Anal Bioanal Chem (2012) 403:755–768

DOI 10.1007/s00216-012-5877-y

ORIGINAL PAPER

Rapid detection and quantification of 35 benzodiazepines in urine

by GC-TOF-MS
Kathrin Arnhard & Rupert Schmid & Uwe Kobold &
Roland Thiele

Received: 2 December 2011 / Revised: 7 February 2012 / Accepted: 14 February 2012 / Published online: 25 March 2012
# Springer-Verlag 2012

Abstract A rapid and sensitive method for the screening Keywords Benzodiazepines . Gas chromatography/time-of-
and quantification of 35 benzodiazepines in human urine by flight mass spectrometry . Urine . Screening . Drug of abuse
gas chromatography/time-of-flight mass spectrometry was testing . Signal deconvolution
developed and validated. Target analytes were isolated from
1 ml urine by solid-phase extraction using Oasis MCX
extraction columns (extraction recovery between 35 and Introduction
99 %). With a supported liquid–liquid extraction method, a
new modification of conventional liquid–liquid-extraction, a Benzodiazepines are drugs frequently prescribed for the
less time intensive alternative for benzodiazepine extraction treatment of insomnia, anxiety, and epilepsy; however, they
is presented. The sample pretreatment entails the derivatiza- also have a relatively high dependence potential and are
tion of the benzodiazepines with N,O-bis(trimethylsilyl)tri- therefore often abused in cases of crime, suicide, or to
fluoroacetamide plus 1 % trimethylchlorosilane. Separation relieve withdrawal symptoms of other drugs [1, 2].
of all benzodiazepines was done within 9.5 min, and detec- For the determination of drugs and medicines, urine is
tion was based on full mass spectra for each analyte. A one of the preferred specimens, because the collection of
deconvolution algorithm was used for unresolved chromato- samples is easy and the concentrations of drugs are relative-
graphic peaks to identify coeluted substances. The subse- ly high. Metabolites play an important role in urine analysis
quent quantification was done using significant masses. The because elimination with the urine only occurs after metab-
limit of quantification is 10 ng/ml for most of the com- olization of the original substance. In the case of benzodia-
pounds. Linearity is in the range between 10 and 350 ng/ zepines, many of these metabolites are active and still have
ml. Reproducibility was observed with coefficients of vari- therapeutic effects.
ation below 2 % at concentrations of 50 and 200 ng/ml. The The major metabolic pathways of benzodiazepines
accuracy is between 88 and 108 % depending on the re- (phase I reactions) lead to formation of nordiazepam and
spective analyte and the concentration. temazepam (both active metabolites) through demethylation
or hydroxylation reactions. If the metabolites of phase I
reactions are sufficiently polar, they can be excreted at this
stage. However, most of the phase I products are not quickly
eliminated and undergo a subsequent reaction in which they
Electronic supplementary material The online version of this article are conjugated with glucuronic acid (see Fig. S1). Only after
(doi:10.1007/s00216-012-5877-y) contains supplementary material, this step do they lose their bioactivity.
which is available to authorized users. In unknown urine samples one cannot predict which
K. Arnhard (*) : R. Schmid : U. Kobold : R. Thiele (*) benzodiazepine or which of its metabolites are present. That
Roche Diagnostics GmbH, is why sensitive analytical screening methods are required.
Nonnenwald 2,
To develop a method which is suitable to screen multiple
82377 Penzberg, Germany
e-mail: [email protected] analytes at the same time, an analytical system which pro-
e-mail: [email protected] vides a fast scan rate and also has sufficient sensitivity to
756 K. Arnhard et al.

detect low concentrations of the compounds had to be Germany), and acetonitrile (J.T. Baker, Griesheim, Germany)
chosen. With respect to the sensitivity and because of its were high-performance LC grade. Anhydrous Na2HPO4
universality and resolution efficiency, gas-chromatographic (p.a.), NaH2PO4·H2O, and glacial acetic acid (p.a.) were used
separation seemed to be the best choice [3–7]. Another to prepare 100 mM phosphate buffer pH 6 and 7.5. Phosphoric
significant advantage of gas chromatography (GC)–mass acid and sodium chloride were used for 100 mM phosphate
spectrometry (MS) is that separated analytes can be identi- buffer pH 3. For pH 4.5 acetate buffer, 100 mM sodium
fied by comparison of their full mass spectra using reference acetate was used. Borate buffer pH 9 was prepared with boric
databases if the system is run in scan mode. acid and sodium hydroxide (all Merck, Darmstadt, Germany).
Usually, quadrupoles are the most widespread detectors in β-Glucuronidase (type H-2 from Helix pomatia) was
GC–MS, because they are comparatively favorable considering obtained from Sigma (Steinheim, Germany), and bis(tri-
the price–performance ratio and application range. However, methylsilyl)trifluoroacetamide (BSTFA)+1 % trimethyl-
for screening methods the detector should be able to scan fast chlorosilane (TMCS) (Silyl-991) was obtained from
enough to be run in full-scan mode and thereby achieve enough Chromatographie Service (Langerwehe, Germany).
(more than ten) data points across a peak. A time-of-flight For method validation, the certified reference material
(TOF) detector is predestined for this kind of requirement. Liquichek urine toxicology control (level C1) from Bio-Rad
So despite the tendency in recent years for liquid chro- (Irvine, CA, USA) was used.
matography (LC)–MS(/MS) methods to be preferred for the
detection of thermolabile and more polar drugs [8–12], a
Materials
GC-TOF-MS system for simultaneous screening and quan-
tification of 35 benzodiazepines was developed (see the
For SPE, Oasis MCX extraction cartridges (30 μm, 60 mg;
structures of the benzodiazepines in Fig. S2).
Waters, Eschborn, Germany), Bond Elut Certify large res-
The pretreatment of the urine samples was a key issue during
ervoir capacity (LRC) cartridges (200 mg), and Bond Elut
development of the method. Therefore, we systematically op-
AccuCAT LRC cartridges (200 mg; Agilent Technologies,
timized a solid-phase extraction (SPE) method, using strong-
Waldbronn, Germany) were used. The SLE was done with
cation-exchanger SPE columns, with regard to the influence of
Extrelut® NT 3 (Merck, Darmstadt, Germany), Isolute SLE
the pH and the eluent. Moreover, a supported liquid–liquid
+(5 ml, Biotage, Düsseldorf, Germany), and Isolute HM-N
extraction (SLE) method, which is less time-consuming, was
(5 ml; Biotage) columns.
developed as an alternative to conventional SPE.
A FactorFour VF-DA (12 m×0.2 mm, 0.33 μm) GC
column (Agilent Technologies, Waldbronn, Germany) was
used for the chromatographic separation .
Experimental

Chemicals
Preparation of standard solutions and calibrators
2-Hydroxyethylflurazepam, 7-aminoclonazepam, 7-aminoflu-
nitrazepam, α-hydroxyalprazolam, α-hydroxymidazolam, α- Benzodiazepine stock solutions (10 μg/ml) of oxazepam-d5,
hydroxytriazolam, desmethylflunitrazepam, and oxazepam-d5 desmethylflunitrazepam, 2-hydroxyethylflurazepam, α-hydro-
(at a concentration of 100 μg/ml) were purchased from LGC xyalprazolam, α-hydroxymidazolam, and α-hydroxytriazolam
Promochem (Wesel, Germany). 3-Hydroxyflunitrazepam, 4- in methanol, stock solutions of 7-aminoflunitrazepam and 7-
hydroxymidazolam, demoxepam, lormetazepam, N-ethylnor- aminoclonazepam (10 μg/ml) in acetonitrile, and stock solu-
dazepam, and N-ethyloxazepam were obtained from Lipomed tions with a concentration of 100 μg/ml of the remaining
(Weil am Rhein, Germany). 7-Aminonitrazepam was acquired benzodiazepines in methanol were prepared for urinary calibra-
from Toronto Research Chemicals (North York, ON, Canada) tors. For each benzodiazepine, a single stock solution was
and halazepam was acquired from United States Pharmacopeial prepared. A urinary calibrator was therefore spiked with 35
Convention (Rockville, MD, USA). Alprazolam, bromazepam, stock solutions except the oxazepam-d5 solution (internal stan-
chlordiazepoxide, diazepam, estazolam, flunitrazepam, loraze- dard), which was not added before sample preparation. Care
pam, midazolam, oxazepam, prazepam, temazepam, and triazo- was taken in the preparation of urinary calibrators to ensure a
lam were bought from Sigma-Aldrich (Schnelldorf, Germany). maximum organic content of 10 % in the respective calibrator.
Clonazepam, clorazepate, delorazepam, medazepam, nitraze- All stock solutions were stored at 4 °C.
pam, norchlordiazepoxide, nordiazepam, and pinazepam were Eleven calibration levels containing all benzodiazepines
obtained by in-house synthesis (Roche, Indianapolis, IN, USA). at concentrations of 8, 10, 25, 50, 75, 100, 200, 300, 350,
Methanol (Sigma-Aldrich, Schnelldorf, Germany), ethyl 400, and 500 ng/ml were prepared in human drug-free urine,
acetate, butyl acetate, n-hexane (Merck, Darmstadt, and the solutions were stored at -20 °C.
Rapid detection and quantification of 35 benzodiazepines 757

The 100 ng/ml internal standard solution of oxazepam-d5 thin layer. The aqueous phase is absorbed by the stationary
was also prepared in drug-free urine from a 10 μg/ml stock phase in this step. An efficient liquid–liquid extraction occurs
solution in methanol. when a suitable, water-immiscible organic solvent is applied.
The lipophilic substances are extracted from the aqueous
phase into the organic phase, and the aqueous phase remains
Biological samples on the sorbent.

The urine to prepare the calibrators was obtained from Derivatization

pooled urine samples from volunteers.
The last step of sample pretreatment involved a silylation of
Sample preparation the benzodiazepines. The eluates from either SPE or SLE
were transferred to 4-ml screw-cap vials and were evaporated
pH adjustment, internal standard spiking, and hydrolysis to dryness at 40 °C under a gentle stream of nitrogen. Ethyl
of aglycons from metabolites acetate (50 μl) and BSTFA+1 % TMCS (50 μl), as the
derivatization reagent, were added afterwards and the vial
A urine sample (1 ml) was spiked with 1 ml internal stan- tops were tightly screwed. Prior to injection of 1 μl into the
dard solution (oxazepam-d5, 100 ng/ml in drug-free urine). GC-MS system, the derivatization was performed at 70 °C for
This was followed by the addition of 1 ml 100 mM acetate 30 min on a heating block.
buffer (pH 4.5) and 200 μl of the enzyme β-glucuronidase
before the samples were incubated for 3 h at 65 °C to cleave
all glycated benzodiazepine metabolites. Then the samples Gas chromatography–time-of flight mass spectrometry
were centrifuged for 10 min at 2,000 rpm.
Chromatography
Solid-phase extraction
Analysis was done using an Agilent 7890 GC instrument
Subsequently, the supernatants (3 ml) were applied to Oasis operating in splitless mode (splitless time 120 s, back inlet
MCX extraction cartridges, which had previously been condi- purge flow 60 ml/min, back inlet septum purge flow 3 ml/
tioned with 1.5 ml methanol and 1.5 ml deionized water. After min). Helium was the used the carrier gas with a constant flow
the samples, under low vacuum (below 5 mmHg), had passed rate of 1 ml/min. Separation was performed on a VF-DA
through the cartridge, they were rinsed in four steps with 3 ml column (12 m×0.2 mm, 0.33 μm, Agilent). A temperature
deionized water, 1 ml 100 mM acetic acid, 3 ml acetonitrile/ gradient was applied starting from 80 °C (1 min) to 330 °C
100 mM phosphate buffer pH 6 (20:80, v/v), and 2 ml n-hexane. (1 min) with a ramp rate of 30 °C/min. The temperature of the
The column bed was dried under a full vacuum for 5 min. The transfer line was 225 °C and the injector temperature was set
analytes were eluted with 1.5 ml ethyl acetate/n-hexane (50:50, to 280 °C. The liner was a low pressure drop liner (4 mm×
v/v) and 1.5 ml dichloromethane/2-propanol/ammonia (78:20:2, 6.3 mm×78.5 mm) with intermediate polarity deactivation
v/v/v). The SPE workflow is shown in Fig. S3. with wool packing from Restek (Bad Homburg, Germany).

Supported liquid–liquid extraction Mass spectrometry

As an alternative to SPE, SLE was used for sample prepara- The TOF mass spectrometer, a Pegasus HT (Leco, St. Jo-
tion. The pH adjustment, spiking with the internal standard, seph, MI, USA), was operated in electron ionization mode
and the hydrolysis of aglycons were performed as described in (70 eV). Mass spectra were registered with a scan rate of 50
“pH adjustment, internal standard spiking, and hydrolysis of spectra per second to allow true signal deconvolution, in a
aglycons from metabolites.” The supernatants were subse- mass range of 50–500m/z. The temperature of the ion source
quently basified with 1 ml 100 mM borate buffer pH 9. was 200 °C, the solvent-delay time was set to 3.5 min, and a
Afterwards, the samples were applied to Extrelut NT 3 extrac- detector voltage of 1,650 V was applied. For quantification,
tion columns. The samples were allowed to spread over the one significant ion trace for each analyte was chosen.
chemically inert matrix of the column for 15 min. The extrac-
tion was done using 3×5 ml butyl acetate. SLE columns have Data analysis
a high capacity for retaining aqueous samples. When the urine
sample is loaded onto the extraction column, which is packed All acquired data were analyzed by ChromaTOF version 4.24.
with wide-pore, diatomaceous earth, the analytes spread over The detection of the sample components was achieved by
the supporting surface, which acts as a stationary phase, in a comparing the mass spectra of each peak with spectra stored
758 K. Arnhard et al.

in an in-house library database. Peaks were detected with a automatically extracted without interferences from the sys-
signal-to-noise ratio greater than 10 and a minimum similarity tem background, matrix background, and coeluted analytes,
match of 600, which means that the mass spectra of a peak must after the Automated Peak Find algorithm has located all of
have a minimal accordance of 600 (from a maximum of 1,000) the analytes because already minimal retention time differ-
with a reference spectrum in the mass spectra library in order ences are reflected in the mass spectra. TSD is the result of
to be assigned as a specific analyte. The mass threshold combining a fast acquisition and continuity of mass
(relative abundance of base ion) was set at 50. spectra.
The ChromaTOF program is equipped with True Signal Ion traces which are shared between coeluted analytes are
Deconvolution™ (TSD). With TSD, a mass spectrum is properly proportioned to provide a complete spectrum. The

Table 1 Comparison of different sample extraction methods regarding the recovery rates (%)

SLE SPE (Oasis MCX) SPE (AccuCAT)

Extrelut HM-N SLE+ single-elution two-step elution

2-hydroxyethylflurazepam 60.3 64.7 48.7 42.7 53.7 40.7

4-hydroxymidazolam 33.1 23.1 27.9 21.2 43.0 24.3
7-Aminoclonazepam 42.2 43.6 7.1 29.0 51.4 42.6
7-Aminoflunitrazepam 32.9 54.8 23.0 68.8 61.3 37.8
7-Aminonitrazepam 9.1 23.6 8.9 59.4 44.9 29.7
alpha-hydroxyalprazolam 84.3 87.1 63.9 59.8 74.0 60.6
alpha-hydroxymidazolam 104.6 63.2 61.5 69.9 77.0 47.1
alpha-hydroxytriazolam 94.9 97.6 74.0 69.5 61.2 60.9
alprazolam 88.9 95.9 63.9 54.3 57.7 51.7
bromazepam 75.3 88.9 53.0 71.1 72.1 50.9
chlordiazepoxide 79.9 72.8 53.9 78.0 54.8 16.0
clonazepam 64.9 97.9 50.5 25.3 73.2 61.5
delorazepam 74.7 70.6 66.7 58.1 62.0 49.3
demoxepam 79.0 90.1 69.2 55.8 98.8 58.5
desmethylflunitrazepam 63.5 67.3 45.7 80.1 40.7 34.6
diazepam 84.2 91.9 69.1 32.0 67.9 52.4
estazolam 86.9 97.6 58.8 53.9 49.3 46.6
flunitrazepam 90.4 104.5 102.7 62.0 35.0 43.2
flunitrazepam 3-hydroxy 74.1 224.8 25.1 18.4 46.8 46.7
halazepam 102.2 86.5 87.6 61.1 74.4 58.7
lorazepam 72.9 80.4 59.3 62.1 69.2 61.1
lormetazepam 84.2 87.7 69.1 57.7 71.1 60.9
medazepam 65.0 48.0 47.5 72.9 74.6 60.6
midazolam 52.4 64.9 38.6 65.4 78.2 47.7
N-ethylnordazepam 98.0 99.9 87.1 61.1 71.2 58.7
N-ethyloxazepam 95.0 97.5 80.7 64.6 82.4 64.6
nitrazepam 84.9 91.9 70.3 49.6 61.1 48.2
norchlordiazepoxide 2.7 5.1 0.0 41.1 73.1 15.1
oxazepam 89.2 86.1 98.1 78.0 67.1 74.2
pinazepam 103.4 88.7 106.2 52.5 52.1 46.5
prazepam 75.8 82.4 63.1 56.7 69.4 57.1
temazepam 85.1 84.4 71.7 55.7 67.3 54.5
triazolam 94.2 92.2 68.4 54.8 51.4 49.6

Supported liquid–liquid extraction (SLE) was performed with three different column types: Extrelut NT 3 (Merck), Isolute HM-N (Biotage), and
Isolute SLE+(Biotage): samples (pH 9) were eluted with butyl acetate. Solid-phase extraction (SPE) was performed with Oasis MCX and
AccuCAT (strong cation exchanger combined with strong anion exchanger) cartridges: for Oasis MCX cartridges, a single elution using 3 ml
dichloromethane/2-propanol/ammonia (78:20:2, v/v/v) and a two-step elution with 1.5 ml ethyl acetate/n-hexane (50:50, v/v) and 1.5 ml dichloro-
methane/2-propanol/ammonia (78:20:2, v/v/v)
Rapid detection and quantification of 35 benzodiazepines 759

resulting spectrum (Peak True) is then used for identifi- in-house database was implemented in the NIST database
cation of the analyte by spectra interpretation and by a and works according to the same algorithm as the NIST’s
library search using an in-house-prepared database. This reverse library search.

Fig. 1 Total ion current (TIC)

chromatogram of two blank a
urine samples (without the
4e+007
internal standard): a solid-phase
extraction (SPE) sample
preparation (Oasis MCX) and b
supported liquid–liquid 3.5e+007
extraction sample preparation
(Extrelut NT 3). Note that
blanks of body fluids are 3e+007
always unique owing to the
variance of their composition
2.5e+007

2e+007

1.5e+007

1e+007

5e+006

T ime (min:sec) 4:10.00 5:00.00 5:50.00 6:40.00 7:30.00 8:20.00 9:10.00

b
4e+007

3.5e+007

3e+007

2.5e+007

2e+007

1.5e+007

1e+007

5e+006

T ime (min:sec) 4:10.00 5:00.00 5:50.00 6:40.00 7:30.00 8:20.00 9:10.00

760 K. Arnhard et al.

Validation Limits of quantification and limits of detection

Validation experiments were performed following corporate The limit of quantification (LOQ) was defined as the lowest
procedures, which are aligned to ICH guidelines for analyt- concentration of the urinary calibrators which could be
ical method validation [13]. quantified with precision within ±25 %. In addition, a
signal-to-noise ratio above 10 was required for the LOQ
Linearity level. The upper LOQ was chosen as the highest concentra-
tion of the calibrator which could be applied to obtain linear
Calibration curves were prepared at concentrations rang- calibration curves (R≥0.99). The limit of detection was set
ing from 8 to 500 ng/ml (8, 10, 25, 50, 75, 100, 200, to the lowest concentration where the signal of the com-
300, 350, 400, 500 ng/ml) for all analytes to cover 80– pound was threefold higher than the background noise.
120 % of the working range. They were established by
linear least-squares regression, by plotting the relative Precision and accuracy
response (area analyte/area internal standard) as a func-
tion of analyte concentration. The linearity was evaluated Precision was examined by the calculation of relative stan-
from the resulting plots and the correlation coefficients dard deviations of urine samples spiked with benzodiaze-
(R≥0.99). pines at three different concentrations: 10 ng/ml (n06), the

Fig. 2 Chromatogram of an
urinary calibrator (200 ng/ml)
after extraction with Oasis
MCX cartridges and
derivatization containing all 35
benzodiazepines plus the
internal standard (ISTD)
oxazepam-d5 (100 ng/ml). TMS ISTD
trimethylsilane

242+341+324+434+429+388+256+270+297+371+394+282+352+280+357+343+310+286+387+288+269+377+355+398+372+205+279+381+313+415

1 medazepam 13 7-aminonitrazepam TMS 26 7-aminoflunitrazepam TMS

2 demoxepam / 14 nitrazepam TMS 27 4-OH-midazolam TMS
nordiazepam/clorazepate 15 pinazepam 28 3-OH-flunitrazepam TMS
3 halazepam 16 chlordiazepoxide 29 alpha-OH-midazolam
4 oxazepam D5 TMS (ISTD) 17 N-ethyloxazepam 30 Estazolam
5 oxazepam 2TMS 18 temazepam TMS 31 Alprazolam
6 delorazepam TMS 19 midazolam 32 alpha-OH-alprazolam TMS
7 bromazepam TMS 20 7-aminoclonazepam TMS 33 Triazolam
8 lorazepam 2TMS 21 flunitrazepam 34 alpha-OH-triazolam TMS
9 diazepam 22 clo n azep am TM S
10 N-ethylnordazepam 23 2-OH-ethylflurazepam
11 norchlordiazepoxide TMS 24 lormetazepam TMS
12 desmethylflunitrazepam TMS 25 prazepam
Rapid detection and quantification of 35 benzodiazepines 761

lower LOQ for most of the analytes, 50 ng/ml (n06), and previously described sample preparation protocol, evaporated to
200 ng/ml (n03). Moreover, a certified reference material dryness and derivatized. The relative response (analyte/internal
(Bio-Rad Liquichek urine toxicology control, level C1) standard) obtained when injecting the methanolic solution was
containing α-hydroxyalprazolam at a concentration of used as 100 % of the extraction recovery value for the respective
80 ng/ml was analyzed in six repetitions. analyte. See Fig. S5 for a schematic depiction of how the
Accuracy was investigated by comparing the nominal con- extraction recoveries were evaluated.
centration with the calculated concentration after calibration.

Extraction recovery Results and discussion

The extraction recoveries of the SPE and SLE methods were Optimization of the SPE method
calculated by comparing the peak area ratio of the individual
analyte with that of the internal standard. Therefore, all benzo- To achieve high extraction recoveries and thus enhance the
diazepines including the internal standard at a concentration of sensitivity of the analytical method, sample extraction methods
100 ng/ml were added to a blank urine sample before sample were optimized regarding the pH of the samples, the type and
preparation. As a control sample, a methanolic solution contain- volume of solvents used for removing the matrix and the
ing the same analyte concentrations and the internal standard elution of the analytes, and the extraction cartridges.
was prepared. This methanolic solution was not subject to the During the development of a suitable sample pretreat-
extraction procedure. The eluates from sample extraction as well ment the following SPE cartridges were used: Waters Oasis
as the methanolic solution were then, in accordance with the MCX, Agilent Bond Elut Certify LRC and Agilent AccuCAT

Fig. 3 True Signal Deconvolution demonstrated with the example of the chromatogram in the left corner depicts selected ion traces for lormeta-
coeluted analytes lormetazepam trimethylsilane (TMS) and prazepam: the zepam TMS (377 Da) and prazepam (269 Da). The corresponding mass
main chromatogram shows the TIC of an extracted and silylated urine spectra are shown at the bottom: left the mass spectra of prazepam; right
sample, and contains 35 benzodiazepines (each 100 ng/ml); the enlarged the mass spectra of lormetazepam TMS
762 K. Arnhard et al.

LRC. The Oasis MCX and Bond Elut Certify LRC cartridges Two different elution possibilities were investigated in
were comparable regarding the recovery rates because both the next step: a single elution using 3 ml dichloromethane/
sorbents consist of a strong cation exchanger. Because of a 2-propanol/ammonia (78:20:2, v/v/v) and a two-step elution
less disturbing matrix background in the chromatograms with 1.5 ml ethyl acetate/n-hexane (50:50, v/v) and
when using Oasis MCX cartridges, these cartridges were 1.5 ml dichloromethane/2-propanol/ammonia (78:20:2,
favored (see Fig. S4). The extraction recoveries were better v/v/v). The two-step elution was found to be the more
with the Oasis MCX cartridges than with the mixed-mode efficient method owing to the higher recovery rates
AccuCAT LRC cartridges (strong cation and strong anion (Table 1).
exchanger) (see Table 1).
The pH of the urine samples was set at pH 4.5 with Supported liquid–liquid extraction
100 mM sodium acetate buffer before SPE. These slightly
acidic conditions were applied following previously described Analogous to the SPE optimization, the effect of the pH (pH
methods for the extraction of benzodiazepines [14–16]. 6–9) was investigated for SLE as well. The best results were

Table 2 Limits of quantification (LOQ), limits of detection (LOD), correlation coefficients, equations, retention times (RT), and mass-
spectrometric detection

LOQ LOD Qualifier Quantifier ion equation R Absolute R.T.

[ng/ml] [ng/ml] m/z (min:sec)

2-OH-ethylflurazepam 10 1.2 full spectra 288 y ¼ þ0:034037x þ 0:05693 0.9988 7:46.04

3-OH-flunitrazepam 50 12.2 full spectra 372 y ¼ þ0:004559x  0:05436 0.9991 7:59.20
4-OH-midazolam 10 2.4 full spectra 398 y ¼ þ0:005195x þ 0:00803 0.9986 7:53.28
7-aminoclonazepam 10 2.1 full spectra 429 y ¼ þ0:014323x þ 0:03135 0.9979 7:41.40
7-aminoflunitrazepam 10 3.6 full spectra 355 y ¼ þ0:019991x þ 0:04198 0.9966 7:52.22
7-aminonitrazepam 10 3.7 full spectra 394 y ¼ þ0:031137x þ 0:09249 0.9985 7:21.88
alprazolam 10 2.8 full spectra 279 y ¼ þ0:006540x þ 0:10068 0.9972 8:36.70
bromazepam 10 2.1 full spectra 388 y ¼ þ0:016748x þ 0:05772 0.9976 7:10.04
chlordiazepoxide 10 2.5 full spectra 283 y ¼ þ0:004852x þ 0:06688 0.9994 7:28.32
clonazepam 50 15.3 full spectra 387 y ¼ þ0:005648x  0:05233 0.9984 7:45.58
delorazepam 10 2.1 full spectra 376 y ¼ þ0:014378x  0:01875 0.9997 7:04.64
demoxepam 10 0.4 full spectra 341 y ¼ þ0:114371x þ 0:93357 0.9982 6:41.96
desmethylflunitrazepam 25 7.5 full spectra 352 y ¼ þ0:001632x þ 0:05931 0.9999 7:20.40
diazepam 10 0.8 full spectra 256 y ¼ þ0:033850x þ 0:30912 0.9986 7:14.84
estazolam 10 1.8 full spectra 259 y ¼ þ0:004540x þ 0:10319 0.9986 8:31.22
flunitrazepam 10 7.3 full spectra 286 y ¼ þ0:001260x  0:03106 0.9998 7:43.16
halazepam 10 0.8 full spectra 324 y ¼ þ0:031815x þ 0:32961 0.9989 6:43.28
lorazepam 10 0.6 full spectra 429 y ¼ þ0:038379x þ 0:17970 0.9980 7:13.24
lormetazepam 10 1.1 full spectra 377 y ¼ þ0:024828x þ 0:02265 0.9994 7:48.32
medazepam 10 0.9 full spectra 242 y ¼ þ0:039248x þ 0:89050 0.9964 6:39.88
midazolam 10 1.4 full spectra 310 y ¼ þ0:028269x þ 0:66425 0.9993 7:39.80
N-ethylnordazepam 10 2.5 full spectra 270 y ¼ þ0:035047x þ 0:12251 0.9978 7:16.20
N-ethyloxazepam 10 0.7 full spectra 357 y ¼ þ0:032907x þ 0:18623 0.9966 7:35.28
nitrazepam 10 3.7 full spectra 352 y ¼ þ0:005171x  0:02806 0.9983 7:25.44
norchlordiazepoxide 10 3.8 full spectra 340 y ¼ þ0:003650x þ 0:06168 0.9972 7:17.38
oxazepam 10 0.5 full spectra 429 y ¼ þ0:040596x þ 0:13229 0.9994 6:55.18
pinazepam 50 10.7 full spectra 280 y ¼ þ0:014867x þ 0:14295 0.9998 7:27.38
prazepam 10 1.1 full spectra 269 y ¼ þ0:026186x  0:01904 0.9971 7:48.56
temazepam 10 1.3 full spectra 343 y ¼ þ0:032433x þ 0:04022 0.9989 7:36.40
triazolam 25 6.2 full spectra 313 y ¼ þ0:006872x þ 0:01547 0.9998 8:52.74
alpha-OH-alprazolam 10 1.3 full spectra 381 y ¼ þ0:012900x þ 0:03946 0.9975 8:44.38
alpha-OH-midazolam 10 4.4 full spectra 310 y ¼ þ0:032365x þ 0:10273 0.9986 7:59.86
alpha-OH-triazolam 10 1.1 full spectra 415 y ¼ þ0:007782x  0:01566 0.9989 9:04.68
Table 3 Intra-assay and interassay precision and accuracy data

10 ng/ml 50 ng/ml 200 ng/ml

Precision (SD and Accuracy (n06, %) Precision (SD and Accuracy Precision (SD and Accuracy Precision (SD and Accuracy Precision (SD and Accuracy
RSD (%), n06) RSD (%), n06) (n06, %) RSD (%), n06) (n06, %) RSD (%), n012) (n012, %) RSD (%), n03) (n06, %)

Day A Day B Day A+B

SD RSD SD RSD SD RSD SD RSD SD RSD

2-OH-ethylflurazepam 0.43 4.03 105.68 0.52 1.04 99.79 0.88 1.72 101.86 0.87 1.73 100.83 0.22 0.45 99.13
3-OH-flunitrazepam < LOQ < LOQ < LOQ 0.40 0.81 99.70 1.21 2.36 102.28 1.09 2.16 100.99 0.44 0.89 99.41
4-OHmidazolam 0.40 4.55 87.80 0.51 1.01 100.42 0.77 1.56 99.63 0.66 1.31 100.03 0.75 1.50 100.28
7-aminoclonazepam 0.45 4.26 105.00 0.26 0.53 99.42 1.30 2.75 94.72 1.52 3.13 97.07 0.17 0.35 99.18
7-aminoflunitrazepam 0.41 3.69 109.72 0.40 0.80 100.23 0.57 1.15 99.13 0.55 1.10 99.68 0.23 0.47 99.89
7-aminonitrazepam 0.37 3.52 105.70 0.48 0.95 100.32 1.27 2.57 98.47 1.03 2.08 99.40 0.50 1.00 100.49
Rapid detection and quantification of 35 benzodiazepines

alprazolam 0.74 7.04 105.37 0.60 1.21 99.84 0.49 0.97 100.41 0.54 1.08 100.13 0.51 1.03 99.38
bromazepam 0.82 7.75 106.35 0.38 0.76 100.58 0.42 0.85 99.59 0.46 0.93 100.09 0.20 0.40 100.57
chlordiazepoxide 0.49 4.66 105.27 0.40 0.80 100.47 0.52 1.03 100.17 0.45 0.89 100.32 0.35 0.70 100.59
clonazepam < LOQ < LOQ < LOQ 0.50 1.01 99.69 1.41 2.61 108.33 2.47 4.75 104.01 0.62 1.25 100.02
delorazepam 0.33 3.24 100.93 0.47 0.94 99.99 0.56 1.12 99.70 0.50 1.00 99.85 0.40 0.80 99.31
demoxepam 0.29 2.66 107.87 0.36 0.71 100.08 0.32 0.64 99.80 0.33 0.66 99.94 0.53 1.06 100.16
desmethylflunitrazepam < LOQ < LOQ < LOQ 0.66 1.34 99.48 0.43 0.87 99.40 0.53 1.07 99.44 0.92 1.86 99.30
diazepam 0.87 8.96 97.45 0.33 0.67 99.87 0.23 0.46 100.49 0.32 0.64 100.18 0.40 0.80 99.82
estazolam 0.66 6.65 98.52 0.55 1.11 99.65 1.09 2.17 100.14 0.83 1.66 99.89 0.31 0.62 98.79
flunitrazepam < LOQ < LOQ < LOQ 0.51 1.02 99.78 1.77 3.56 99.20 1.25 2.51 99.49 0.68 1.37 100.13
halazepam 0.47 4.60 103.12 0.53 1.06 100.60 0.31 0.62 100.04 0.44 0.88 100.32 0.49 0.96 100.98
lorazepam 0.34 3.37 101.92 0.42 0.84 99.78 0.31 0.63 100.02 0.36 0.72 99.90 0.13 0.26
lormetazepam 0.70 7.15 97.58 0.29 0.57 100.43 0.47 0.93 100.12 0.38 0.75 100.27 0.31 0.62 100.53
medazepam 0.78 8.34 94.03 0.37 0.74 100.08 1.04 2.14 97.48 1.01 2.05 98.78 0.52 1.04 100.28
midazolam 0.67 6.51 102.23 0.30 0.61 100.65 0.32 0.64 100.16 0.32 0.65 100.41 0.20 0.39 100.76
N-ethylnordazepam 0.57 5.77 98.72 0.32 0.64 99.50 0.28 0.56 99.47 0.28 0.57 99.49 0.23 0.45 99.60
N-ethyloxazepam 0.97 9.68 99.83 0.58 1.15 100.45 0.44 0.88 100.75 0.50 0.99 100.60 0.50 0.98 100.75
nitrazepam 0.72 6.65 107.97 0.51 1.03 99.74 0.27 0.54 99.92 0.39 0.79 99.83 0.57 1.13 100.07
norchlordiazepoxide 0.60 5.57 108.30 0.23 0.47 99.87 0.87 1.75 99.49 0.62 1.23 99.68 0.17 0.35 99.63
oxazepam 0.37 3.49 105.20 0.26 0.51 100.71 0.47 0.93 100.13 0.39 0.78 100.42 0.36 0.71 100.86
pinazepam < LOQ < LOQ < LOQ 0.43 0.86 99.99 0.67 1.33 100.27 0.54 1.08 100.13 0.23 0.47 100.31
prazepam 0.43 4.07 106.15 0.33 0.65 100.04 0.46 0.92 99.86 0.38 0.76 99.95 0.18 0.36 99.64
temazepam 0.25 2.49 99.82 0.40 0.79 100.19 0.34 0.68 100.77 0.38 0.76 100.48 0.53 1.05 100.36
triazolam < LOQ < LOQ < LOQ 0.65 1.29 99.91 0.54 1.07 100.22 0.57 1.14 100.07 0.95 1.90 99.79
763
764 K. Arnhard et al.

Accuracy (n06, %) Precision (SD and Accuracy Precision (SD and Accuracy Precision (SD and Accuracy Precision (SD and Accuracy
(n06, %)
obtained at pH 9, therefore in line with those of previously

100.30
99.99

99.51
described methods [17–19] (see Table S1). Butyl acetate
was better than ethyl acetate and dichloromethane/2-prop-
anol for elution of the benzodiazepine .
(n06, %) RSD (%), n012) (n012, %) RSD (%), n03)

RSD

0.69
0.40
0.94
Furthermore, after method optimization, three different
200ng/ml

SLE column types were compared: Isolute SLE+(Biotage),

Isolute HM-N (Biotage), and Extrelut NT 3 (Merck). The

0.34
0.20
0.47
SD best extraction recoveries were achieved using Extrelut NT
3 columns, as shown in Table 1.
If one compares SPE with SLE regarding the extraction
100.79
100.14
100.28
recoveries (urine spiked with all benzodiazepines at a con-
centration of 100 ng/ml each), the recovery for most of the
analytes was higher for SLE (Table 1). However, the recov-
RSD

2.26
0.90
1.42

ery of norchlordiazepoxide and 7-aminonitrazepam was

Day A+B

very low for extraction of the benzodiazepines with SLE.

It is for this reason that the SPE method was employed for
1.14
0.45
0.71
SD

the benzodiazepine screening method presented. In addition,

the solvent consumption for the SLE method of 15 ml per
101.30
100.05
100.88

sample is relatively high. Furthermore, the matrix back-

ground when using SLE for analyte extraction is higher than
for extraction of the benzodiazepines with SPE (see Fig. 1).
(n06, %) RSD (%), n06)

RSD

Nevertheless, SLE is advantageous with respect to the time

3.19
1.15
1.72

needed to perform the extraction and the simpler handling of

the columns, because no vacuum is required.
Day B

1.62
0.58
0.87
SD

Analytical method
100.27
100.23
99.67

The separation of 35 benzodiazepines was achieved using a

VF-DA GC column as shown in Fig. 2. With a run time of
9.5 min, the method presented is an example of today’s fast
RSD (%), n06)

RSD

0.61
0.66
0.77

GC possibilities. Identification of the benzodiazepines was

done with ChromaTOF and our own library for benzodia-
50ng/ml

Day A

zepines. The library was established after each analyte has

0.31
0.33
0.39
SD

been run separately. The most intense ion, one for each
benzodiazepine, was selected for quantification, but identi-
fication was based on full mass spectra. Even if analytes are
not completely separated, ChromaTOF with its integrated
TSD algorithm can separate coeluted analytes because al-
104.93

108.10

SD standard deviation, RSD relative standard deviation

93.47

ready minimal retention time differences are reflected in the

mass spectra (Fig. 3).
The quantification was achieved with six-point to eight-
Precision (SD and

point calibration curves (concentrations of each benzodiaz-

RSD

3.78
8.38
7.15
RSD (%), n06)

epine of 10, 25, 50, 75, 100, 200, 300, and 350 ng/ml) by
plotting the peak area ratio of the internal standard (oxaze-
10ng/ml

pam-d5 at a concentration of 100 ng/ml) against the peak

0.40
0.78
0.77
SD

area ratio of the respective analyte. Owing to the limited

linear range of the TOF mass analyzer, an extended linear
alpha-OH-midazolam
alpha-OH-alprazolam
Table 3 (continued)

range up to 500 ng/ml was not applicable. The previously

alpha-OH-triazolam

described 8 ng/ml calibrator was only used for demonstra-

tion and confirmation of the linear range, but was not used
in the working range (10–350 ng/ml) for quantification.
Linearity for every benzodiazepine is given with correla-
tion coefficients greater than 0.99 in a calibration range
Rapid detection and quantification of 35 benzodiazepines 765

between 50 and 350 ng/ml for pinazepam, clonazepam, and The extraction recoveries of benzodiazepines using
3-hydroxyflunitrazepam, between 25 and 350 ng/ml for Oasis MCX SPE cartridges at 100 ng/ml range between
desmethylflunitrazepam, flunitrazepam, and triazolam, and 35 and 99 %. The respectively low recoveries for fluni-
between 8 and 350 ng/ml for all remaining analytes. The trazepam (35 %) and desmethylflunitrazepam (41 %)
correlation coefficients as well as the LOQs and limits of were probably due to the weak affinity of the extraction
detection are presented in Table 2. Table 3 shows the coef- cartridge for the analytes.
ficients of variation (CVs) obtained for intra-assay and inter-
assay precision values for three concentrations (10, 50, and Analyte identification
200 ng/ml) as well as accuracy data. The CVs are all lower
than 10 % at the lowest LOQ of most of the analytes and are For identification of the analytes, commercial spectral
even below 2 % for the other two concentrations (50 and databases were used, on the one hand, and an in-house
200 ng/ml). The aforementioned accuracy ranges between library especially for benzodiazepines was developed, on
88 and 108 % at the 10 ng/ml level, and ranges between 99 the other hand. If only screening and confirmation of
and 101 % for the 50 and 200 ng/ml level, depending on the benzodiazepines were performed, the MS library, which
analyte. Table 3 demonstrates the precision of the benzodi- was developed in-house, was preferred because of higher
azepine screening method developed. The analysis of the similarities between sample spectra and the library’s refer-
certified reference material (Bio-Rad) confirmed the good ence spectra. Despite the fact that electron ionization mass
precision and accuracy, with a CV of 1.5 % and an accuracy spectra are, because of a universal ionization energy of
of 101 % (n06). Moreover, it has to be kept in mind that 70 eV, quite reproducible, it is not surprising that slightly
only one internal standard (oxazepam-d5) was used for all higher similarities are achieved when comparing sample
analytes. spectra with an MS library which was established for the

Fig. 4 Thermal decomposition of demoxepam and clorazepate to solution of demoxepam (silylated; middle) and a methanolic solution
nordiazepam: selected ion traces of a methanolic solution of nordiaze- clorazepate, also silylated (bottom)
pam, which was silylated (top); selected ion traces of a methanolic
766 K. Arnhard et al.

same system. See Table S2 for a comparison of similarity nordiazepam trimethylsilane (Fig. 4).The most appropriate
matches when using different MS libraries. Furthermore, reason seemed to be a thermal degradation occurring in the
not all benzodiazepines investigated here are referenced in hot injector. A solution to this problem was pursued by
commercial spectral databases, particularly not as deriva- using a cold-injection system with cryofocusing, but the
tives. When screening for unknowns, we applied both effect still occurred. The decomposition could also have
identification approaches. Thus, possible benzodiazepines been caused by the very fast heating after the cold trapping.
present in samples could be reliably assigned with high We assumed that the thermal decomposition was caused by
similarities, and other sample components could be identi- the derivatization reaction. To solve this problem, the ana-
fied with the help of commercial databases: the NIST/EPA/ lytes were derivatized at room temperature overnight. Fur-
NIH Mass Spectral Library contains more than 200,000 thermore, other derivatization possibilities, acylation [N-
entries and the Mass Spectral and GC Data of Drugs, methylbis(trifluoroacetamide)] and a combination of acyla-
Poisons, Pesticides, Pollutants and Their Metabolites data- tion and silylation, were investigated, but again without
base (PMWTox) contains 8,000 entries. success. Additionally, the decomposition could also derive
from the GC capillary. Further experiments need to be per-
Thermal decomposition of demoxepam/clorazepate formed to prove the cause of the decomposition of these two
benzodiazepines.
Quantification of demoxepam and clorazepate was not pos-
sible owing to thermal decomposition of these analytes. This
effect of decomposition for demoxepam was previously Discussion
reported by Essien et al. [20]. When each analyte (metha-
nolic stock solution) was injected separately, the same main Despite recently published excellent LC-MS methods
peak at 6 min 40 s occurred and was identified as and increasing use of LC-MS-based separations [15,

Fig. 5 Extracted ion chromatogram of a patient sample, where three benzodiazepines and seven other drugs/therapeutic drugs could be identified
(sample preparation using Oasis MCX SPE)
Rapid detection and quantification of 35 benzodiazepines 767

16, 21], GC-MS is still the method of choice in drug and completely anonymous. Various benzodiazepines were
abuse testing. The capillary GC separation technique identified and quantified. Furthermore, other drugs/thera-
provides excellent separation efficiency owing to low peutic drugs such as amphetamine, methadone, methaqua-
height equivalent to a theoretical plate values. The 70- lone, codeine, diclofenac, paracetamol, and propoxyphene
eV electron ionization mode is still a universal standard were confirmed in the samples or in certified reference
that allows the use of commercial spectral databases for material (see the extracted ion chromatogram of a patient
unknown screenings. This is an unbeatable advantage sample in Fig. 5).
compared with the electrospray ionization methods used
in LC-MS. However, using GC single quadrupole sys-
tems is always a trade-off between sensitivity (selected Conclusion
ion monitoring) and specificity (full-scan mode). A GC-
TOF-MS system was chosen for the detection and quan- A single method for the screening and quantification of 35
tification of benzodiazepines because of the high data benzodiazepines in urine using full mass spectra to identify
acquisition rate in full-scan mode. That means there is the analytes was developed. The validity of the method was
no scanning of mass ranges, but always a simultaneous shown, following a corporate validation protocol which is
recording of full spectra. The maximum data recording aligned to the ICH guidelines for analytical method valida-
rate is 500 spectra per second, which ensures an optimal tion. The detection and identification were done by applying
description even of very narrow peaks. In addition, this a self-assembled mass spectra library which is supported by
enables ChromaTOF to separate coeluted peaks using a commercially available libraries for unknown screening. No
special algorithm. The high resolving power of GC previously described method (neither GC-MS nor LC-MS)
enables a faster separation of the analytes as shown has the capability to quantify so many targeted benzodiaze-
for the GC method presented (9.5 min). The LOQ pines and screen for unknowns at the same time. With a run
(10 ng/ml) is sufficient in respect of the intended use time of 9.5 min and an LOQ of 10 ng/ml in 1 ml urine, the
(unknown screening and quantification in therapeutically method demonstrates a good combination of speed and
relevant ranges). LC-MS/MS methods are often used for sensitivity. Furthermore, it is planned to extend the method
targeted analysis of therapeutic drugs and drugs of to a general unknown screening for drugs of abuse and
abuse owing to their unmatched sensitivity in multiple therapeutic drugs.
reaction monitoring mode. The main disadvantage of
LC-MS (selected ion monitoring mode and multiple
reaction monitoring mode) is its inability to detect un- References
expected, unknown analytes. This makes the GC-TOF-
MS method a superior and versatile tool for fast and 1. Voß U (2010) Benzodiazepine: Wie groß ist das Abhängigkeitspo-
sensitive unknown screening and quantitative determina- tential? http://www.uni-leipzig.de/~pharm/phfn/AngstVoss.pdf.
tion of benzodiazepines in urine samples. Accessed 28 Dec 2010
2. Woods JH, Katz JL, Winger G (2008) Abuse and therapeutic use of
The optimized SPE sample preparation showed good
benzodiazepines and benzodiazepine-like drugs. http://www.acnp.org/
recovery of all 35 benzodiazepines. In combination with g4/GN401000172/CH168.html. Accessed 7 Oct 2008
just one stable-isotope-labeled internal standard, very 3. Aebi B, Bernhard W (2002) Chimia 56(3):48–52
good accuracy and an extremely high precision were 4. Maurer HH (2004) Clin Chem Lab Med 42(11):1310–1324
5. Karch SB (1998) Drugs of abuse handbook, 1st edn. CRC, Boca
achieved. The SLE method is an excellent alternative to
Raton
SPE owing to it having fewer manual steps, and two 6. Gunnar T, Ariniemi K, Lillsunde P (2005) J Chromatogr 818:175–189
benzodiazepines were discriminated (norchlordiazepoxide 7. Adamowicz P, Kała M (2010) Forensic Sci Int 198(1–3):39–45.
and 7-aminonitrazepam). The SLE method would be the doi:10.1016/j.forsciint.2010.02.012
8. Abdel-Rehim M (2010) LC/MS in Drug Development. Part I:
method of choice to replace old-fashioned liquid–liquid
Historical Perspective and Current Role. Department of Chemistry
separations using separation funnels. However, to support and Biomedical Science, Faculty of Technology and Science,
and confirm the applicability of the SLE method pre- Karlstad University. http://www.j-chrom-sci.com/index.php?
sented, it should also be validated. option0com_content&view0article&id053&Itemid019. Accessed
18 Jan 2011
9. Zhou LZ (2005) In: Ahuja S, Dong MW (eds) Handbook of
pharmaceutical analysis by HPLC, vol 6. Academic, New York,
Practical application pp 499–568
10. Maurer HH (2006) J Mass Spectrom 41(11):1399–1413
11. Miyaguchi H, Kuwayama K, Tsujikawa K, Kanamori T, Iwata YT,
The method described has already been successfully applied
Inoue H, Kishi T (2006) Forensic Sci Int 157(1):57–70
to real patient samples and showed its potential. The sam- 12. Smink BE, Brandsma JE, Dijkhuizen A, Lusthof KJ, de Gier JJ,
ples investigated were leftovers from an unrelated project Egberts ACG, Uges DRA (2004) J Chromatogr B 811(1):13–20
768 K. Arnhard et al.

13. International Conference on Harmonisation of Technical Requirements 17. Hailey DM (1974) J Chromatogr A 98(3):527–568. doi:10.1016/
for Registration of Pharmaceuticals for Human Use (2005) Validation s0021-9673(00)84071-6
of analytical procedures: text and methodology Q2(R1). http://www. 18. Brodie RR, Chasseaud LF, Taylor T (1978) J Chromatogr A 150
ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/ (2):361–366. doi:10.1016/s0021-9673(00)88195-9
Quality/Q2_R1/Step4/Q2_R1__Guideline.pdf. Accessed 31 Jan 2012 19. Drouet-Coassolo C, Aubert C, Coassolo P, Cano J-P (1989) J Chro-
14. Marin SJ, Coles R, Merrell M, McMillin GA (2008) J Anal matogr B: Biomed Sci Appl 487:295–311. doi:10.1016/s0378-4347
Toxicol 32(7):491–498 (00)83039-x
15. Ishida T, Kudo K, Hayashida M, Ikeda N (2009) J Chromatogr B 20. Essien H, Lai SJ, Binder SR, King DL (1996) J Chromatogr B:
Anal Technol Biomed Life Sci 877(25):2652–2657 Biomed Sci Appl 683(2):199–208
16. ElSohly MA, Gul W, Avula B, Murphy TP, Khan IA (2008) J Anal 21. Vogliardi S, Favretto D, Tucci M, Stocchero G, Ferrara SD
Toxicol 32(8):547–561 (2011) Anal Bioanal Chem 400(1):51–67