. Two-standard calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2. One-standard calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Methodologies for analytical calibration/quantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1. External calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2. Matrix-matched calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3. Standard addition calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4. Internal calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5. Calibration by internal normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33
35
37
38
38
38
39
40
42
43
44
45
45
45

1. Introduction

Corresponding author. Tel.: +34 958243296.


E-mail address: lcuadros@ugr.es (L. Cuadros-Rodrguez).

0021-9673/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2007.03.030

A chemical measurement process (CMP) is an analytical


method of dened structure that has been brought into a state of
statistical control, given the measurement conditions [1]. Calibration and validation are key operations in a CMP because they

34

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

provide traceability and comparability to the measurement, acting the former as a metrological interface between measurement
standards and analytical results. Nevertheless, in analytical sciences, calibration can be considered from two points of view
as metrological calibration as well as analytical calibration, not
being its requirements and aims similar necessarily. From a practical view, analytical calibration implies to verify or to state
the relationship between the measurement signal and the analyte/s quantity, whereas metrological calibration, based on the
analytical one, is just related to a measuring system understood as a set of one or more measuring instruments and often
other devices, including any reagent and supply, assembled
and adapted to give measured quantity values within specied
intervals for quantities of specied kinds [2]. The metrological
calibration implies the daily evaluation of the metrological performance of a measuring system in order to provide the quality
of the analytical results. Therefore, with a calibrated chemical
measurement system could be possible to obtain good results
although it provides erroneous indications, because these ones
can be corrected due to, in the calibration process, the error
is established. That is why, a metrological calibration leads
to the characterization of a dependence relationship between
the error committed by the measurement system and the value
of the measure, expressed in terms of parameters as deviation, bias, correction, correction factor or calibration factor.
This relationship is only established for the measurement values represented by the calibration standards used, but it can be
extended interpolating into the interval covered by them through
a correction/calibration function. Specific information about the
metrological approach of the calibration in a CMP can be found
in references [3,4].
Calibration in metrology, has been recently defined in a new
revision of the International Vocabulary of Metrology (VIM)
as the operation that, under specied conditions, in a rst step
establishes a relation between the quantity values with measurement uncertainties provided by measurement standards and the
corresponding indications [of a measuring system]1 with associated measurement uncertainties and, in a second step, uses this
information to establish a relation for obtaining a measurement
result from an indication [2] where a measurement standard
is referred to the realization of the denition of a given quantity,
with stated value and measurement uncertainty, used as a reference and it can be provided by a measuring system, a material
measure, or a certied reference material [2]. This new twostep definition generalizes the application of the previous VIMs
definition [5] still in effect, simplifies the wording/phrasing and
improves substantially its understanding. The main innovation
comes from the second step since as the VIM itself quoted, in
this definition the first one has been traditionally perceived as
the proper calibration. As it will be highlighted in this paper, this
second step could include all the analytical operations closely
related with calibration.
The aim of the chemical analysis of an object or a material,
is to obtain analytical information of it represented by the value

The text in brackets has been added by authors

of an input analytical quantity, X, through the measurement of


a physicochemical magnitude, Y, denoted as output analytical
quantity. That is why it is necessary to have a measurement
function, Y = f(X), which permits to link the values of the output analytical quantity with those corresponding to the input
analytical quantity. This one is a function of quantities, the value
of which, when calculated using known quantity values for the
input quantities in a measurement model, is a measured quantity
value of the output quantity in the measurement model [2].
In most cases, that link, always supported in a physicochemical principle, cannot be expressed in a general way by means
of a straightforward mathematical function, being necessary to
use empirical models adapted to each analytical system. At this
point, is where analytical calibration stress, due to it is the operation why an empirical mathematical function is established.
Analytical calibration is always an indirect calibration, because
is not possible to measure the quantity of a substance in a
system. There are different regulations where it can be found
some calibration definitions, that strictly speaking are definitions of analytical calibration, which are consistent with the one
established in the most recent VIMs version. Thus, a ISO standard [6] states that [analytical] calibration is a complete set of
operations which estimates under specied conditions the calibration function from observation of the response variable, y,
obtained on reference states (described by the values of variable states, x). Also, an IUPAC recommendation [7], defines the
[analytical] calibration as the operation that determines the functional relationship between measured values (signal intensities),
y-variable, and analytical quantities characterizing types of
analytes and their amount (content, concentration), x-variable.
It can be seen that in these definitions appear the most usual
analytical names for the y-variable (signal intensity, response),
and x-variable (type of analyte, analyte amount). This IUPAC
recommendation distinguishes two different kind of calibration
functions. The first one, for species identification and qualitative
analysis, and the second for quantitative analysis, being its main
purpose to obtain a function that allows to calculate amounts of
an analyte as a function of an instrumental signal (quantification
function). In most cases, the calibration function has to take into
account the response relations for all relevant constituents and
interferences.
These calibration functions are characteristics only for the
standards used and depend on the actual instrumental and operational conditions due to the presence of influence quantities,
i.e. quantities that, in a direct measurement does not affect
the quantity that is actually measured, but affects the relation
between the signal intensity and the measurement results. Small
changes in the influence quantity values could change the indication of the measurement system and that is why metrological
calibration by itself does not guarantee the comparability of the
measurements. In order to assure that the measurement system
remains in a calibration state, and that certain influence magnitudes are under statistical control, some periodical controls,
named verification operations, are required. Verification, in
this context, is defined as the provision of objective evidence
that a given item fulls specied requirements, taking any measurement uncertainty into consideration [2]. The verification

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

35

2. Metrological fundamentals: measurement function


for analytical calibration/quantication
The simplest way to write a measurement function for each
particular analytical calibration/quantification is:
Y = (x, m) X

Fig. 1. Flow-chart showing up the relationships between both analytical calibration, metrological calibration and analytical quantification.

is a confirmation that stated performance properties or quality


requirements of a measuring system, are achieved. So, when a
measurement system has been verified, the maintenance of its
calibration state is assured.
Thus, the main objective of calibration is to establish a
mathematical function from measurement standards which subsequently is applied to infer analytical information from the
samples/materials analysed on the basis of an initial hypothesis
in which implicitly is admitted that, samples and standards are
metrologically equivalents. In order to obtain a quantification
free of errors, is necessary to satisfy two basic requirements:
(i) standards and samples composition must be as similar as
possible and (ii) standards and samples must have an identical
behaviour in the measurement system. In any case, the standard
must be representative of the sample; usually the first situation
although desirable, is not essential being enough to fulfil the
second requirement to assure representativeness. The interrelation between both analytical calibration and quantification is
arranged in sequence in Fig. 1.
In general, the representativeness is satisfied using standards
of the analyte in the very material to be analysed. Nevertheless,
in those separation methods in which a multi-analyte quantification is going to be done, the calibration for each separate
analyte can be tedious or even impossible (if there are no
available standards), being necessary to take into account other
options.
In this paper, the necessary conditions to assure the representativeness and the strategies that can be adopted to minimize the
errors in the quantification when this one was not fulfilled, are
going to be established. Thus, the metrological fundamentals of
the calibration/quantification, the calibration schemes depending on the number of standards employed and the methodologies
that can be used for the calibration/quantification in separation
sciences, will be widely explained, emphasizing the practical
requirement for making measuring and discussing common misconceptions and errors that might arise. Consequently, names
as external calibration, internal calibration, internal normalization calibration, matrix-matched calibration, standard addition
calibration, and signal-ratio calibration together with terms as
external standard, surrogate or internal standard will be discussed and clarified.

where X and Y represent, respectively, the analyte amount (input


analytical quantity) and the analytical signal (output analytical quantity) and (x, m) is the measurement factor which
depends on the value x of the magnitude X, and on the type of
material (matrix) m where it was. This general expression, valid
in all circumstances, indicates that for each analyte amount and
matrix, the analytical signal is obtained by multiplying the true
analyte amount by a specific measurement factor. Nevertheless,
the simplicity of this expression is deceitful, because in practice
is impossible to characterize the -parameter and to know each
one of the values it can take.
In an ideal situation, the measurement factor would just be
characteristic of the type of analyte and independent of the analyte amount and/or the matrix; in this case, the CMP could be
labelled as selective. However, in analytical sciences the measuring systems show frequently low selectivity and the value
of the measures depend on different constituents in the material subjected to the CMP and the specific analytical system
used defined as, the range of circumstances that contribute to
the reliability of analytical measurements, including measurement system, reagents, procedures, test materials, personnel,
environment and quality assurance trials (adapted from [8]).
The selectivity of a measuring system is its capability, using
a specied measurement procedure, to provide measurement
results, for one or more measurands (quantities intended to be
measured), that do not depend on each other nor on any other
quantity in the system undergoing measurement [2]. Actually,
the analytical signal depends on: (i) the properties of the analyte species involved; (ii) the instrumental parameters of the
particular equipment; (iii) the experimental conditions (mostly
adjusted by the analyst); and (iv) the presence of accompanying substances [9]. In order to convert the previous expression
in other one more accessible from a practical point of view, to
improve the selectivity of the CMP, minimizing the lack of selectivity of the measuring system, some different strategies can be
applied. One of them, is to use a separation technique previous the measurement step in order to: (i) remove or diminish
the matrix constituents (clean-up) to do the -parameter independent from the rest of the constituents of the material, and
(ii) to separate the sample in fractions (in a continuous or in
a batch way), to obtain only an analyte in each one to assure
that the indication provided by the measuring system is only
attributed to the very analyte. This meaning of selectivity term
is coherent with IUPAC definition [10]: the extent to which the
method can be used to determine particular analytes in mixtures or matrices without interferences from other components
of similar behaviour.
As a consequence of this fact, the general expression of the
measurement function above described, could give rise to dif-

36

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

ferent mathematical expressions. Danzer [11] analyses a general


functional relationship adapted to the special features of a CMP
in analytical chemistry, discussing the variables that affect the
value of the analytical signal and the consequences over some
important characteristics of analytical methods.
One of the easiest way to describe the measurement function
for an analyte (not taken in consideration the matrix) is:
Y = Y0 + Yanal = Y0 + S(x) X
which indicates that the value of the analytical signal depends
on two components: one, independent of the quantity of analyte, denoted as blank signal (Y0 ) and the other, the net signal
generated by the analyte (Yanal ). This last is expressed as a
product of two terms, where S(x) is the sensitivity of the
measurement system (not necessarily a constant because its
value can be dependent on the actual x-value), and X, the
amount/concentration of analyte. This expression belongs to the
analytical calibration function obtained in the validation step of
the CMP, when the system is fitted to a linear response. Thus if
this equation is rearranged in the way:


Y0
Y=
+ S(x) X
X
the term in brackets will be the measurement factor:
Y0
=
+ S(x)
X
The blank signals may be described by three different terms,
depending upon their origin as Currie stated [1]: (i) the
instrumental background. The null signal (which for certain
instruments may be set to zero, on the average) obtained in the
absence of any analyte- or interference-derived signal; (ii) the
(spectrum or chromatogram) baseline. It comprises the summation of the instrumental background plus signals in the analyte
(peak) region of interest due to interfering species; and (iii) the
analyte blank. This signal arises from contamination from the
reagents, sampling procedure, or sample preparation steps which
correspond to the very analyte being sought. In our opinion, the
blank signals are produced by: (i) the instrumental background;
(ii) the signal in the analyte peak region due to other accompanying species which can be exogenous, e.g. from the reagents
(the so-called reagent blank), or endogenous, e.g. from interferents or matrix sample (the so-called matrix blank); and (iii) the
signal due to the presence of the very analyte in the reagents.
The combination of instrumental background and reagent blank
constitutes the method blank.
The blank signal origins a (non-analytical) net signal (it
can be considered constant for each sample in each analytical system) which is independent of those signals attributed to
matrixanalyte interactions, closely related with the existence
of a matrix effect. Ideally, when a chromatographic method is
applied, the blank signal should be zero, since the analyte must
pass through the measurement system separated from the other
components of the sample and the instrumental background also
may be set to zero.
The assessment of the blank (and its variability) is mandatory
for accurate low-level measurements, but require expert chem-

ical knowledge concerning the analytical system in question,


bearing in mind that, in a lot of cases it is very difficult to obtain
an analytical blank for the overall CMP (i.e. a sample that is identical to those being taken for analysis but containing none/not
detectable levels of the analyte(s) sought). When the absence
of a matrix effect is verified, this blank can be substituted by a
reagent or procedural blank [12] carrying out a complete analysis using the solvents and reagents only, in the absence of any
sample.
Sensitivity, has been defined as the quotient of the change in
the indication and the corresponding change in the value of the
quantity being measured [2]. It denotes the quantity of analytical
signal yielded by each unit of analyte amount in the measured
material and it can be constant or dependent of the very analyte
amount. It would be desirable that sensitivity could achieve the
highest value possible in every CMP, in order to obtain the best
performance characteristics for the analytical method implied.
From a calibration experiment appropriately designed in
which the analytical signal is measured from the measurement
standards, the measurement function is explicitly stated as a
representative calibration function, which can be written in a
general way as:
y = f (x)
Usually it is obtained fitting the different (x, y) pairs, by applying
a proper regression algorithm. The calibration features, blank
signal and sensitivity, are estimated by the independent term
and the gradient, respectively, and they are identified with the
intercept and the slope when the measurement function is plotted
as a linear calibration curve.
For any standard analyte amount, the measurement factor
could be estimated by mean of a response factor (RF) from
the following equation:
y
RF =
x
To carry out the quantification of an analyte in a sample, the
corresponding output analytical magnitude (y-variable) is measured and a numerical value of the input analytical magnitude
(x-variable) is calculated from the inverse calibration function,
denoted as evaluation function [1,7] or analytical function,
according to the following expression:
xspl = f 1 (yspl )
where f1 is the inverse calibration function and the subscript
spl indicates the sample. In a similar way, each analyte amount
in the sample could be characterized by means of a calibration
factor (CF), defined as:
CF =

xspl
1
=
RF
yspl

Thus, this feature, could be used to estimate the actual analyte


amount in a sample according to:
xspl = CF yspl
Table 1 shows the most frequent possibilities of calibration
curves as well as the RFs and CFs associated to them, being the

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

37

Table 1
More frequent calibration curves and different features for the analytical separation sciences
Calibration curve

Calibration function

Signal blank

Sensitivity

Response factor

Linear

y = bx
y = a + bx

0
a

b [constant]
b [constant]

b [constant]
a/x + b [depends on x]

Parabolic

y = cx2
y = a + cx2
y = a + bx + cx2

0
a
a

2cx [depends on x]
2cx [depends on x]
b + 2cx [depends on x]

cx [depends on x]
a/x + cx [depends on x]
a/x + b + cx [depends on x]

linear and parabolic calibration functions the most frequently


used in analytical separation sciences.
As it can be seen, in the linear functions sensitivity is constant
and therefore independent of the amount of analyte. Nevertheless, the response factor is only constant when the blank
signal is null. In these circumstances, sensitivity and response
factor are equivalent and the calibration factor is the inverse
of each one of them. Thus, the verification of these two conditions (linear response and null blank signal) is crucial in
order to avoid errors associated with the selection of certain
quantification strategies based in the use of a only measurement standard for calibration purpose, as will be pointed out in
Section 4.
3. Schemes for analytical calibration/quantication
The multi-standard calibration [13] is the usual scheme for
the analytical calibration and the performance characteristic
establishment of a CMP. It is based on the measurement of a
calibration standards set, including the blank if this is manifest,
evenly spaced over the analytical method working range and
preparing replicates of each one in an independent way. Then,
applying an adequate regression algorithm (not necessarily the
Gauss least-square fitting method) to the data, a calibration
function is obtained. These days, new regression methods e.g.
robust regression, are been taken into consideration. A regression involves two steps: (i) selecting a mathematical model
(measurement function) and fitting the corresponding mathematical function from the experimental data; and (ii) validation
of the model and checking of certain initial hypotheses. The
consistency of the regression results always depends on the
grade of the experimental random errors. Therefore, if these
ones are important, the conclusions of the applied statistical
tests are not reliable; for calibration purpose, this fact has special relevance in relation to linearity or goodness-of-fit tests. It
must be taken into account that regression is only a statistical
tool useful to obtain/estimate calibration functions, being other
strategies also possible. To deal with the statistical approach
of the calibration, out of the aims of this paper, there are
available numerous guidelines [7,1416] and technical papers
[1721].
In order to assure the accuracy of the results in samples with
low analyte content, i.e. to avoid a leverage effect on the calibration curve, it is advisable to use a homogeneous arrangement
of the calibration standards. Thus, for example, if in a 5100
analyte amount arbitrary units, the levels chosen are 5, 25 and
100, the highest standard will have a considerable leverage effect

on the calibration curve, being the estimation of the calibration


function strongly affected.
For multi-analyte analytical methods, a very common situation when continuous separation techniques are applied, the
calibration standards are a set of mixtures containing the different analyte standards to be quantified in the adequate amounts.
The multi-standard is feasible when the number of analytes
allows it and analyte measurement standards are available. The
measurement analyte standards used as calibrants (a measurement standard specically used in calibrating, also so-called
calibrators [2]) in analytical sciences, are reference materials,
that can be essentially classified in two types [22]:
(1) Substance reference materials (substance RM), prepared
from single analytes. They include: (a) pure substances characterized for chemical purity and/or trace purities, and (b)
standard solutions, prepared from pure substances (singles
or mixed) in the working solvents used.
(2) Matrix reference materials (matrix RM) characterized for
the composition of specified major, minor and trace chemical constituents. They may be prepared from the matrices to
be analysed or by mean of synthetic mixtures. Thus, this RM
contains the analytes of interest plus the principal chemical
compounds characterizing the matrix to be matched.
In relation to the number of calibration standards which must
be used, and the number of replicates at each calibration level,
different recommendations in recognized written standards and
guidelines can be found. So, IUPAC advises for method validation purposes, the use of six or more calibration standards
that should be run, preferably triplicate on more, in a randomized way [23]. In the same way, the ISO 8466 states, for an
initial assessment of the calibration performance, to employ at
least five calibration standards, although it recommends 10, and
10 replicates of the lowest and highest standards [14]. Another
schemes with a non-balanced uniform-level design for calibration purposes have been recommended [24].
These severe requirements are tempered when the analytical method is running in a routine way, being enough with
three or four calibration standards, duplicated if possible; this
short-working calibration for routine analysis is termed as continuing calibration [13]. If the calibration curve has been well
established during method validation and a suitable calibration
control scheme is put into practice, it may be even possible to
use two alternative schemes: (i) two-standard calibration and (ii)
one-standard calibration.

38

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

3.1. Two-standard calibration


When the linearity of the calibration function has been
ensured in the validation step, a routine calibration can be
obtained from only two calibration standards, preferably measured in replicate. The standards must be selected in such way
that the sample analyte/s quantity be included in the covered
range by them. That is why the quantification can be done interpolating between both standards values, e.g. by applying the
next equation:
xspl =

x2 (yspl y1 ) x1 (yspl y2 )
y2 y 1

where x1 and x2 are each one of the two standards analyte


amounts, y1 and y2 the arithmetic means of the corresponding
analytical signals, and y spl is the arithmetic mean of the analyte
analytical signal in the sample. Also, quantification can be done
using the expression of the straight line through two points (both
calibration standards) as follows:
xspl =

y spl y 1,2
+ x 1,2 ,
b

where b =

y 2 y 1
x2 x 1

being y 1,2 and x 1,2 the global means of the two measured
analytical signals and analyte amounts of the two standards,
respectively, and b is the slope.
When it is known that the intercept is zero, the calibration
factors of both standards are the same and very similar to the
inverse of the slope. Then, the quantification could be easily carried out calculating a pooled calibration factor, CF1,2 , according
to the expressions:
xspl = CFpool y spl ,

where CFpool = CF =

x 1,2
1
=
y 1,2
b

The two-standard calibration can be advisable, when the


expected range of analyte amount in the samples to be analysed
is narrow, but it could be a very laborious calibration scheme
when the samples is unknown. For pesticide residue analysis in
particular, the use of two-standard calibration has been stated
to be acceptable providing the two calibrant x-value differs in a
factor not greater than 4, and where the mean response factors,
derived from replicate determinations at each standard, does not
vary in more than a 20%. This is indicative of an acceptable
linearity of the observable analytical signals [12].
A particular type of two-standard calibration is the one
labelled bracketing calibration [15,16]. The basis of this strategy is to decrease, as much as possible, the interval for which the
calibration function is linear, delimiting it with two calibrants
which bracket the analyte amount values of the test samples
closely; both calibrants and samples, should be measured at least
twice. Assuming this fact, bracketing calibration can be applied
when there is some doubt about the linearity of the calibration
function in the range where the samples are measured and it is
specifically useful when the quantification of analyte amount in
samples is required with a high accuracy degree (in terms of
trueness and precision), e.g., in doping control analysis.
It can be also convenient when the measurement system
shows a significant drift in its detector response, being necessary

a low test samples number with the aim to minimize the total
analysis time. Whether the number of test samples is great, a
batch processing (of them) can be done.
3.2. One-standard calibration
The calibration could be performed from only one calibrant
when two conditions are fulfilled: (i) the calibration function
must be linear in the interval of analyte amount ranged from
the standard value to zero, and (ii) the blank signal must be
null in the interval previously referred to. So, it is mandatory
to check this last requirement before this calibration scheme
was applied. During the experiment, the calibrant must be replicated twice at least. The quantification is carried out using the
calibration factor calculated from the pair of values, standard
analyte amount/average analytical signal, concerning the calibrant applying the expressions:
x1
xspl = CF y spl , where CF =
y 1
The linearity requirement can be avoided when the analyte
amount in the selected calibration standard is very near to the
one in the sample which is possible when the features of the
sample are previously known. Then the one-standard calibration can be easy going applied (for example, for quality control
of manufactured products).
As for the preceding case, for pesticide residue analysis, it has
been stipulated that one-standard calibration may provide more
accurate results than multi-standard calibration if the detector
response is variable with time [12]. In this case, it must be taking into account that, unless further extrapolation is supported
by evidence of acceptable linearity of the calibration function,
the sample analytical signal should be within 10% of the calibration standard analytical signal if the maximum residue limit
is exceeded; if it is not exceeded, the sample response should be
within 50% of the calibration response.
4. Methodologies for analytical
calibration/quantication
In order to the calibration function, previously established
from standards, could be applied to the samples (Fig. 1), different calibration methodologies [13] can be used, bearing in mind
factors like: (i) the availability of representative standards; (ii)
the analytes nature; (iii) the lack of precision or drift of the measurement system; and (iv) the possibility of disturbance caused
in the output analytical signal by other components concomitants the analyte, considered them individually (interferent) or
in a whole way (matrix effect). Obviously any of the previously
mentioned schemes: multi-, two- or one-standard calibrations
can be applied in all the methodologies.
The calibration standards are prepared from RMs containing
the analyte or a surrogate, that is, a pure substance (compound or element) similar to analyte of interest in chemical
composition, separation and measuring which is taken to be
representative of the native analyte; this must be absent or in
a negligible initial concentration in the sample. The calibration

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

standards can be measured in a preparation apart from the sample


(external methodology) or in one including the sample (internal
methodology). Usually, a surrogate is used in an internal methodology and, in this case, it is termed as internal standard (IS).
The IS is chemically distinct from the analytes and therefore
will not have identical chemical properties. However, it is usually selected in such way it was closely related with the analytes
and can represent their analytical behaviour to the highest degree
practicable. The most favourable way to apply IS, is to use an
isotopically labelled analyte, so the chemical properties of the
internal standard are virtually identical to the very analyte. Both,
the very analyte and the labelled one, can be measured separately
by mass spectrometry (this principle is the basis of the isotope
dilution mass spectrometry (IDMS) technique).
In calibration, the IS can be applied with two aims: (i) to
estimate a calibration feature as response factor or calibration
function (it will be the subject of a detailed comment and it will
be pointed out further on Section 4.4); and (ii) to compensate
uncontrolled analytical signal variations in the measuring system. This strategy, in our opinion is not properly a methodology
for calibration, and should be called signal-ratio calibration
although the general expression to refer to it is internal standard
method.
The main characteristic of the signal-ratio calibration is based
on the use of a relative signal (signal-ratio), defined as the quotient between the signal attributed to the analyte (analyte signal)
and to the internal standard:
yR =

ya
yIS

where the subscripts a and IS are related to the analyte and


the internal standard.
It has been widely used to correct mechanical losses of
analytes in procedures such as manual injection in gas chromatography. Nowadays, the use of internal standards has been
Table 2
Types and main distinctive features of different calibration methodologies

39

extended to capillary electrophoresis, in order to maintain


the injection required precision during long routine injection
sequences and/or analysis of samples in complex matrices. In
fact, many imperceptible factors affect the volume introduced
into the capillary during a pressure injection in CE. These factors, mainly related to variability in the pressure/timing of the
injection and to changes in sample solution viscosity, can be
efficiently eliminated by means of the signal-ratio calibration
[25].
The signal-ratio calibration can also be used to correct a moderate proportional (no-additive) matrix effect (see Section 4.2),
since the premise is that, a matrix effect changes the signals from
analyte and the internal standard in the same degree, so that calibration on the ratio signals against analyte amount, will remain
invariant to changes in matrix (different composition between
the calibration standards and the sample analysed).
Ideally, the standard should be added exactly to the test
dissolution before the injection, in enough amount so that its analytical signal be equal or higher than the highest analyte signal
expected; thus, the signal-ratios will always be lower than 1, minimizing the imprecision of these signals and ensuring normally
distributed measurement errors [26].
The distinctive features of the different methodologies,
explained in next subsections, are summarized in Table 2, in
which the minimum number of preparations for a metrological
calibration/quantification are included.
4.1. External calibration
The external calibration (EC) is the most commonly
employed calibration methodology and it is named so, because
the calibration standards do not make up of the sample test
portion but they are prepared and analysed separately from the
samples. In this sense, standards and samples form part of different analytical preparations which are measured in a sequential

40

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

way. In bibliography, EC has been also termed as solvent calibration, standard calibration or even normal calibration.
As external standard (ES), a solution of a RM substance in the
working solvent is used. It usually contains the analyte, although
it is not essential, and the calibration could be carried out from a
surrogate. The use of a surrogate in external calibration is only
appropriate when there are no a RM containing the analyte. An
application, in which two surrogates are used as ES, can be found
in a recent paper on the determination of polyphenols in olive
oil by CE-UV [27].
When a multi-standard calibration is performed, the linear
measurement function to estimate from data regression is:
Y = Y0 + S X
which give rise to a calibration curve:
ya,std = aEC + bEC xa,std
where the subscript std are related to calibration standard, the
intercept aEC is the method blank (Y0 ) and the slope bEC is the
sensitivity of the measurement system (S). The quantification
can be then performed from:
xa,spl =

ya,spl aEC
bEC

where the subscript spl are now related to sample. In certain cases, a no-linear calibration function, as polynomial,
exponential, sigmoidal, etc., could be applied.When a signalratio external calibration is performed, the calibration functions
should be established from signal-ratios; e.g. for a linear function, the equation to be fitted is:
ya,std
R


= ystd
= aEC
+ bEC
xa,std
yIS,std
and the quantification equation is:
xa,spl =


(ya,spl /yIS,spl ) aEC

bEC

To do this, it is mandatory that the sample test portion and all


the calibration standards were prepared with the same internal
standard amount.
A representative example of this procedure would be
the1613 EPA method for the determination of tetra- through
octa-chlorinated dioxins and furans by isotope dilution
high-resolution capillary column gas chromatography/highresolution mass spectrometry [28]. In it, a five-standard external
calibration is used in order to estimate the linear signal-ratio
calibration function used for quantification purposes. A 13 C
isotopically labelled dioxine is used as internal standard.
In routine analytical determinations, of the use of response
and/or calibration factors obtained from one-standard calibration is also quite applied, being it advised too in some official
analytical method, for example, the Harmonised Methods of the
International Honey Commission for determination of sugars in
honey by HPLC or GC [29]. In this case, the equations to apply
are:
xa,std
1
RF =
;
CF =
ya,std
RF

xa,spl = CF ya,spl = xa,std

ya,spl
ya,std

When the signal-ratio calibration is applied, it can be established


a relative response factor (RRF) or a relative calibration factor
(RCF) which are calculated from the ratio signals. The equations
used are:
RRF =

RF
ya,std /xa,std
ya,std /yIS,std
xIS,std R
=
=
=
ystd ;
RFIS
yIS,std /xIS,std
xa,std /xIS,std
xa,std

RCF =

1
RRF

xa,spl = xIS,spl RCF

ya,spl
R
= xIS,spl RCF yspl
yIS,spl

It is habitual, but not essential, to add the same amount of internal standard to all the analytical preparations (calibrants and
samples). Then, due to xIS,std = xIS,spl , both terms can be eliminated in the previous equation and a pseudo-relative response
factor (RRF ) or pseudo-relative calibration factor (RCF ) can
be calculated according to:
RRF =

yR
ya,std /yIS,std
= std ;
xa,std
xa,std

xa,spl = RCF

RCF =

1
RRF

R
yspl
ya,spl
R
= RCF yspl
= xa,std R
yIS,spl
ystd

EC is suitable for the analytical methods that could be considered


as matrix-effect free methods (the so-called Type II methods
in ISO Guide 32 [30]). But it have a main limitation that derives
from the assumption that, the differences between the matrices
of the standards (usually a working solvent) and the sample have
no effect on the calibration function, bearing in mind that if these
differences are ignored, additive and/or proportional systematic
matrix errors may be introduced. In general, the EC methodology is recommended if the analyst has a good knowledge of the
experimental conditions and the contribution of interference to
the measurement is irrelevant or can be kept constant; nevertheless, in the last case, a proper correction of the matrix should be
applied [31], for example, by performing a correction function
[32].
4.2. Matrix-matched calibration
The matrix of calibrants and sample test portion never are
exactly equal, although the sample has been previously subjected
to a clean-up procedure. Sometimes, this fact is irrelevant and
any errors in the determination are produced, but in other cases,
a systematic error appears due to the matrix (matrix systematic
error) not being possible then to apply external calibration.
In separation techniques, this matrix effect is due to the influence on the measurement of the analyte amount, of one or
more undetected sample components and derive from various
physical and chemical processes. It tends to be variable and
unpredictable in occurrence, being in some cases, difficult or
impossible to eliminate. The presence or absence of such effect,

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

may be demonstrated by comparing the response produced from


the analyte in a working solvent solution with that obtained from
the same amount of analyte in the presence of the sample or sample extract. Usually, the matrix produces a change in the signal
which can be: (i) constant and independent of the analyte amount
(additive or translational effect); (ii) variable and proportional
to that amount (proportional or rotational effect); or (iii) the
combination of both type of effects [31,33].
Whether matrix problems are suspected (Type III methods in
ISO Guide 32 [30]), more reliable calibration may be obtained
using as calibration standard, a matrix RM (when available)
or a pure substance RM in conjunction with free-analyte sample matrix prepared freshly (most usual). This is the so-called
matrix-matched calibration (MC) and may make up for matrix
effects although it does not eliminate the underlying cause. That
is why, the intensity of effect may differ from one matrix or
sample to another, and can be also affected by the concentration of matrix. In fact, a MC is a particular type of EC in
which the calibration standards are prepared in a simulated sample that initially it does not contain analyte. Nevertheless, the
MC will not overcome chromatographic interferences caused by
overlapping/unresolved peaks from co-extracted compounds. In
addition, when matrix effect is sample dependent it is necessary
to use standard addition calibration (see next section).
Metrologically, the matrix effect origins a modification in
the measurement function, due to the appearance of new terms.
Supposing a linear behaviour, it can be represented as:
Y = Y0 + Ym + Yanal = Y0 + Ym + p S X
where Ym indicates the matrix sample contribution to the blank,
i.e. the matrix blank (this term characterizes the additive matrix
effect and can have positive or negative values) and p characterizes the proportional one (It can reach values upper and down 1)
being it a feature defined in a similar way to the so-called selectivity index in IUPAC nomenclature [23]. It can be expressed
as:
p=

Sm
;
S

and so : sm = p S

where Sm is the analyte measurement system sensitivity when


the matrix sample is present.
The presence of a significant matrix effect brings about a MCfunction different to that corresponding to EC-function which
for a linear function, can be stated as:
ya,std = aMC + bMC xa,std
where the intercept and the slope can be expressed as:
aMC = aEC + ym = y0 + ym , and bMC = pS = pbEC . It can be
deduced that, when ym = 0 and p = 1, there is not a matrix effect
and both calibration functions are equivalents. A particular case
in which ym > 0 and p > 1 is shown in Fig. 2, as an example.
In the laboratory, the sample must be prepared in such way
the sample amount in the test portion was similar to the matrix
amount used to prepare the calibration standards.
xa,spl =

ya,spl aMC
bMC

41

Fig. 2. A particular example of the relationship between different types of calibration curve plots: external calibration (EC), matrix-matched calibration (MC),
standard addition calibration (AC) and empirical matrix-matched calibration or
matrix-corrected calibration (MCC ). y0 , method blank signal; ym , matrix blank
signal; ya,spl , signal from actual analyte in sample; xa,spl , actual analyte amount
in sample; aEC , intercept from EC-curve; aMC , intercept from MC-curve; aAC ,
intercept from AC-curve; aYC , intercept from Youden calibration curve (or YCcurve); bEC , slope from EC-curve; bMC , slope from MC-curve; bAC , slope from
AC-curve.

If is necessary, the MC-function also could be performed from


signal-ratio calibration by using a suitable internal standard,
being the equations to apply:
ya,std
R


= ystd
= aMC
+ bMC
xa,std
yIS,std
xa,spl =


(ya,spl /yIS,spl ) aMC

bMC

This methodology is particularly recommended in procedures for pesticide or drug residue analysis and other
contaminants in food and biological matrices [34,35]. Contrary
to the common belief, the reliability of quantification using
modern analytical techniques as GCMS or even LCMS (or
tandem MS/MS) might not be adequate. Results can be adversely
affected by the lack of selectivity caused by matrix sample effect
on ion suppression.
Owing to the variety of matrices and samples in pesticide
residue analysis, the MC-functions are prepared only in a series
of representative matrices, this is, a sample material or an extract
of a single food or feed used to represent a commodity group as
an indicator of matrix effect in the analysis of broadly similar
commodities. Matrix representativeness is usually determined
among similar biological material (or tissues) according to its

42

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

content in constituents like, water, acids, sugars, lipids, secondary plant metabolites, etc. The utility of these representative
matrices, has been demonstrated preparing MC calibrations in
cucumber extracts for the determination of 2030 pesticides by
GCMS [36] or LCMS [37] in several vegetable samples.

considered a true calibration, since no measurement standards


are used, but just increasing sample amounts. The YC-curve is
obtained from a plot of the measured signal against the sample
(not analyte) amount (xspl ):

4.3. Standard addition calibration

and the corresponding intercept (aYC ) is the total sample blank


(so-called total Youden blank) [33,39]. Consequently the YCintercept, in agreement with the MC-intercept (see Fig. 2a and
b), is a measure of the total sample blank:

The establishment of a matrix-matched calibration needs a


matrix free from analyte which it is only possible when the
analytes are exogenous sample components (e.g. additives or
contaminants or additives). Nevertheless, if the analytes are
endogenous (e.g. aminoacids in biological fluids or pigments
in vegetables), and a matrix effect has been verified, it is necessary to apply the method of standard addition, properly named
standard addition calibration (AC). There are different ways in
which the AC can be applied which have been recently reviewed
[38].
It is a particular internal calibration methodology, usually based on the analysis of several test portions made,
adding increased amounts of analyte standard to several sample aliquots. From these data, an AC-function can be obtained,
which will have the same slope as the MC-function but higher
intercept, assuming that it can be considered as a MC in which
the matrix contains analyte. Fig. 2a shows a linear AC-curve
together with the corresponding linear MC- and EC-curves.
In this case, the measurement linear function is expressed by
the equation:
Y = Y0 + Ym + Sm (Xnative + Xadded )
= Y0 + Ym + Sm Xnative + Sm Xadded
where Xnative indicates the analyte amount in each test portion
which is related with the analyte originally present in the sample and Xadded represents the analyte amount added to each test
portion. The AC-curve fitted (measured analytical signal versus
added analyte amount) is given by:
ya,(spl+added) = aAC + bAC xa,added
whose intercept (see Fig. 2a and b) is:
aAC = y0 + ym + Sm xa,spl
where Sm is the AC-slope (bAC ), y0 is the method blank, ym is
the matrix blank, ya,spl and ya,(spl + add) are the analyte signals in
the original and added samples, respectively, and xa,add is the
added analyte amount.
From this expression is possible to quantify the analyte
amount initially present in the sample (xa,spl ):
xa,spl =

aAC (y0 + ym )
aAC (y0 + ym )
=
Sm
bAC

and as it can be seen, it is necessary to correct the intercept


value (aAC ) with the total sample blank (method blank + matrix
blank).
On the other hand, in order to obtain a bias free result by
means AC, it is necessary to carry out a Youden calibration
(YC) to estimate the total blank of the sample. It cannot be

ya,spl = aYC + bYC xspl

aYC = aMC = aEC + ym = y0 + ym


and the quantification equation is:
xa,spl =

aAC aYC
Sm

When the total sample blank is null, the former equation can be
simplified to:
aAC
xa,spl =
bAC
which gives the extrapolated value of the x-intercept. Only in
this case, the analyte amount can be calculated as the quotient
between the intercept and the slope of the AC-curve. This fact
is the basis of the graphic quantification method that generally
appears in text books. Nevertheless, if the total sample blank
nullity is not fulfilled, this last approach can produce serious
errors in the quantification.
When a one-standard addition calibration is made, the
quantification step requires the analysis of at least two test portions, one of them containing an amount of analyte standard
added. Then the sample analyte amount is obtained applying
the next equation which is a particular case of the general one
above explained:
RF =

ya,(spl+add) ya,spl
;
xa,add

xa,spl = CF ya,spl = xa,add

CF =

1
RF

ya,spl
ya,(spl+add) ya,spl

being ya,spl and ya,(spl + add) the analyte signals in the original and
the added samples, respectively; and xa,add is the added standard
analyte amount.
As in the previous methodologies, the AC-calibration could
be performed from signals-ratio. The equations for calibration
and quantification are equivalent to the previous ones without
more than to substitute the analytical signals for the respective
analytical signals-ratio.
This type of calibration is the only methodology in which
the results are not affected by systematic matrix errors, independently of the kind and magnitude of the matrix effect, so it
permits to quantify the amount of analytes in any kind of samples. That is why a in-house validation of analytical methods
based on the standard additions and Youden calibrations has
been proposed [40]. Its principal disadvantage lies in the fact that
a calibration for each sample is needed, which implies a lot of
work in routine analysis. To avoid this problem, a simulation of a

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

MC-curve by means of an empirical calibration curve obtained


from the YC- and AC-curves features, has been recently proposed. This new calibration function, named matrix-corrected
calibration (MCC ) [41], is a hypothetical matrix-matched calibration characterized by a linear curve where the x-variable is
the analyte amount (in the presence of matrix) and the y-variable
is the analytical signal (Fig. 2b).
To establish the MCC , is necessary to set up only two calibrations: AC and YC, which must be representative. From the
intercept (aAC ) and the slope (bAC ) of the AC and the intercept
(aYC ) of the YC, both MCC -intercept and MCC -slope can be
calculated by the equations:
aMCC = aAC ya,spl = aYC ;

bMCC = bAC

which are exclusive of each type of matrix, since each matrix


has its own AC-curve because the AC-slope, which corrects the
proportional error introduced by the matrix, may change when
different matrix amounts are used and consequently bMCC =
bAC and they are only valid provided that the amount of sample
used for analysis remains constant.
Thus, when a matrix effect exists, the analyte amount from
each measured analytical signal in any sample, can be calculated
from:
ya,spl aMCC
ya,spl aYC
=
xa,spl =
bMCC
bAC
4.4. Internal calibration
There are some analytical situations in which an internal standard, employed as calibrant, is added to the sample and analysed
as a whole together with the analyte. From the same analytical
preparation, a response/calibration factor is obtained from the
IS which is applied on the analyte signal for quantification using
the following equations:
RFIS =

yIS,spl
;
xIS,spl

CFIS =

1
RFIS

xa,spl = CFIS ya,spl = xIS,spl

ya,spl
R
= xIS,spl yspl
yIS,spl

Metrologically, this is the real internal calibration (IC), a onestandard internal calibration methodology formally different to
the signal-ratio calibration; however this last is the most known
application of the use of internal standard in analytical sciences
which sometimes has derived in an inappropriate use of the term
internal calibration.
The preceding equations could be combined and the quantification would be obtained from the following simplified one:
xa,spl =

xIS,spl
R
ya,spl = xIS,spl yspl
yIS,spl

With the aim to enhance the accuracy of the results, a modification of the methodology that uses two internal standards, has
been proposed and it has been called double standard internal

43

method. Initially, the quantification involved the average geometrical value calculation of the two results obtained with each
standard:

xa,spl = ya,spl CFIS1 CFIS2
This equation can be significantly simplified when: (i) two
closely related compounds, upper and lower, of the same
homologous series to which the analyte belongs, are used
as internal standards (e.g. methanol and n-propanol could be
used for ethanol determination); (ii) the analyte and internal standards calibration factors are approximately equals
(CFIS,1 CFa CFIS,2 ), being the product of the relative calibration factors analyte/internal standards around 1 (CF1 CF2 = 1)
due to its differences are compensated each other [42]; and
(iii) both internal standards are added in the same amount,
xIS,1 = xIS,2 = xIS . Then, the analyte amount in the sample can
be calculated according to:


1
R
R
=xIS,spl y1,spl
y2,spl
xa,spl =ya,spl xIS,spl
yIS1,spl yIS2,spl
R and yR represent the analyte/internal standard
where ya,1
a,2
signal-ratios 1 and 2, respectively, obtained from the sample
analysis.
The main advantage argued by the authors, arises from that
this quantification method does not need a previous calibration
(the calibration is implicit in the quantification), so its practical application is very easy. It only requires the addition of
a known and equal amount of both internal standards to each
sample. Once the sample treatment has been finished, the analyte and internal standards signals are measured. The results
obtained show a great accuracy degree, even whether during
sample treatment any analyte losing is produced. Its more important disadvantage lies in the necessity of having two analyte
homologous (internal standards), initially absents in the sample.
To apply an IC the measurement system has to be able to distinguish between the signals from the internal standard(s) and
the very analyte, that is why, both substances must have a different behaviour in the measurement system but similar analytical
properties in order to be measured in a quasi-simultaneous way.
In the same portion of the test sample, the signal of the standard
and the very analyte are measured; from the first one a calibration factor is obtained, which is used to quantify the amount of
analyte from the second signal.
An exclusive feature of IC is that it is possible to carry out the
quantification of several analytes (generally from the same family) in the same sample portion, starting from an only internal
calibration, this is, the internal standard can represent to several
analytes in a simultaneous way in the same sample. As consequence, it would be possible to calculate the (percentage) mass
fraction of each analyte using only the values of the measured
analytical signals by:

ya,spl
xa,spl
FCIS ya,spl
ca,spl (%) =  100 = 
100 =  100
xi
(FCIS yi )
yi
where ca,spl indicates the percentage mass fraction of the analyte
a; xa,spl is the mass of this analyte in the sample and ya,spl is

44

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346


the analytical signal due to it. xi,spl represents the sum of the
masses of all the analytes and, yi,spl represents the sum of the
analytical signals attributed to these ones. This strategy is only
possible when the calibration factors are similar for all the analytes, and equal to the calibration factor of the internal standard.
The reliability of internal calibration as calibration/
quantification methodology is based on the ability of the IS to
act as surrogate. In metrological terms, this fact implies that the
response factor (or sensitivity) of the IS is constant and equal
to the calibration factors of each one of the substituted analytes,
into the interval of application of the method. That can be true
for those universal no-specific detectors which respond directly
to the substance amount (mass or concentration), as it happens
with the most classical GC detectors. However, this condition
is harder to fulfil in LC, due to the majority of common detectors used, produce a dissimilar signal amount depending on the
judged analyte. Even, when MS detectors are used, and depending on the ionization system, it could be consider this premise
is satisfied when they work in full-scan mode (not in SIM
mode). In any case, and as far as possible, it is necessary to verify
that this requirement is achieved using a measurement standard
containing both analyte and internal standard.
The main advantage of internal calibration is the reduction
of analysis time, since it only needs one analytical preparation
for calibration and quantification. Moreover, this methodology
permits to make up for the losing of analyte during sample preparation and, in a moderate way, the matrix effect. That is why it is
advisable when at least one of the following circumstances are
produced: (i) the previous preparation sample process is long
and complicated; (ii) a long time for the measure is required
(e.g. chromatogram run time); and (iii) there is no or it is impossible to acquire a material with analyte standard. Obviously, the
main limitation of this methodology is the availability of an adequate IS. As an additional precautionary measure, the amount of
IS must be such its analytical signal (peak area or peak height)
be of the same order that those corresponding to the analyte;
otherwise, significant errors in quantification can be found, over
all if the internal standard signal is not entirely linear.
UE olive oil official analytical methods for the determination of the composition and content of waxes [43], sterols [44],
stigmastadienes [45], and aliphatic alcohols [46], by capillarycolumn gas chromatography, constitute examples of common
application of IC. In those methods, an only calibration IS is
used for the quantification of some analytes of the same chemical
family.
An imaginative proposal recently reported is the echo-peak
internal standard calibration [47]. This is a novel calibration methodology which simulates the use of internal standard
by injecting consecutively, within a short period of time, an
unknown sample and a standard analyte solution. As a result,
the standard analyte peak elutes closely to the sample analyte
peak, thus performing a echo peak. For quantification purposes, a calibration function is obtained from the signal-ratios
of the standard and the reference standard. The concentration of
analyte in the unknown sample is calculated from the peak area
ratio of sample and reference. As an internal standard method,
the echo-peak technique provide the possibility to compensate

the signal decreasing during the analytical sequence when the


detector response shows a significant instability with the time.
4.5. Calibration by internal normalization
When the measuring system sensitivity changes from one
analyte to another, the application of the internal standard calibration requires to normalize the measured analytical signals by
determining a normalization factor (NF) for each analyte in
relation to a reference compound (usually, an internal standard).
This calibration methodology is called calibration by internal
normalization or normalized signal calibration.
Despite its denomination, it is properly a methodology of
double calibration externalinternal, applied in two steps: (i) in
the first, a multiple external standard measurement to determine
the calibration factors of all the components (analytes and internal standard) and, the normalization factors of the analytes in
relation with the internal standard, are used; and (ii) in the second, the calibration factor of the internal standard, previously
added to the sample, is determined and the analytes are quantified. Thus, the advantages of both methodologies are joined and
the drawbacks of each one of them when are applied separately,
are minimized.
The NF values of the analytes depend on the detector
employed, so it is necessary to estimate them for each measuring system. For this purpose, in the first step, a multiple standard
calibration, which includes a known amount of each analyte and
internal standard, must be prepared and measured previously to
the sample. Then, the different normalization factors are calculated as the quotient between the calibration factors of the
considered analyte (CFa ) and the internal standard (CFIS ):
NF =

CF
xa,std /ya,std
xa,std yIS,std
xa,std
1
=
=

=
R
CFIS
xIS,std /yIS,std
xIS,std ya,std
xIS,std ystd

Consequently, the NFa,std is the relative calibration factor of each


analyte in relation to a reference compound (internal standard).
It can be underlined that this NF equation is similar to the one
used to calculate the relative calibration factors (RCF) in Section
4.1.
In the second step, the calculated NF values are used to estiN )
mate a normalized signal for each analyte in the sample (ya,spl
from the measured signals (ya,spl ):
N
= NF ya,spl
ya,spl

These normalized signals indicate the hypothetical signals that


would be measured if the calibration factors were similar for
all of them, and equal to the calibration factor of the internal
standard. Finally, the normalized signals can be used for quantification of the analytes in the sample, by applying the next
expression:
N
xa,spl = CFIS ya,spl
= CFIS NF ya,spl

where it can be seen that, the calibration factor of each analyte


(CF), can be unfolded into two factors:
CF = NF CFIS

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

45

To assure the validity of this methodology, it is necessary that


the calibration factor of internal standard keep constant.
As it has been stated in the previous section, is possible to
obtain directly the (percentage) mass fractions of each analyte
in a sample containing several analyte from the same chemical
family. Whether only the analyte mass fraction in the sample is
required, any other of the analytes in the sample can be used
as reference compound to normalize the signals, not being necessary to add an internal standard neither in the measurement
standard nor in the sample.
In this case, the NF values are calculated directly from the
mass fraction (in percentage or non) of each analyte i in the
calibration standard (ca,std ) as:

xi,std / xi,std yref,std
xi,std yref,std


NF =
xref,std yi,std
xref,std / xi,std yi,std
ci,std yref,std
=

cref,std yi,std

nal calibration with an only standard (an one-standard external


calibration) until the most tedious, for routine analysis, standard
addition calibration from several calibration standards prepared
and measured in replicate.
The selection of the appropriate methodology should be
based on a good knowledge of the metrological features of
the CMP and/or from a previous judgment on particular situations such as lack of stability or drift of the measuring system,
existence of the (additive or proportional) matrix effects, availability of proper calibration standard (from the analyte or some
surrogates), etc. The paper also contains a lot of practical
recommendations and sufficient guidance to properly apply a
chromatographical (or from other separation technique) calibration, in order to avoid misconceptions and errors for lack of
knowledge.

where the subscript


ref is referred to the analyte reference,

and the term
xi,std is the sum of the amounts of all analyte,
including the analyte reference. Thus, the composition of one
analyte a in the sample (ca,spl ), expressed as mass fraction (in
percentage), is calculated as:

We acknowledge financial support from the Spanish Ministerio de Educacion y Ciencia (MEC), Direccion General de
Investigacion (Project No. CTQ2006-15066-C02-02).

xa,spl
ca,spl (%) = 
xi,spl

N
CFref ya,spl
100 = 
100
N )
(CFref yi,spl

N
ya,spl
NFa ya,spl
=  N 100 = 
100
(NFi yi,spl )
yi,spl

The calibration by internal normalization is a methodology


which has been scarcely used in analytical quantification. However, it exits a significant application example in the UE official
analytical method for the determination of methyl esters of fatty
acids by gas chromatography in olive oil [48], a representative
example in which both ones, an internal standard and an analyte,
are used as reference compound.
5. Conclusion
An overview on the problem of the analytical calibration/qualification for the separation sciences has been shown up.
The paper intends to establish a description of the metrological
fundamentals with the aim of reconciling the vocabulary and the
subjects of the measurement science with those corresponding
to the chemical analysis, when separation techniques are previously applied. We believe necessary that the chromatographers
familiarize with this terminology that nowadays is considered
as a fundamental topic of the analytical chemistry.
In addition, on the basis of the conventional multi-standard
calibration, the principles, advantages and limitations of alternative calibration schemes as two- and one-standard calibration
have been presented. The different calibration methodologies,
that could be applied, have been explained and their scope have
been discussed. We have tried to show how the analysts can have
a battery of possibilities that embraces from the simplest exter-

Acknowledgement

References
[1] Nomenclature in Evaluation of Analytical Methods including Detection
and Quantification Capabilities, IUPAC Recommendations 1995 (prepared
by L.A. Currie), Pure Appl. Chem. 67 (1995) 16991723; reproduced in
L. A. Currie, Anal. Chim. Acta 391 (1999) 105.
[2] International Vocabulary of MetrologyBasic and General Concepts and
Associated Terms (VIM), third ed., Final Draft, August 2006, http://www.
ncsli.org/vim/wg2 doc N318 VIM 3rd edition 2006-08-01%20(3).pdf.
[3] L. Cuadros Rodrguez, L. Gamiz Gracia, E. Almansa Lopez, J. Laso
Sanchez, Trends Anal. Chem. 20 (2001) 195.
[4] A. Switaj-Zawadka, P. Konieczka, E. Przyk, J. Namiesnik, Anal. Lett. 38
(2005) 353.
[5] International Vocabulary of Basic and General Terms in Metrology (VIM),
second ed., International Organization for Standardization (ISO), Geneva,
1993.
[6] ISO 11843-1, Capability of DetectionPart 1: Terms and Definitions,
International Organization for Standardization (ISO), Geneva, 1997.
[7] Guidelines for Calibration in Analytical ChemistryPart 1: Fundamentals and Single Component Calibration, IUPAC Recommendation 1998
(prepared by K. Danzer and L.A. Currie), Pure Appl. Chem. 70 (1998)
993.
[8] Harmonized Guidelines for Internal Quality Control in Analytical Chemistry Laboratories, IUPAC Technical Report 1995 (prepared by M.
Thompson and R. Wood), Pure Appl. Chem. 67 (1995) 649.
[9] Absolute Methods in Analytical Chemistry, IUPAC Technical Report 1995
(prepared by A. Hulanicki), Pure Appl. Chem. 67 (1995) 1905.
[10] Selectivity in Analytical Chemistry, IUPAC Recommendations 2001 (prepared by J. Vessman and col.), Pure Appl. Chem. 73 (2001) 1381.
[11] K. Danzer, Anal. Bioanal. Chem. 380 (2004) 376.
[12] Quality Control Procedures for Pesticide Residue Analysis, Document
SANCO/10232/2006, European Union, Brussels, 2006.
[13] L. Cuadros Rodrguez, L. Gamiz Gracia, E. Almansa Lopez, J.M. Bosque
Sendra, Trends Anal. Chem. 20 (2001) 620.
[14] ISO 8466, Water QualityCalibration and Evaluation of Analytical Methods and Estimation of Performance Characteristics (Parts 1 and 2),
International Organization for Standardization (ISO), Geneva, 1990 and
1993.
[15] ISO 11095, Linear Calibration using Reference Materials, International
Organization for Standardization (ISO), Geneva, 1992.

46

L. Cuadros-Rodrguez et al. / J. Chromatogr. A 1158 (2007) 3346

[16] LGC/VAM/2003/032, Preparation of Calibration CurvesA Guide to Best


Practice, Laboratory of the Government Chemist (LGC), Cambridge, 2003.
[17] J.N. Miller, Analyst 116 (1991) 3.
[18] D.L. MacTaggar, S.O. Farwell, J. AOAC Int. 75 (1992) 594.
[19] K. Baumann, H. Waetzig, Proc. Cont. Qual. 10 (1997) 59.
[20] K. Baumann, Proc. Cont. Qual. 10 (1997) 75.
[21] R. de Levie, Crit. Rev. Anal. Chem. 30 (2000) 59.
[22] E.A.-04/14, The Selection and Use of Reference Materials, European Cooperation for Accreditation, 2003.
[23] Harmonized Guidelines for Single-Laboratory Validation of Methods of
Analysis, IUPAC Technical Report 2002 (prepared by M. Thompson,
S.L.R. Ellison and R. Wood), Pure Appl. Chem. 74 (2002) 835.
[24] A. Gonzalez Casado, A.M. Garca Campana, L. Cuadros Rodrguez, J.L.
Vilchez Quero, R. Blanc Garca, LC GC Int. 11 (1998) 726.
[25] K. Altria, LC GC Eur. 1 (September) (2002) 1.
[26] L. Cuadros Rodrguez, A. Gonzalez Casado, A.M. Garca Campana, J.L.
Vlchez, Chromatographia 47 (1998) 550.
[27] A.M. Gomez Caravaca, A. Carrasco Pancorbo, B. Canabate Daz, A.
Segura Carretero, A. Fernandez Gutierrez, Electrophoresis 26 (2005)
3538.
[28] Method 1613, US Environmental Protection Agency (EPA), Washington,
DC, 1994, http://www.epa.gov/waterscience/methods/1613.pdf.
[29] Methods 7.2 and 7.3, Harmonised Methods of the International Honey
Commission (IHC), Berne, 2002, http://www.apis.admin.ch/enghish/host/
pdf/honey/ICHmethods e.pdf.
[30] ISO Guide 32, Use of Certified Reference Materials, International Organization for Standardization (ISO), Geneva, 1997.
[31] M. Thompson, S.L.R. Ellison, Accred. Qual. Assur. 10 (2005) 82.
[32] L. Cuadros Rodrguez, A.M. Garca Campana, E.M. Almansa Lopez, F.J.
Egea Gonzalez, M.L. Castro Cano, A. Garrido Frenich, J.L. Martnez Vidal,
Anal. Chim. Acta 478 (2003) 281.

[33] M.J. Cardone, J. AOAC Int. 66 (1983) 1283.


[34] Guidelines, for Single-Laboratory Validation of Analytical Methods
for Trace-Level Concentration of Organic Chemicals, Association of
Official Analytical Chemist (AOAC)/Food and Agriculture Organization of the United Nations (FAO)/International Atomic Energy Agency
(IAEA)/International Union of Pure and Applied Chemistry (IUPAC).
[35] A.R.C. Hill, S.L. Reynolds, Analyst 124 (1999) 953.
[36] J.L. Martnez Vidal, F.J. Arrebola, A. Garrido Frenich, J. Martnez
Fernandez, M. Mateu Sanchez, Chromatographia 59 (2004) 321.
[37] J.L. Martnez Vidal, A. Garrido Frenich, T. Lopez Lopez, I. Martnez Salvador, L. Hajjaj el Hassani, M. Hassan Benajiba, Chromatographia 61
(2005) 127.
[38] P. Koscielniak, J. Kozak, Crit. Rev. Anal. Chem. 36 (2006) 27.
[39] R.C. Castell, M.A. Castillo, Anal. Chim. Acta 423 (2000) 179.
[40] A.M. Garca Campana, L. Cuadros Rodrguez, J. Aybar Munoz, F. Ales
Barrero, J. AOAC Int. 80 (1997) 657.
[41] I. Fernandez Fgares, A. Gonzalez Casado, L. Cuadros Rodrguez, J. Chromatogr. B 779 (2004) 73.
[42] I.G. Zenkevich, E.D. Makarov, J. Chromatogr. A, in press.
[43] Commission Regulation (EC) No. 183/93, Annex IV, Off. J. Eur. Commun.
L22 (1993) 62.
[44] Commission Regulation (EC) No. 2568/91, Annex V, Off. J. Eur. Commun.
L248 (1991) 15.
[45] Commission Regulation (EC) No. 656/95, Annex XVII, Off. J. Eur. Commun. L69 (1995) 05.
[46] Commission Regulation (EC) No. 769/2002, Annex XIX, Off. J. Eur. Commun. L128 (2002) 23.
[47] J. Zrostlikova, J. Hajslova, J. Poustka, P. Begany, J. Chromatogr. A 973
(2002) 13.
[48] Commission Regulation (EC) No. 2568/91, Annex X-A, Off. J. Eur. Commun. L248 (1991) 36.

You might also like