Forensic Science International 185 (2009) 59–66

Contents lists available at ScienceDirect

Forensic Science International

journal homepage: www.elsevier.com/locate/forsciint

Estimation of the measurement uncertainty of methamphetamine and

amphetamine in hair analysis
Sooyeun Lee *, Yonghoon Park, Wonkyung Yang, Eunyoung Han, Sanggil Choe, Miae Lim, Heesun Chung
National Institute of Scientific Investigation, 331-1 Sinwol-7-dong, Yangcheon-gu, Seoul 158-707 Republic of Korea

A R T I C L E I N F O A B S T R A C T

Article history: The measurement uncertainties (MUs) were estimated for the determination of methamphetamine
Received 27 January 2008 (MA) and its main metabolite, amphetamine (AP) at the low concentrations (around the cut-off value of
Received in revised form 15 December 2008 MA) in human hair according to the recommendations of the EURACHEM/CITAC Guide and ‘‘Guide to the
Accepted 18 December 2008
expression of uncertainty in measurement (GUM)’’. MA and AP were extracted by agitating hair with 1%
HCl in methanol, followed by derivatization and quantification using GC–MS. The major components
Keywords: contributing to their uncertainties were the amount of MA or AP in the test sample, the weight of the test
Measurement uncertainty
sample and the method precision, based on the equation to calculate the mesurand from intermediate
Hair analysis
values. Consequently, the concentrations of MA and AP in the hair sample with their expanded
Methamphetamine
Amphetamine uncertainties were 0.66  0.05 and 1.01  0.06 ng/mg, respectively, which were acceptable to support the
Method precision successful application of the analytical method. The method precision and the weight of the hair sample gave
the largest contribution to the overall combined uncertainties of MA and AP, for each.
ß 2009 Elsevier Ireland Ltd. All rights reserved.

1. Introduction attributed to the measurand and expressed the number after .

Quality assurance is a major concern in the area of forensic
toxicology because the analytical results have a great effect on
administrative and legal consequences. In recent years, many
analytical chemistry laboratories have employed a quality
management system, compliant to international quality standards
such as ISO/IEC 17025. For forensic laboratories, the Crime
Laboratory Accreditation Programs of the American Society of
Crime Laboratory Directors/Laboratory Accreditation Board
(ASCLD/LAB) have been well known as accreditation programs
and the ASCLD/LAB-International Program was newly established
in 2003 based on the ISO/IEC 17025 standards and the ASCLD/LAB-
International Supplemental Requirements [1]. These programs
require applicants and/or accredited laboratories to demonstrate
the reliability of analytical data objectively in various ways: proper
validation of analytical methods, regular participation in profi-
ciency tests, etc.
The estimation of the measurement uncertainty (MU) is
another accreditation requirement of ISO/IEC 17025. According
to the EURACHEM/CITAC Guide [2], the MU is defined as a
parameter, associated with the result of a measurement, that
characterizes the dispersion of values that could reasonably be
* Corresponding author. Tel.: +82 2 2600 4937; fax: +82 2 2600 4939.
E-mail address: spp26625@yahoo.co.kr (S. Lee).
Nowadays, it is recommended that the quantitative findings should
be presented with their MU to evaluate the reliability of the results
[3,4].
The application of drug analysis in hair has significantly
increased in both forensic and clinical toxicology: illegal drug use,
postmortem cases, drug-facilitated crimes, workplace drug testing,
doping control, drug treatment or management programs, clinical
diagnosis, gestational drug exposure, therapeutic drug monitoring
(TDM), etc. [5,6]. In Korea, hair analysis is critical because it is
accepted by law enforcement agencies as one of important
corroborative facts for drug abuse. The hair analysis provides
information not only on chronic drug use but also on drug use
period according to the rate of hair growth [5], which has a key
effect on legal decision. Usually, the result of hair testing is
reported as positive and negative, for which several criteria were
proposed for obtaining a positive result: evaluation of possible
passive contamination, identification of metabolites, use of
metabolites-to-parent drug ratios, use of assay values of deconta-
mination washes and use of threshold values (cut-off values) [7,8].
In case other criteria are satisfactory, the determination of positive
or negative depends significantly on the quantification results, so
the cut-off value can be a key criterion. However, it can be difficult
to determine if a quantified value is positive or negative when the
value is very close to the cut-off value. In the human sports testing
field, the decision limit, which is the threshold plus the MU, is
adopted to establish a doping offence [9,10]. Even though the MU

0379-0738/$ – see front matter ß 2009 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.forsciint.2008.12.012
60 S. Lee et al. / Forensic Science International 185 (2009) 59–66
Fig. 1. Hair analysis procedure.
could be applied in a different way to the identification of a
prohibited substance to supply legal evidence, it is necessary to
establish the MU at the cut-off value as one of supplementary
criteria in forensic analysis.
Methamphetamine (MA) has received the most attention as a
drug of abuse in Korea. It undergoes some N-demethylation to
amphetamine (AP), its major active metabolite [11]. As stated in
our standard operating procedure (SOP), a MA concentration of
greater than the cut-off value (0.5 ng/mg) [7] and an AP greater
than lower limit of detection (LLOD) along with appropriate
compound ion ratios are required to prove an individual’s MA use.
Therefore, this research aims to estimate the MU of MA and AP at
the low concentrations (around the cut-off value of MA) in human
hair.
2. Experimental
2.1. Chemicals
Methanol, ethyl acetate and hydrochloric acid (HCl) were analytical grade. MA
hydrochloride and AM sulfate were obtained from Lipomed AG (Switzerland) for
the preparation of a spiked hair sample. The solutions of MA, AP, MA-d5 and AP-d5
(1 mg/ml, 99% for each) from Cerilliant (TX, USA) were used to prepare the stock and
working standard solutions. Dimethylsulfoxide (DMSO) and Trifluoroacetic
anhydride (TFAA) were purchased from Sigma–Aldrich (MO, USA).
2.2. Preparation of a hair sample
The preparation of a hair sample was carried out using a slight modification of a
method described elsewhere [13]. Briefly, each ca. 60 mg of MA hydrochloride and
AM sulfate were dissolved in a small volume of distilled water in a 1000 ml glass
beaker. Then, 250 ml of 0.02 M HCl in DMSO and 250 ml distilled water were added
in succession in the beaker in an ice bath. Drug-free hair (ca. 10 g) was soaked in the
solution for 24 h. The hair was removed, rinsed in a sufficient volume of methanol
several times and air-dried.
2.3. Preparation of stock and working standard solution
Each 1 mg/ml of MA and AP was diluted in a 10 ml volumetric flask (Flask 10)
with 1% HCl in methanol (Stock solution, 99 mg/ml). The working standard solution
was prepared by two steps: pipetting (Pipette 1) 1 ml of the stock solution into a
100 ml volumetric flask (Flask 100) and diluting with 1% HCl in methanol
(Intermediate standard solution) followed by pipetting (Pipette 1) 1 ml of the
intermediate standard solution into a 10 ml volumetric flask (Flask 10) and diluting
with 1% HCl in methanol (Working standard solution). A 10–100 ml autopipette
(Autopipette A) and a 200–1000 ml autopipette (Autopipette B) were used to make
five sets of calibrators (2.5, 5, 10, 25, 50, 75 and 100 ng).
2.4. Hair analysis
The experimental procedure (Fig. 1) given here was fully validated [12]. Shortly,
triplicate hair samples were accurately weighed (ca. 10 mg), cut into very small
pieces of less than 1 mm and agitated with 3 ml of 1% HCl in methanol for 20 h at
38 8C. MA-d5 and AP-d5 were added as internal standards. The hair extract was
evaporated to dryness at 45 8C under N2 gas and then the residue was derivatized
with 100 ml of TFAA/ethyl acetate (1:1) at 65 8C for 15 min. The excess derivatizing
reagent was removed and the residue was reconstituted in ethanol for GC–MS
analysis (Agilent 6890/5973 GC-MS system). The GC was equipped with a 30-m-
long, 0.25-mm-i.d., 0.25-mm-film-thickness HP-5MS capillary column. The inlet
temperature was 260 8C and the helium flow rate was 1.0 ml/min. The oven was
programmed to operate at an initial temperature of 100 8C for 1 min, to increase the
temperature to 270 8C at a heating rate of 20 8C/min and to hold at 270 8C for
10 min. The MS was operated in selected ion monitoring (SIM) mode. The TFAA
derivatized ions for MA, AP, MA-d5 and AP-d5 were as follows: MA, m/z 154, 118,
110, 91; AP, m/z 140, 118, 91; MA-d5, 158, 122; AP-d5, 144, 122.
2.5. Evaluation of method precision
Method precision was examined by analyzing hair samples spiked with low
(8 ng), medium (40 ng) and high (80 ng) concentrations of MA and AP, respectively,
using the same method described in the previous hair analysis section. The six
aliquots of each sample were analyzed on the first day, followed by triplicates for
the four consecutive days.
2.6. MU estimation
The estimation of the MU was performed in compliance with the EURACHEM/
CITAC Guide [2] and ‘‘Guide to the expression of uncertainty in measurement
(GUM)’’ [14]. In brief, there are six steps involved in developing the MU:
Specification of the measurand; Identifying uncertainty sources; Quantifying
uncertainty; Calculating the combined uncertainty; Calculating the degrees of
freedom; Calculating the expanded uncertainty. Therefore, this study followed the
process to derive an expanded uncertainty. Firstly, the mesurand was clearly
 S. Lee et al. / Forensic Science International 185 (2009) 59–66 61

expressed as a mathematical equation based on the experimental method.

Secondly, the possible sources of uncertainty were listed according to the equation
and an initial cause and effect diagram was proposed. Thirdly, considering the
relationship among parameters which have an effect on the sources, an expanded
cause and effect diagram was given. Then, the uncertainties of individual
components were evaluated and quantified using analytical data. In order to
adjust different units of measurement, the individual standard uncertainties (SU)
were changed into the relative standard uncertainties (RSU) equivalent. Next, each
RSU was combined by the root-sum-of-squares (RSS) method to give the combined
standard uncertainty (CSR). After that, the degree of freedom of the CSR was
calculated and an appropriate coverage factor was determined. Finally, an
expanded uncertainty was given using the coverage factor.

3. Results
Fig. 2. Initial cause and effect diagram.
3.1. Specification of the measurand

According to the EURACHEM/CITAC Guide [2], it is necessary to of the method precision (fprecision) as follows:
define what is being measured and express quantitatively the
c0
relationship among the measurands in the first step. The C MA or AP ¼  f precision ðng=mgÞ (3)
W
concentration of MA or AP in the test sample (CMA or AP) was
calculated by a simple mathematical model:
3.2. Identifying uncertainty sources
c0
C MA or AP ¼ ðng=mgÞ (1)
W The relevant uncertainty sources are shown in the cause and
effect diagrams (Figs. 2 and 3). From Eq. (1), CMA or AP was
where c0 is the amount of MA or AP in the test sample and W is the associated with c0 and W, which were the major bones in the
weight of the test sample. Because c0 was derived from a initial cause and effect diagram (Fig. 2). c0 was calculated from a
calibration curve, CMA or AP was also given by calibration curve and W was subjected to two main sources of
uncertainty: variation of a balance and repeatability of measure-
A 0  B0 1 ment. This initial cause and effect diagram was expanded with
C MA or AP ¼  ðng=mgÞ (2)
B1 W the added correction factor, method precision. Moreover, the
working standard solution was considered as another uncer-
where A0 is the peak area ratio of MA or AP to IS (MA-d5 or AP-d5) tainty source of c0 because hair samples spiked with standard
and B0 and B1 is the intercept and the slope of the calibration curve, solutions were used instead of a certified reference material. Also,
respectively. Furthermore, since the method precision should be the repeatability of measurement and the variation of solvent
considered as a major source of the MU in any quantification volume due to temperature change overlap with the uncertainty
method [9], the equation was expanded using the correction factor of the method precision, where those factors are already covered

Fig. 3. Expanded cause and effect diagram.

62 S. Lee et al. / Forensic Science International 185 (2009) 59–66

[9]. Therefore, the effect of the repeatability of weighing hair was
not included and those of the repeatability and the temperature
change in using a 10 ml volumetric flask and 1 ml pipette to
prepare the working standard solution were not considered.
However, the repeatability and the temperature effect of the
volumetric apparatus for the preparation of both the stock
solution and the intermediate standard solution were regarded as
contributors to the estimation of the MU since the solutions were
made beforehand. Finally, the expanded cause and effect diagram
was given as Fig. 3.
3.3. Quantifying uncertainty
3.3.1. c0
In order to create the RSU of c0 (ur(c0)), the two uncertainty
components, one from the preparation of the working standard
solution and the other from the calibration curve, were combined
as follows: for MA,
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ur ðc0 Þ ¼ u2r ðStdÞ þ u2r ðCalÞ ¼ u2r ðStdÞ þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
0:0011
¼ 0:01312 þ ¼ 0:0131
6:5799
and for AP,
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
0:0022 2
ur ðc0 Þ ¼ 0:01312 þ ¼ 0:0131
10:0763
where ur(Std) is the RSU of the working standard solution, ur(Cal) is
that of the calibration curve and u(Cal) is the SU of the calibration
curve. Since the working standard solutions of MA and AP were
prepared using chemicals with same purity and identical
apparatus, the uncertainties ur(Std) were equal. For the calculation
of ur(Std), following uncertainties were considered: uncertainty in
the preparation of the stock solution, ur(Stock); uncertainty in the
use of Autopipette A to make the calibrators, u(AutopipetteA);
uncertainty in the use of Autopipette B to make the calibrators,
u(AutopipetteB); uncertainty in the use of Pipette 1 to make the
intermediate standard solution, u(Pipette1); uncertainty in the use
of Pipette 1 to make the working standard solution, u0 (Pipette1);
uncertainty in the use of Flask 10 to make the working standard
solution, u(Flask10); uncertainty in the use of Flask 100 to make the
 2
uðCalÞ intermediate standard solution, u(Flask100). The detailed calcula-
c0 tion of ur(Std) was summarized in Table 1.
According to the manufacturer’s certificate of analysis, it was
assumed that the purity of each solution of MA and AP was
100  1%, which was converted to 1.0  0.01 mg/ml. A rectangular

Table 1
A summary of the calculation of ur(Std) for MA and AP.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
       0   2  2
uðAuto pi petteAÞ 2 uðAuto pi petteBÞ 2 uðPi pette1Þ 2 u ðPi pette1Þ 2 uðFlask10Þ uðFlask100Þ
ur ðStdÞ ¼ u2r ðStockÞ þ 3  þ4 þ þ þ þ
Vol:ðAuto pi petteAÞ Vol:ðAuto pi petteBÞ Vol:ðPi pette1Þ Vol:ðPi pette1Þ Vol:ðFlask10Þ Vol:ðFlask100Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
           
0:0001 2 0:0005 2 0:0095 2 0:0035 2 0:0144 2 0:3751 2
¼ 0:00702 þ 3  þ4 þ þ þ þ ¼ 0:0131
0:1 1 1 1 10 100
Uncertainty in the preparation of the stock solution: ur(Stock)
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

uðPurityÞ
2 
0:006 2
  2ffi
ur ðStockÞ ¼ Purity þ uðFlask10Þ
Volume
¼ 0:99 þ 0:0391 10 ¼ 0:0070

uðPurityÞ ¼ 0:01
pffiffi ¼ 0:006 mg=ml
3

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
uðFlask10Þ ¼ u2 ðTol10Þ þ u2 ðTem p10Þ þ u2 ðRe peat10Þ ¼ 0:01442 þ 0:03632 þ 0:00162 ¼ 0:0391 ml

uðTol10Þ ¼ 0:025
pffiffi ¼ 0:0144 ml
3

pffiffi ¼ 0:0363 ml; ð10  5  1:259  103 Þ ¼ 0:0630 ml

uðTem p10Þ ¼ 0:0630
3

uðRe peat10Þ ¼ 0:0049

pffiffiffiffi ¼ 0:0016 ml
10

Uncertainty in the use of Autopipette A to make the calibrators: u(AutopipetteA)

uðAuto pi petteAÞ ¼ 0:0002

2 ¼ 0:0001 ml
Uncertainty in the use of Autopipette B to make the calibrators: u(AutopipetteB)
uðAuto pi petteBÞ ¼ 0:001
2 ¼ 0:0005 ml

Uncertainty in the use of Pipette 1 to make the intermediate standard solution: u(Pipette1)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uðPi pette1Þ ¼ u2 ðTol1Þ þ u2 ðTem p1Þ þ u2 ðRe peat1Þ ¼ 0:00352 þ 0:00362 þ 0:00812 ¼ 0:0095 ml

uðTol1Þ ¼ 0:006
pffiffi ¼ 0:0035 ml
3

pffiffi ¼ 0:0036 ml; ð1  5  1:259  103 Þ ¼ 0:0063 ml

uðTem p1Þ ¼ 0:0063
3

uðRe peat1Þ ¼ 0:0255

pffiffiffiffi ¼ 0:0081 ml
10

Uncertainty in the use of Pipette 1 to make the working standard solution: u0 (Pipette1)
u0 ðPi pette1Þ ¼ uðTol1Þ ¼ 0:006
pffiffi ¼ 0:0035 ml
3

Uncertainty in the use of Flask 10 to make the working standard solution: u(Flask10)
uðFlask10Þ ¼ uðTol10Þ ¼ 0:025
pffiffi ¼ 0:0144 ml
3
Uncertainty in the use of Flask 100 to make the intermediate standard solution: u(Flask100)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
uðFlask100Þ ¼ u2 ðTol100Þ þ u2 ðTem p100Þ þ u2 ðRe peat100Þ ¼ 0:09242 þ 0:36342 þ 0:00662 ¼ 0:3751 ml
uðTol100Þ ¼ 0:16
pffiffi ¼ 0:0924 ml
3
pffiffi ¼ 0:3634 ml; ð100  5  1:259  103 Þ ¼ 0:6295 ml
uðTem p100Þ ¼ 0:6295
3

uðRe peat100Þ ¼ 0:0208

pffiffiffiffi ¼ 0:0066 ml
10
 S. Lee et al. / Forensic Science International 185 (2009) 59–66 63

distribution is used when a certificate or other specification gives Table 2

Calibration results.
limits without specifying a level of confidence [2]. Therefore, the
uncertainty of purity (u(Purity)) was calculated by the semi-range of Amount (ng) MA AP
0.01 divided by the function for the rectangular distribution
2.5 0.1151 0.0564
(Table 1). 5 0.1718 0.1009
Moreover, the uncertainty of tolerance of the glassware was 10 0.2814 0.2117
estimated using results obtained from manufacturer’s certificates 25 0.5620 0.4642
50 1.0220 0.9747
and the function of the rectangular distribution. Since the
75 1.5439 1.5069
glassware had been calibrated at 20 8C but the temperature of 100 2.0287 1.9678
our laboratory is adjusted at 20  5 8C, the effect of temperature on
B1 0.0195 0.0197
the volume of methanol was also considered using the coefficient of
B0 0.0720 0.0001
expansion of methanol, 1.259  103. The uncertainty of repeat- r 0.9987 0.9991
ability of the glassware was calculated using the standard deviation
B1, slope; B0, intercept; r, correlation coefficient.
from ten measurements (Table 1).
The uncertainties of Autopipette A and B in their certificates of
calibration were 0.0002 and 0.001 ml, respectively, with a 95% respectively. Therefore, the uncertainty u(Cal) of MA was given by
degree of confidence. Therefore, the uncertainties were standar- as follows:
dized to a 68% degree of confidence (Table 1). sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
S 1 1 ðc0  cÞ2
The calibration curve was given by Aj = ciB1 + B0 where Aj is the uðCalÞ ¼ þ þ
jth measurement of the peak area ratio of the ith calibration B1 p n Sxx
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
standard, ci is the amount of the ith calibration standard and B1 and 0:0000562 1 1 ð6:5799  38:21Þ2
B0 are the slope and the intercept of the calibration curve, ¼ þ þ ¼ 0:0011 ng
0:0195 3 35 8658:93
respectively. The seven calibration standards were measured five
times each (Fig. 4), providing the results of mean values of the with the residual standard deviation S given by
different calibration standards, slopes and intercepts for MA and Pn 2
j¼1 ½A j  ðB0 þ B1  c j Þ
AP in Table 2. The B1 and B0 of MA are 0.0195 and 0.0720 with a S¼ ¼ 0:0000316
n2
correlation coefficient r of 0.9987 and those of AP are 0.0197 and
0.0001 with r of 0.9991, for each. The triplicate hair samples were and
analyzed, leading to c0 of 6.5977 and 10.0763 ng for MA and AP, X
n
Sxx ¼ ðc j  cÞ2 ¼ 8658:93
j¼1

where B1 is the slope, p is the number of measurements to

determine c0, n is the number of measurements for the calibration,
c0 is the amount of MA in the test sample, c is the mean value of the
different calibration standards (n number of measurements), i is
the index for the number of calibration standards and j is the index
for the number of measurements to obtain the calibration curve.
The uncertainty u(Cal) of AP was calculated as the same manner,
resulting in 0.0022 ng.

3.3.2. W
The uncertainty of weighing the hair sample (u(W)) was
equivalent to that of the variation of a balance. Thus, for the
estimation of u(W), the uncertainties of tolerance (u(Tol)) and
resolution (u(Resol)) of the balance were combined as follows:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uðWÞ ¼ u2 ðTolÞ þ u2 ðResolÞ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
¼ ð2:050  104 Þ þ ð2:887  105 Þ ¼ 2:070  104 g
¼ 0:2070 mg

0:00041
uðTolÞ ¼ ¼ 2:050  104 g
2

0:0001=2
uðResolÞ ¼ pffiffiffi ¼ 2:887  105 g
3

Using the value obtained, the RSU of weighing was ur

(W) = 0.2070/10 = 0.0207.

3.3.3. Method precision

The uncertainties of the method precision for MA and AP were
calculated using the pooled standard deviation (sp) given by
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uPN
u 2
i¼1 vi  si sp
sp ¼ t P N
and u ¼ pffiffiffiffiffi
i¼1 i v m
Fig. 4. Calibration curves of MA and AP (r, correlation coefficient).
64 S. Lee et al. / Forensic Science International 185 (2009) 59–66

Table 3 3.4. Calculating the overall combined uncertainty

Results of the evaluation of method precision.

Spiked Day Mean Standard Degrees of SU (ng) RSU The concentration of MA or AP in the test sample (CMA or AP) was
amount deviation freedom calculated using the equation (1) as follows:
(ng)
6:5799
(A) MA
C MA ¼  1:0 ¼ 0:66 ng=mg
10
8 1 6.69 0.30 5 0.1772 0.0221
2 7.97 0.69 2 10:0763
3 8.62 0.45 2 C AP ¼  1:0 ¼ 1:01 ng=mg
10
4 8.20 0.26 2
5 7.15 0.21 2
The RSUs of three components, c0, W and method precision,
40 1 36.04 0.46 5 0.4346 0.0109
2 42.08 0.31 2 were combined and the CSUs of MA and AP were given by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3 45.16 0.95 2
uc ðC MA Þ ¼ C MA  u2r ðc0 Þ þ u2r ðWÞ þ u2r ð f precision Þ
4 41.53 0.65 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
5 44.93 2.05 2 ¼ 0:66 0:01312 þ0:02072 þ0:02532 ¼ 0:0232 ng=mg
80 1 76.43 0.92 5 0.4464 0.0056
2 85.40 1.23 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3 87.09 1.54 2 uc ðC AP Þ ¼ C AP  u2r ðc0 Þ þ u2r ðWÞ þ u2r ð f precision Þ
4 80.18 0.50 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
5 85.45 0.46 2 ¼ 1:01 0:01312 þ0:02072 þ0:01912 ¼ 0:0313 ng=mg

CSU – – – – 0.6477 0.0253

3.5. Calculating the degrees of freedom
(B) AP
8 1 6.88 0.29 5 0.1333 0.0167
2 7.88 0.33 2 The degrees of freedom for MA and AP were approximated by
3 8.22 0.36 2 the Welch-Satterthwaite formula:
4 8.39 0.20 2 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u4 ðyÞ
5 7.66 0.31 2 veff ¼ PN c 4
i¼1 ui ðyÞ=vi
40 1 36.82 0.82 5 0.2996 0.0075
2 42.48 0.44 2
3 43.26 0.92 2 where veff is the effective degree of freedom, uc(y) is the combined
4 42.47 0.42 2
5 43.36 0.04 2
standard uncertainty, ui(y) is the individual standard uncertainty,
80 1 77.11 1.24 5 0.4403 0.0055 and vi is the degree of freedom of ui(y). A summary of the
2 84.95 1.13 2 calculation of the degrees of freedom for MA and AP is shown in
3 86.60 0.29 2 Table 4. The degrees of freedom veff of MA and AP were 74.4 and
4 81.01 0.24 2
144.9, for each.
5 85.16 1.01 2

CSU – – – – 0.5490 0.0191 3.6. Calculating the expanded uncertainty

SU: standard uncertainty; RSU: relative standard uncertainty; CSR: combined
standard uncertainty.
where vi is the degree of freedom of the ith sample, si is the
standard deviation of the ith sample and m is the number of
independent measurements. The results of the evaluation of the
method precision for MA and AP with their SU, RSU and CSU are
shown in Table 3. The uncertainties ur(fPrecision) of MA and AP were
0.0253 and 0.0191, respectively.
The degrees of freedom of MA and MA were large enough to
consider the coverage factor (k) as 2 at the 95% significance level.
Thus, the expanded uncertainties of CMA and CAP were given by
UðC MA Þ ¼ 2  0:0232  0:05 ng=mg
UðC AP Þ ¼ 2  0:0313  0:06 ng=mg
Therefore, the concentrations of MA and AP in the hair sample
with their expanded uncertainties were 0.66  0.05 and
1.01  0.07 ng/mg, respectively.

Table 4
A summary of the calculation of the degrees of freedom for MA and AP.

A. MA
Uncertainty factor c0 W Method precision
1.158×109
vi
Working standard
solution
Calibration curve ∞ 19.8137
∞ 33
veff 74.4

B. AP
Uncertainty factor c0 W Method precision
4.392×108
vi
Working standard
solution
Calibration curve ∞ 20.6578
∞ 33
veff 144.9
c0, the amount of MA or AP in the test sample; W, the weight of the test sample; vi, the degrees of freedom of the individual
standard uncertainty; veff, the effective degrees of freedom.
 S. Lee et al. / Forensic Science International 185 (2009) 59–66
4. Discussion and conclusion
The estimation of the MU has become an issue in the quality
control of clinical, doing and forensic laboratories because it has an
effect on a clinical diagnosis and a judicial decision as well as the
interpretation of analytical results. Even though the assessment of
the MU was originated from the area of clinical chemistry, its
importance has been lately realized in forensic analytical
chemistry, where uncertainty not only has to be calculated with
precision, but it also has to be both small and reliable enough to
support effective decision making [10]. Moreover, the documenta-
tion of the MU is essential to acquire an accreditation. Our
laboratory has been especially concerned about the reliability of
analytical findings for the determination of MA and AP in hair
because we need to prepare for the recent social changes such as
the internationalization of MA-related crimes as well as the
introduction of a new judicial system in Korea.
Until now, many studies on the estimation of the MU in
analytical procedures have been conducted [4,9,15–17]. In those
studies, the different components contributing to the MU were
considered depending on the analytical method. In doping control
for horses, the preparation of the calibrators and the test sample,
the method precision and recovery were the four main sources for
the testosterone quantification and the largest uncertainty was
originated from the method precision [9]. Gullberg reported that
biological/sampling, analytical and traceability components were
involved to calculate the MU in forensic breath-alcohol analysis
[15], among which the biological/sampling component contrib-
uted most to the overall combined uncertainty. In the study of the
simultaneous determination of V(V) and Mo(VI) at trace levels in a
synthetic water sample after precolumn chelation and extraction
with N-benzoyl-N-phenylhydroxylamine, the followings were
contributed to the overall uncertainty: the operational or working
uncertainty from the preparation of the stock, intermediate and
the calibration solutions, the inherent uncertainty from liquid–
liquid extraction, chromatographic separation and chemical
calibration and the uncertainty from method bias [4]. The overall
relative uncertainty in the determination of lead in blood was
obtained by combining the uncertainties of the precision studies,
the analysis of certified reference materials (CRMs) and recovery
tests [16]. According to another study published recently, jasmonic
acid in Lemna minor L. was determined by liquid chromatography
with fluorescence detection with its MU. The authors of this study
elucidated that sampling, sample processing and chromatographic
determination were considered as sources of the MU and the
method recovery and sample homogeneity were the main
contributors to uncertainty [17].
The method recovery was regarded as a effective factor in
previous studies [9,16,17]. In hair analysis, it is very difficult to
determine the extraction recovery because hair is a solid matrix.
Thus, determining if the recovery contributes to uncertainty
depends on extraction efficiency. The method employed in the
present paper was fully validated and showed good recoveries for
MA and AP [12]. However, this recovery was determined using
spiked hair samples as usual. Therefore, the extraction efficacy was
compared with another extraction method, the ultrasonication-
based method, which was also fully validated, using spiked and
authentic hair samples as well as NIST SRM 2379 certified
reference material. As a result, both of the methods were
acceptable to analyze MA and AP in hair [12]. Therefore, it was
considered that the recovery did not have a significant effect on the
MU in the current method.
The homogeneity of hair samples could affect the MU in actual
cases because the amounts of MA and AP in authentic hair can vary
drastically between individuals, between hairs of an individual or
even along hair shaft, depending on drug use patterns. Never-
65
Fig. 5. Contribution of the different sources to the overall combined uncertainties.
theless, the homogeneity was not considered in the current study
because the fortified sample was used. Further study on this is
definitely required in authentic hair samples.
In our study, we estimated the MU of MA and AP in hair analysis
combining the uncertainties from the amount of MA or AP in the
test sample, the weight of the test sample and the method
precision based on the equation to calculate the mesurand from
intermediate values. As a result, the RSUs of each parameter for MA
and AP were as follows: For MA, ur(c0) = 0.0131, ur(W) = 0.0207,
ur(fprecision) = 0.0253; for AP, ur(c0) = 0.0131, ur(W) = 0.0207, ur(f-
precision) = 0.0191. The influence of the method precision on the
overall combined uncertainty was most significant while that of
the calibration curve was negligible in MA. However, the weight of
the hair sample had a great effect on the overall combined
uncertainty in AP (Fig. 5).
According to recent publications, the MU varied depending on
a measured concentration [15,18]. The higher breath alcohol
concentration showed the higher standard uncertainty [15].
However, the relative uncertainty of benzodiazepines decreased
as the concentration increased [18]. From this point of view, Leung
et al. estimated the MU at the threshold level of testosterone in
horse urine and suggested that the MU should be established at
the various concentrations [9]. Therefore, we also determined the
MU of MA and its main metabolite, AP, around the cut-off value of
MA.
Consequently, this study showed a procedure and results to
estimate the MU of MA and AP around the cut-off value of MA,
which has been accepted in hair analysis by our laboratory. The
concentrations of MA and AP in the hair sample with their
expanded uncertainties were 0.66  0.05 and 1.01  0.06 ng/mg,
respectively. The method precision and the weight of the hair sample
gave the largest contribution to the overall combined uncertainties of
MA and AP, for each. The expanded uncertainties of both MA and AP
were acceptable, which supported the successful application of the
analytical method.
Acknowledgments
The authors thank Mr. Sungjoo Jeon and Youngsun Oh at the
Korea Testing and Research Institute for Chemical Industry (KTR)
in Korea for their support.
66 S. Lee et al. / Forensic Science International 185 (2009) 59–66
References
[1] http://www.ascld-lab.org/dual/aslabdualaboutascldlab.html.
[2] Quantifying uncertainty in analytical measurement, EURACHEM/CITAC Guide,
second ed., 2000.
[3] H.H. Maurer, Demands on scientific studies in clinical toxicology, Forensic Sci. Int.
165 (2007) 194–198.
[4] G. Bagura, M. Sánchez-Viñasa, D. Gázqueza, M. Ortegab, R. Romeroa, Estimation of
the uncertainty associated with the standard addition methodology when a
matrix effect is detected, Talanta 66 (2005) 1168–1174.
[5] P. Kintz, Value of hair analysis in postmortem toxicology, Forensic Sci. Int. 142
(2004) 127–134.
[6] F. Musshoff, B. Madea, New trends in hair analysis and scientific demands on
validation and technical notes, Forensic Sci. Int. 165 (2007) 204–215.
[7] R. Wennig, Potential problems with the interpretation of hair analysis results,
Forensic Sci. Int. 107 (2000) 5–12.
[8] P. Kintz, P. Mangin, What constitutes a positive result in hair analysis: proposal for
the establishment of cut-off values, Forensic Sci. Int. 70 (1995) 3–11.
[9] G.N. Leung, E.N. Ho, W.H. Kwok, D.K. Leung, F.P. Tang, T.S. Wan, A.S. Wong, C.H.
Wong, J.K. Wong, N.H. Yu, A bottom-up approach in estimating the measurement
uncertainty and other important considerations for quantitative analyses in drug
testing for horses, J. Chromatogr. A. 1163 (2007) 237–246.
[10] E. Spirito, F. Botre, The role of measurement uncertainty in doping analysis, Int. J.
Risk Assess. Manage. 5 (2005) 374–386.
[11] R.C. Baselt, Disposition of Toxic Drugs and Chemicals in Man, Chemical Toxicology
Institute, Foster City, CA, 2000, p. 527.
[12] S. Lee, Y. Park, W. Yang, E. Han, S. Choe, S. In, M. Lim, H. Chung, Development of a
reference material using methamphetamine abuser’s hair samples for the deter-
mination of methamphetamine and amphetamine in hair, J. Chromatogr. B 865
(2008) 33–39.
[13] M.J. Welch, L.T. Sniegoski, S. Tai, Two new standard reference materials for the
determination of drugs of abuse in human hair, Anal. Bioanal. Chem. 376 (2003)
1205–1211.
[14] Guide to the expression of uncertainty in measurement, International organiza-
tion for standardization, 1995.
[15] R. Gullberg, Estimating the measurement uncertainty in forensic breath-alcohol
analysis, Accred. Qual. Assur. 11 (2006) 562–568.
[16] M. Patriarca, M. Castelli, F. Corsetti, A. Menditto, Estimate of uncertainty of
measurement from a single-laboratory validation study: application to the
determination of lead in blood, Clin. Chem. 50 (2004) 1396–1405.
[17] J. Kristl, B. Krajnčič, D. Brodnjak-Vončina, M. Veber, Evaluation of measurement
uncertainty in the determination of jasmonic acid in Lemna minor L. by liquid
chromatography with fluorescence detection, Accred. Qual. Assur. 12 (2007) 303–
310.
[18] E.F. Dussy, C. Hamberg, T.A. Briellmann, Quantification of benzodiazepines in
whole blood and serum, Int. J. Legal Med. 120 (2006) 323–330.