Talanta 82 (2010) 19951998

Contents lists available at ScienceDirect

Talanta
journal homepage: www.elsevier.com/locate/talanta

Short communication

Intra-laboratory assessment of method accuracy (trueness and precision) by

using validation standards
A. Gustavo Gonzlez , M. ngeles Herrador, Agustn G. Asuero
Department of Analytical Chemistry, University of Seville, 41012 Seville, Spain

a r t i c l e

i n f o

Article history:
Received 27 June 2009
Received in revised form 19 July 2010
Accepted 27 July 2010
Available online 4 August 2010

a b s t r a c t
Assessment of accuracy of analytical methods is a fundamental stage in method validation. The use of
validation standards enables the assessment of both trueness and precision of analytical methods at the
same time. Procedures of intra-laboratory testing of method accuracy using validation standards are
outlined and discussed.
2010 Elsevier B.V. All rights reserved.

Keywords:
Accuracy
Trueness
Precision
Validation standards

1. Introduction
The accuracy of an analytical method is a primarily role for validation purposes [16]. In a former article [7] the assessment of
method accuracy from recovery assays based on spiked matrices
and spiked samples was revised. However in the mentioned paper,
accuracy was considered in the perspective of a systematic component of the error, today superseded by the term trueness. In his
modern sense, accuracy is a performance characteristic that refers
to the total error (systematic and random errors) and comprises two
components: Trueness and precision, whose quantitative expressions are the bias and the standard deviation respectively [8,9].
These two gures of merit can be assessed independently, but it is
possible to assess accuracy in a holistic way according to the measurement uncertainty and accuracy proles [1013]. In-house or
intra-laboratory assessment of trueness and precision can be performed at a time when validation standards (VS) are available. VS
have to be prepared in the same matrix as the expected for future
samples. Certied or internal reference materials represent the best
way to obtain VS, but spiked samples can be considered as a suitable alternative [1416]. In the case of pharmaceutical formulations
or other manufactured products where a placebo is available, the
bias or precision study can be carried out using spiked placebos.
But, when the placebo is not available, selected stable samples fortied to a suitable level of analyte may be prepared. VS must be

Corresponding author. Tel.: +34 954557173; fax: +34 954557168.

E-mail address: agonzale@us.es (A.G. Gonzlez).
0039-9140/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.talanta.2010.07.071

stable, homogeneous and as similar as possible to the future samples to be analyzed and they represent, at the validation stage, the
future samples that the analytical procedure will have to quantify.
Each VS have to be prepared and treated independently as a future
sample. This independence is essential for a good estimation of the
between-conditions variance in the assessment of accuracy; other
important performance characteristics are assumed to be consistent with the tness for purpose of the analytical method. Thus, it
will be supposed that the method is suitably selective and sensitive and the possible matrix effects have been previously studied
and corrected at the calibration stage [5,6]. In such a case, a corrected inverse prediction equation will be available to transform
the measured analytical signal coming from the sample into the
analyte concentration. Once these requirements meet, the test of
method accuracy can be carried out. The aim of the present paper is
to outline and discuss the most suitable and practical procedures for
assessing the trueness and precision of an analytical method when
VS are available, either independently or in a global way through
the measurement uncertainty and the accuracy proles.

2. Assessment of trueness and precision from a nested

design and ANOVA calculations
According to ICH guidelines [1], precision may be considered
at three levels: repeatability, intermediate precision and reproducibility. Repeatability expresses the precision evaluated under
the same experimental conditions over a short time interval. Sometimes it is termed as intra-assay or within-run precision and refers
to the pure random error associated with the assay measurement

1996

A.G. Gonzlez et al. / Talanta 82 (2010) 19951998

process. Intermediate precision applies to within-laboratory variations: different days, different analysts, different equipment and so
forth. Intermediate precision is sometimes called between-run or
inter-assay precision. But as Peters and Maurer pointed out [17],
stricto senso, intermediate precision is the total precision under
varied conditions that it is expected within laboratory in a future
assay, whereas so called inter-assay, between-run or between-day
precision only measure the precision components caused by the
respective factors. Nevertheless, if a between-day precision study
is performed by spacing out the measurement days in such a way
that other items (analysts, equipment, stock solutions, glassware,
etc.) really changed, then this precision measurement could be
considered as a truly intermediate precision estimation. Reproducibility, in its turn, expresses the between-laboratories precision
like in collaborative studies. Reproducibility only has to be studied, if a method is supposed to be used in different laboratories.
Unfortunately, some authors use the term reproducibility for
intra-laboratory precision studies at the level of intermediate precision [18]. On the other hand, the trueness of an analytical procedure
expresses the closeness of agreement between the mean value
obtained from a series of measurements and the value which is
accepted either a conventional value or an accepted reference value
like VS. Trueness can be expressed in terms of recovery or absolute or relative bias [11]. Both trueness and intermediate precision
studies can be carried out by predicting the actual concentrations
of a series of VS selected. Following the golden rules of method
validation [19], the analytical procedure should be validated separately for each kind of matrix considered as a whole (including
sample treatments prior to analysis) and covering the full range
of analyte concentrations. Accordingly it is advisable to perform
the accuracy study with VSs at least for three concentration levels
m (low, medium and high) covering the dynamic working range
previously established in the calibration stage, with a number of n
replicates at each concentration. The ICH Q2(R1) document [1] recommends three replicates and the FDA document [2] consider ve
replications. Accordingly, 35 replications are advisable. Calculations of intermediate precision and bias have to be carried out on
results instead of analytical responses. Considering the different p
days as the main source of variation for the intermediate precision
study, a one-way analysis of variance (ANOVA) can be performed
for each VS. Thus, for each concentration level m, the predicted concentration of the VS (by using the suitable calibration curve) will be
xij with two indices: i (from 1 to p) corresponding to the different
days and j (from 1 to n) accounting for the replications. From the
2 ) and betweenANOVA the estimations of within-days variance (SW
2
days variance (SB ) are obtained [5,20,21]. The within-days variance
is also known as repeatability variance (Sr2 ) and is given by

p n
Sr2

2
SW

i=1

j=1

(xij x i )2

p(n 1)

(1)

with

n

x
j=1 ij

x i =

p
=

i=1

(xi x )

p1

with

p n
x =

i=1

pn

Analyte (%)

Analyte fraction

Concentration unit

% RSDH

% RSDAOAC

100
10
1
0.1
0.01
0.001
0.0001
0.00001
0.000001
0.0000001

1
101
102
103
104
105
106
107
108
109

100%
10%
1%
0.1%
100 ppm
10 ppm
1 ppm
100 ppb
10 ppb
1 ppb

2
2.8
4
5.7
8
11.3
16
22.6
32
45.3

1.3
1.8
2.7
3.7
5.3
7.3
11
15
21
30

2 ) can be estimated
The intermediate precision variance (SIP
according to [5]:
2
= Sr2 + SB2
SIP

(3)

From this value, the corresponding relative value, RSDIP is computed and can be compared with the expected values issued from
the Horwitz equation and the Horrat parameter [22,23]. Horwitz
[24] found an expression for predicting the expected value of the
relative standard deviation for inter-laboratory trials according to:
%RSDH = 2(10.5 log C)

(4)

where C is the analyte concentration expressed in decimal fraction.

The RSDH value is a primary criterion for evaluating reproducibility
precision [25]. Intermediate precision predicted RSD is approximately one-half to two-thirds the RSDH Horwitz value [26,27]. The
Horrat value is often used as a benchmark for the performance
of analytical methods, which is dened as the ratio of the actual
relative standard deviation RSD calculated from the analytical data
to the predicted Horwitz value:
Horrat =

RSD
RSDH

(5)

In our case, for evaluating the intermediate precision we take

RSD = RSDIP in Eq. (5). The acceptable Horrat value for intermediate
precision studies should not be higher than 1.3 [26].
Aside from Horwitzs parameters, values of RSD according to the
AOAC Peer veried Methods Program [28] can be also considered.
Both Horwitz and AOAC acceptable RSD values as a function of analyte concentration are presented in Table 1. As a quick rule, our
RSDIP result should be compared with one-half the corresponding
RSD value appearing in Table 1.
The assessment of trueness can be performed according the
same ANOVA results. Accordingly trueness can be expressed as the
bias or the recovery obtained for each VS assayed [29].
In the bias calculation, the total mean value x is taken as the
nal result corresponding to the concentration of the VS, whose
estimated true concentration is T. Accordingly, the corresponding
is calculated by:
total bias ()
= x T

x
j=1 ij

Sr2
n

(6)

Assuming that the true concentration T of the VS has a negligible uncertainty, the variance of the bias can be easily calculated
from the ANOVA results [30]:

The between-days variance is calculated from

SB2

Table 1
Acceptable RSD values obtained from the Horwitz function and from the AOAC
Peer Veried Methods program according to the concentration level of analyte
(ppm = parts per millions, ppb = parts per billions).

(2)

S 2 = S 2 (x ) =

2 ((n 1)/n)S 2
SIP
r

(7)

We can apply the Students t-test for assessing the signicance

of bias:
t =

x T
=
S
S

(8)

A.G. Gonzlez et al. / Talanta 82 (2010) 19951998

Table 2
Acceptable recovery percentages according to the concentration level of analyte.
Analyte (%)

Analyte fraction

Concentration unit

Recovery range (%)

100
10
1
0.1
0.01
0.001
0.0001
0.00001
0.000001
0.0000001

1
101
102
103
104
105
106
107
108
109

100%
10%
1%
0.1%
100 ppm
10 ppm
1 ppm
100 ppb
10 ppb
1 ppb

98102
98102
97103
95105
90107
80110
80110
80110
60115
40120

1997

materials, 5% on drug products or pharmaceutical formulations,

15% for biological samples, and so forth [36]. A valid analytical
method should provide with results X that accomplish Eq. (11) very
likely. This can be formally expressed as:
P(|X T | < )

(13)

where is the probability of having measurements inside the

acceptance limits, e.g. 90%. If the true bias () and the true precision () are known, and assuming a normal distribution, we can
write X T = z , z being the -quantile of the standard normal
variate. Eq. (11) can now be expressed as
 < z  < 

If t is less than the critical tabulated value for np 1 degrees of

freedom at the chosen signicance level, then no statistical differences have been observed between the overall mean and the true
value.
But, as indicated above, the recovery term can also be used
for checking trueness and has a more intuitive meaning. The total
recovery for VS is dened as the ratio between the observed estimation of the VS concentration, x and the true value T, expressed
as percentage or as fraction:
R=

x
T

(9)

From Eqs. (6) and (8) we get

R=1+

(10)

Thus, if the bias is negative (positive), the recovery will be lesser

(higher) than 100. The variance of the recovery is easily obtained
and related to those of bias:
S 2
S 2 (x )

=
T2
T2

But this is a utopia because the true performance parameters

are unknown.
 It is advisable to use their estimations; for the bias

and S =

2 + S 2 for the total precision standard deviation [8].

SIP


S=

2 +
SIP

2 ((n 1)/n)S 2
SIP
r

+ kS > 
kS < 

Once recovery is computed, we can check it for suitability by

comparison with the published acceptable recovery range as a function of the analyte concentration [28] as it is depicted in Table 2.
This later procedure is suitable for trueness and intermediate
precision assessment coming from a nested ANOVA performed
with VS, where trueness and intermediate precision are checked
independently. However, it is also possible to assess the accuracy in
a global way, with the use of the called accuracy proles mentioned
above.

lower limit :

The concept of accuracy prole was rst introduced in the

papers of Hubert et al. [31] and Boulanger et al. [32]. The Societ
Francaise des Ciences et Techniques Pharmaceutiques (SFCTP) have
used accuracy proles to assess the accuracy in method validation
according to the concept of acceptability limit [3335]. Rozet et al.,
in his excellent paper [36] illustrate very suitably the concepts of
trueness, precision, accuracy and accuracy proles in the sense of
SFCTP.
When applying an analytical method, the analyst expects that
the difference between the measurement result X and the unknown
true value T of the tested sample be less than a predened acceptance limit :
 < X T <  |X T | < 

(12)

The acceptance limit is not arbitrary but depends on the goals of

the analytical procedure. It is the outcome of a discussion between
the analyst and the client or end-user [12] or is linked to the requirements usually admitted by the practice, for instance, 12% on bulk

(15)

(16)

where k is the called coverage factor, that can be assimilated to

the -quantile of the standard normal variate, when a Gaussian
distribution is assumed. For each VS concentration level, the upper
and lower tolerance interval limits, expressed as percentage, are
given by
upper limit :

3. Assessment of global accuracy according to accuracy

proles

Thus the -Expectation Tolerance Interval (ETI) can be constructed according to

(11)

SR2 =

(14)

100( + kS)
T
100( kS)
T

(17)

We can construct one interval for each VS. Then, the upper limits
of the intervals are connected by straightlines and the lower limits too, leading to two segmented lines. The intersections between
these two segmented lines with the acceptance limit straight lines
y =  and y =  (expressed in %) leads to upper and lower quantication limits [5,10].
The excellent e.noval software (Arlenda, Lige, Belgium) can be
used to obtain the accuracy proles and the validation results of
the analytical method [37]. Some authors, aside from the variance
of intermediate precision and of estimated bias, consider also the
robustness variance as a contribution to the total precision variance
[5,10,38]. Robustness tests can be considered as intra-laboratory
simulations of inter-laboratory studies, if the introduced deliberated variations in the method parameters are suitably selected.
Accordingly, the robustness uncertainty can be easily obtained as
a relative deviation [16], RSDrob , and can be joined to the budget of
the total precision variance RSD.
RSD =

2 + RSD2 + RSD2
RSDIP
rob

(18)

However, the robustness contribution to the total precision variance is generally disregarded by the authors and Eq. (15) is used
instead of (18).
4. Conclusion
Intra laboratory testing of method accuracy, when VS are available can be easily performed from the results of one-way ANOVA
designed experiments. Then bias and intermediate precision can be

1998

A.G. Gonzlez et al. / Talanta 82 (2010) 19951998

tested independently or in a global manner according to the concept of accuracy prole and -Expectation Tolerance Interval once
the acceptance limit has been selected.
Acknowledgement
Authors are grateful to the Junta de Andaluca (Spain) for grant
excellence research project P06-FQM-02029.
References
[1] ICH Harmonised Tripartite Guideline, Validation of Analytical Procedures: Text and Methodology, ICH Working Group, November 2005.
http://www.ich.org/LOB/media/MEDIA417.pdf.
[2] Guidance for Industry: Analytical Procedures and Methods Validation, U.S
Department of Health and Human Services, Food and Drug Administration,
Center for Drug Evaluation and Research (CDER), Center for Biologics Evaluation
and Research (CBER), August 2000. http://www.fda.gov/downloads/Drugs/
GuidanceComplianceRegulatoryInformation/Guidances/ucm122858.pdf.
[3] EURACHEM Guide, The Fitness for Purpose of Analytical Methods, A Laboratory
guide to Method Validation and related topics, EURACHEM Working Group,
December 1998. http://eurachem.org/guides/valid.pdf.
[4] IUPAC, Pure Appl. Chem. 74 (2002) 835855.
[5] A.G. Gonzlez, M.A. Herrador, Trends Anal. Chem. 26 (2007) 227238.
[6] T.P.J. Linsinger, Trends Anal. Chem. 27 (2008) 916923.
[7] A.G. Gonzlez, M.A. Herrador, A.G. Asuero, Talanta 48 (1999) 729736.
[8] International Organization for Standardization (ISO), ISO 5725-1. Accuracy
(trueness and precision) of measurement method and results. Part 1. General
principles and denitions, ISO, Geneva, Switzerland, 1994.
[9] A. Menditto, M. Patriarca, B. Magnusson, Accred. Qual. Assur. 12 (2007) 4547.
[10] A.G. Gonzlez, M.A. Herrador, Talanta 70 (2006) 896901.
[11] Ph. Hubert, J.J. Nguyen-Huu, B. Boulanger, E. Chapuzet, P. Chiap, N. Cohen,
P.A. Compagnon, W. Dew, M. Feinberg, M. Lallier, M. Laurentie, N. Mercier,
G. Muzard, C. Nivet, L. Valat, E. Rozet, J. Pharm. Biomed. Anal. 45 (2007) 7081.
[12] V.R. Meyer, J. Chromatogr. A 1158 (2007) 1524.
[13] M. Feinberg, J. Chromatogr. A 1158 (2007) 174183.
[14] M. Feinberg, B. Boulanger, W. Dew, P. Hubert, Anal. Bioanal. Chem. 380 (2004)
502514.
[15] M. Feinberg, M. Laurentie, Accred. Qual. Assur. 11 (2006) 39.
[16] A.G. Gonzlez, M.A. Herrador, A.G. Asuero, Talanta 65 (2005) 10221030.
[17] F.T. Peters, H.H. Maurer, Accred. Qual. Assur. 7 (2002) 441449.
[18] F.T. Peters, O.H. Drummer, F. Musshoff, For. Sci. Int. 165 (2007) 216224.
[19] D.L. Massart, B.G.M. Vandeginste, L.M.C. Buydens, S. De Jong, P.J. Lewi, J.
Smeyers-Verbeke, Handbook of Chemometrics and Qualimetrics, Part A, Elsevier, Amsterdam, The Netherlands, 1997.

[20] International Organization for Standardization (ISO), ISO-5752-2, Accuracy

(Trueness and Precision) of Measurement Methods and Results. Part 2. Basic
Method for the determination of Repeatability and Reproducibility of a standard Measurement Method, ISO, Geneva, Switzerland, 1994.
[21] T. Luping, B. Schouenborg, Methodology of Inter-comparison Tests and Statistical Analysis of Test results. Nordtest Project No. 1483-99, SP Swedish National
Testing and Research Institute, SP report 2000:35, Boras, Sweden, 2000.
[22] M. Thompson, The Amazing Horwitz Function, AMC Technical Brief No. 17,
Royal Society of Chemistry, Cambridge, UK, July, 2004.
[23] R. Wood, Trends Anal. Chem. 18 (1999) 624632.
[24] W. Horwitz, Anal. Chem. 54 (1982) 67A76A.
[25] I. Chen, R.R. Eitenmiller, J. Food Chem. Toxicol. 72 (2007) C243C247.
[26] W. Horwitz, Validation: An invisible component of measurement, AOAC
Int., Gaithersburg, MD, 2003. http://www.aoac.org/dietsupp6/DietarySupplement-web-site/HorwitzValid.pdf.
[27] E. Pritchard, Quality in the Analytical Chemistry Laboratory. ACOL Series, Wiley,
Chichester, West Sussex, UK, 1995.
[28] L. Huber (Ed.), Validation and Qualication in Analytical Laboratories, Interpharm Press, East Englewood, CO, USA, 1998.
[29] AOAC Peer Veried Methods Program. Manual on policies and procedures,
AOAC Int., 1998. http://www.aoac.org/vmeth/PVM.pdf.
[30] International Organization for Standardization (ISO), ISO/DTS 21748, Guide to
the Use of Repeatability, Reproducibility and Trueness Estimates in Measurement Uncertainty Estimation, ISO, Geneva, Switzerland, 2003.
[31] Ph. Hubert, P. Chiap, J. Crommen, B. Boulanger, E. Chapuzet, N. Mercier, S.
Bervoas-Martin, P. Chevalier, D. Grandjean, P. Lagorce, M. Lallier, M.C. Laparra,
M. Laurentie, J.C. Nivet, Anal. Chim. Acta 391 (1999) 135148.
[32] B. Boulanger, P. Chiap, W. Dew, J. Crommen, Ph. Hubert, J. Pharm. Biomed.
Anal. 32 (2003) 753765.
[33] B. Boulanger, W. Dew, Ph. Hubert, Objectives of pre-study validation and decision rules, in: AAPS Conference. APQ Open Forum, Indianapolis, 2000.
[34] Ph. Hubert, J.J. Nguyen-Huu, B. Boulanger, E. Chapuzet, P. Chiap, N. Cohen,
P.A. Compagnon, W. Dew, M. Feinberg, M. Lallier, M. Laurentie, N. Mercier,
G. Muzard, C. Nivet, L. Valat, J. Pharm. Biomed. Anal. 36 (2004) 579
656.
[35] M. Feinberg, B. Boulanger, W. Dew, Ph. Hubert, Anal. Bioanal. Chem. 380 (2004)
502514.
[36] E. Rozet, C. Hubert, A. Ceccato, W. Dew, E. Ziemons, F. Moonen, K. Michail, R.
Wintersteiger, B. Streel, B. Boulanger, Ph. Hubert, J. Chromatogr. A 1158 (2007)
126137.
[37] Arlenda, Laboratory Solutions, Lige, Belgium. http://www.arlenda.com/enoval.html.
[38] V.J. Barwick, L.R. Ellison, VAM Project 3.2.1, Development and Harmonisation of Measurement Uncertainty Principles. Part D. Protocol for uncertainty
Evaluation from Validation Data, Report No: LGC/VAM/1998/088, January,
2000.