Accepted Manuscript Title: Pre-analytical and analytical variation of drug determination in segmented hair using ultra-performance liquid chromatography–tandem mass spectrometry Author: Marie Katrine Klose Nielsen Sys Stybe Johansen Kristian Linnet PII: DOI: Reference:
S0379-0738(13)00474-X http://dx.doi.org/doi:10.1016/j.forsciint.2013.10.029 FSI 7398
To appear in:
FSI
Received date: Revised date: Accepted date:
13-7-2013 21-10-2013 22-10-2013
Please cite this article as: M.K.K. Nielsen, S.S. Johansen, K. Linnet, Pre-analytical and analytical variation of drug determination in segmented hair using ultra-performance liquid chromatographyndashtandem mass spectrometry, Forensic Science International (2013), http://dx.doi.org/10.1016/j.forsciint.2013.10.029 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Acknowledgments
Acknowledgement
Ac
ce pt
ed
M
an
us
cr
ip t
We would like to thank Jytte Lundsby Jensen for her skillful technical assistance with the laboratory work.
Page 1 of 22
Cover letter.docx
To the journal Forensic Science International
10. July 2013
ip t
We hereby submit the paper “Pre-analytical and analytical variation of drug determination in segmented hair using ultra-performance liquid chromatography–tandem mass spectrometry” for publication in the journal Forensic Science International.
cr
Sincerely
us
Marie Katrine Klose Nielsen M. Sc. Environmental Chemistry
M
Kristian Linnet Professor, MD, DMSci, Head of Section
an
Sys Stybe Johansen Associate Professor, M. Sc. Pharmacy, PhD.
Ac
ce pt
ed
Section of Forensic Chemistry Department of Forensic Medicine Faculty of Health Sciences, University of Copenhagen Frederik V´s vej11, DK-2100 Copenhagen Ø E-mail: [email protected] TEL: +45 35 32 61 03 FAX: +45 35 32 60 85
Page 2 of 22
Title page.docx
Pre-analytical and analytical variation of drug determination in segmented hair using ultra-performance liquid chromatography–
a
a
cr
a,*
ip t
tandem mass spectrometry
Marie Katrine Klose Nielsen , Sys Stybe Johansen and Kristian Linnet .
Section of Forensic Chemistry, Department of Forensic Medicine, Faculty of Health Sciences, University of
us
a
*
an
Copenhagen, Frederik V´s vej 11, DK-2100, Denmark.
Corresponding author at: Section of Forensic Chemistry, Department of Forensic Medicine, Faculty of Health
Ac
ce pt
ed
E-mail address: [email protected]
M
Sciences, University of Copenhagen, Frederik V´s vej 11, DK-2100, Denmark. Tel.: + 45 3532 6312; fax: +45 3532 6085.
Page 3 of 22
*Manuscript (without author details) Click here to view linked References
Pre-analytical and analytical variation of drug determination in segmented hair using ultra-performance liquid chromatography–
ip t
tandem mass spectrometry Abstract
cr
Assessment of total uncertainty of analytical methods for the measurements of drugs in human hair has mainly
us
been derived from the analytical variation. However, in hair analysis several other sources of uncertainty will contribute to the total uncertainty. Particularly, in segmental hair analysis pre-analytical variations associated with the
an
sampling and segmentation may be significant factors in the assessment of the total uncertainty budget. The aim of this study was to develop and validate a method for the analysis of 31 common drugs in hair using ultra-performance
M
liquid chromatography–tandem mass spectrometry (UHPLC–MS/MS) with focus on the assessment of both the preanalytical and analytical sampling variations. The validated method was specific, accurate (80-120%), and precise (CV ≤ 20%) across a wide linear concentration range from 0.025–25 ng/mg for most compounds. The analytical variation
ed
was estimated to be less than 15% for almost all compounds. The method was successfully applied to 25 segmented hair specimens from deceased drug addicts showing a broad pattern of poly-drug use. The pre-analytical sampling
ce pt
variation was estimated from the genuine duplicate measurements of two bundles of hair collected from each subject after subtraction of the analytical component. For the most frequently detected analytes, the pre-analytical variation was estimated to be 26–69%. Thus, the pre-analytical variation was 3–7 fold larger than the analytical variation (7–
Ac
13%) and hence the dominant component in the total variation (29–70%). The present study demonstrated the importance of including the pre-analytical variation in the assessment of the total uncertainty budget and in the setting of the 95%-uncertainty interval (± 2 CVT). Excluding the pre-analytical sampling variation could significantly affect the interpretation of results from segmental hair analysis.
Keywords: Segmental hair analysis, pre-analytical variation, uncertainty, drugs of abuse, pharmaceuticals, drug addicts
1 Page 4 of 22
1. Introduction Hair analysis has gained increasing importance in recent year for the analysis of drugs in therapy compliance control, drug-facilitated sexual assaults, driving license re-granting, chronic drug abuse intoxication and post-mortem toxicology [1–5]. In contrast to other biological matrices analysis of hair allows retrospective investigation of drug
ip t
consumption and evaluation of historic drug use when using segmental hair. These advantages are offset by shortcomings of external contamination leading to false-positive hair results, as well as low drug concentrations,
cr
which require more sensitive analytical methods for both screening and quantification.
Several multi-target liquid chromatography–tandem mass spectrometry methods (HPLC or UHPLC–MS/MS) have
us
been reported for drug determination in human hair [6–10]. The validation of analytical methods for hair analysis has mainly focused on the evaluation of specificity, sensitivity, trueness, and precision of the target analytes. However, it
an
is also important to account for sources of uncertainty from hair sampling, segmentation and washing procedures, etc. Especially, in segmental hair analysis, the precision and accuracy of the method is highly dependent on both the
M
sampling and segmentation procedure [11]. Biological variation associated with the variability in hair growth, the variations in drug incorporation rates, horizontal washout and deposition by diffusion from sweat and sebum may also
ed
contribute to the total uncertainty [11–15]. Another factor that could contribute to the overall variation is the amount of incorporated drug, which depends on several things such as the melamine content, blood flow, blood
ce pt
concentration, as well as the acid dissociation constant (pK a) and the lipophilicity of the drug [13]. The aim of this study was to develop and validate an ultra-performance liquid chromatography–tandem mass spectrometry (UHPLC–MS/MS) method for the analysis of 31 common drugs in human hair with focus on the
Ac
assessment of both the pre-analytical and analytical variations. For this purpose, we have adapted a recently presented method for blood analysis of 31 common pharmaceuticals and drugs of abuse for hair analysis [16].
2. Materials and methods 2.1 Hair samples
Authentic hair from 25 deceased drug addicts was routinely collected during forensic autopsies by trained forensic technicians. Hair samples of approximately 1 cm in diameter were collected with a scissor from the posterior vertex
2 Page 5 of 22
region as close to the scalp as possible and wrapped in tinfoil. The hair samples were stored at room temperature until analysis. Blank hair samples used for negative controls, spiked calibration standards and quality control samples were obtained from drug-free healthy volunteers. Different types of drug-free hair (blond, dark brown, straight and curly)
ip t
were used for the investigations of matrix interferences and specificity. Quality control samples were prepared by soaking 2 g drug-free hair cut in 2–3 mm pieces with a mixture of
removed with four times of 40 ml MeOH, and the spiked hair was dried overnight.
cr
reference standards (20 mg/L) in DMSO (50%) for 10 days with regular shaking. The standard solutions were briefly
us
An external hair control sample was obtained from Arvecon GmbH (Walldorf, Germany) and a certified proficiency
an
test was obtained from the Society of Hair Testing (SoHT).
2.2 Sample preparation of authentic hair samples
M
Two bundles of hair collected from the same case were analyzed as genuine duplicate measurements. All hair samples were carefully aligned and cut into two segments; S1 (5 mm) and S2 (10 mm), where S1 was the proximal
ed
segment (root). About 10 mg of each segment was weighted into 2 mL hard tissue grinding tubes (Precellys MK28-R, Bertin Technologies, France). The hair samples were washed one time with 2-propanol (1 mL), two times with purified
ce pt
water (1 mL) and finally one time with 2-propanol (1 mL). The mixtures were shaken 5 min with the first three washing solutions and 1 min with the last washing solution. The washing solutions were removed after 3 min centrifugation at 3600 rpm at room temperature. The last washing water was analyzed to test for external contamination. The decontaminated hair samples were dried overnight, weighed, and mixed with six stainless steel beads before addition
Ac
of 500 µL extraction media (25:29:24 (v/v) methanol:acetonitrile:2mM ammonium formate (pH 5.3)) and 25 µL IS solution (0.4 mg/L). The hair samples were pulverized in a Precellys 24 tissue homogenizer (Bertin Technologies, France) for three times, with 2 x 30 sec bursts at 5600 rpm. The pulverized hair samples were extracted in an oven at 37:C for 18 h and filtered with MiniUni Prep filter vials (Whatman Inc., New Jersey) as described by Nielsen et al. [8].
2.3 Liquid chromatography–tandem mass spectrometry The hair extracts were analyzed with an UHPLC–MS/MS method previously developed in our laboratory for routine analysis of whole blood samples with small modifications [16]. Cathine and cathinone were replaced with EDDP and
3 Page 6 of 22
oxycodone. The MRM transitions for EDDP were m/z 278→234 and m/z 278→249 as the quantitative and confirmative traces, respectively. The cone voltage was set to 50 V and the collision energies were set to 34 and 22 eV for MRM 1 and MRM 2, respectively. For oxycodone, MRM 1 and MRM 2 were m/z 316→241 and m/z 316→298. The cone voltage was set to 30 V and the collision energies were set to 28 and 10 eV, respectively.
ip t
The analytes in authentic samples were identified by the retention times and the ion ratio of two characteristic
cr
MRM transitions. Tolerance was set to ±1% for the retention time and ± 30% for the ion ratio.
2.4 Method validation
document EN ISO/IEC 17025:2000) and as described by Peters et al. [17].
us
The method was fully validated according to The Danish Accreditation and Metrology Fund (DANAK, guidance
an
The specificities of the methods were investigated in different hair types by analyzing 10 drug-free hair samples of different origins. The same 10 drug-free hair samples were also spiked at two levels: 0.05 and 0.5 ng/mg, and the
M
recoveries and coefficient of variation (CV%) were calculated for each hair sample. Post-column infusion experiments were performed for 10 selected compounds eluting at different time areas:
ed
amphetamine, codeine, ketamine, ketobemidone, MDMA, methamphetamine, oxycodone, chlordiazepoxide, nitrazepam, zopiclone, and their deuterated internal standards. The compounds were divided into four groups based
ce pt
on their retention time and masses. The compounds were injected continuously into the mass spectrometer postcolumn using a syringe pump and a T-piece connected to the UHPLC-system. The syringe pump delivered a constant flow of 10 µL/min standard solution (0.05 mg/L), and the UHPLC-system delivered a constant flow of 0.4 mL/min mobile phase. The baseline responses were monitored following the injection of 10 µL extracts of nine drug-free hair
Ac
samples from the autosampler. These baseline responses were compared to the baseline responses following injection of a blank mobile phase. The absolute recoveries were determined by comparing the absolute peak area for each analyte in fortified hair at 5 ng/mg with the absolute peak area for each analyte in a pure standard solution at similar level in 16 different series. The linear range was investigated by analyzing matrix-extracted calibrators at seven concentration levels from 0.025 to 25 ng/mg. The analysis was repeated in four different series. The calibration curves were obtained by plotting the peak area of the analytes to IS versus the theoretical concentration. Linear regression with 1/x weighting was 2
used, and the linearity was evaluated from the residual plots and the squared correlation coefficient (r ). The linearity
4 Page 7 of 22
range was defined as the range from the lower limit of quantification (LLOQ) to the highest level where the criteria of both precision (CV ≤ 15%) and accuracy (85–115%) were met. The LLOQ was defined as the lowest concentration where CV was ≤ 20% and accuracy was within 80–120%. The evaluation of precision and accuracy were performed at seven concentration levels: (0.025; 0.050; 0.10; 0.50;
ip t
2.5; 10 and 25 ng/mg). The determinations were based on four replicates performed on four days (N=16). The longterm precision was determined from measurements of low and high quality control samples analyzed in 17 series.
cr
The total uncertainty (CVT) was derived from the individual components of variations [18]:
us
The analytical uncertainty (CVA) was derived as an average value of the long-term precision from low and high quality control samples. The variation of the calibrator (CVcal) was estimated from the uncertainty of the purity of the
an
compound and the uncertainty of the preparation of calibrator solution. The sampling variation (CV) was derived from
ed
relative differences of duplicate measurements:
M
measurements of two bundles of hair collected from each subject on the basis of the standard deviation of the
where rdi was the paired difference between measurements of the ith sample divided by the mean of the two measurements. The pre-analytical variation (CVPA) was derived from the sampling variation and the analytical variation
ce pt
from the following relation:
Ac
A 95%-uncertainty interval for authentic hair samples was estimated as the measured value ± 2 CV T.
3. Results
3.1 Validation of the method
The validation data are summarized in Table 1. The method was found to be selective for all the tested analytes since no interfering peaks were observed in the traces of the analytes or the internal standards when analyzing 10 drug-free hair samples except for morphine where a small peak was detected below LLOQ in one sample. When the same 10 drug-free hair samples were spiked at two different levels the average recoveries ranged between 80 and
5 Page 8 of 22
120% for almost all compounds. The coefficient of variation between the recoveries obtained from 10 different types of hair was less than 15% for the majority of the analytes. The post-column infusion experiment showed some ion suppression or enhancement for all the selected compounds, but the suppression or enhancement was in the same order of magnitude as their deuterated internal
ip t
standards. The only compound which was not corrected by the chosen internal standard was ketobemidone, but since the variations between the different hair types was small, the use of matrix-matched standards will minimize the
cr
effect of suppression from endogenous hair matrix compared to a pure standard solution. A deuterated internal standard for ketobemidone is preferable to correct for suppression, but a deuterated analogue was not available at
us
the time the method was validated.
Interference from co-eluting analytes and from other possible drugs in forensic cases has previously been
an
investigated for the chromatographic system by Bjørk et al. [16]. No significant interferences from co-eluting analytes or from 167 tested drugs such as other benzodiazepines, analgesics, antidepressants, antipsychotics, β-blockers,
M
narcotics, and stimulants contributed to positive blank values in the traces of the target analytes. For all targets the average absolute recovery ranged from 42 to 106%. The recoveries for the internal standard
ed
were of the same order of magnitude as the corresponding analyte. Linearity was observed from 0.025 to 10 or 25 2
ng/mg for all compounds except alprazolam with squared correlation coefficients (r ) greater than 0.994. Deviations
ce pt
from the fitted curve were less than 15% for levels above LLOQ and less than 20% at LLOQ for most compounds. A value for LLOQ of 0.025 ng/mg was obtained for all compounds except for cocaine, codeine, morphine, 6monoacetylmorphine, oxycodone, and nordiazepam, where the obtained LLOQ values were 0.050 ng/mg. Cases with
system.
Ac
values above the upper limit of quantification were diluted with extraction media before injection into the UHPLC-
For all analytes the criteria for precision and accuracy were fulfilled at seven concentration levels except for a few compounds, where CV was slightly above the limit for one concentration level. Chlordiazepoxide, morphine, and oxycodone showed CVs up to 23% at one level, while the accuracy was within 85–115% at all levels. The long-term precisions were below 15% at both levels except for chlordiazepoxide (≤25%), ketamine (≤16%), methamphetamine (≤21%), and oxycodone (≤22%). The accuracy of the method was further evaluated for some analytes by analyzing an external quality control sample from Arvecon GmbH in 15 different series. The average accuracies ranged between 83
6 Page 9 of 22
and 105% for amphetamine, benzoylecgonine, cocaine, codeine, EDDP, methadone, methamphetamine, 6monoacetylmorphine, and morphine. The accuracy of the method was also tested for some analytes (amphetamine, benzoylecgonine, cocaine, codeine, MDA, MDMA, methadone, methamphetamine, 6-monoacetylmorphine, morphine and tramadol) by participating in a
ip t
quality proficiency program provided by the Society of Hair Testing (SoHT). All the analytes passed the test with zscores less than 2 (z = (observed value-median of participants)/SD).
cr
The method was successfully applied to 25 authentic hair specimens from deceased drug addicts. Thirty of the target compounds were detected in one or more cases and several drug groups were found in each case, indicating a
us
broad poly-drug use. Most of the drugs were found in both hair segments and there were no significant differences
an
between the mean concentration levels in the two segments according to a Wilcoxon’s rank test (P > 0.05).
3.2 Evaluation of components of variations
M
The components of variation for all the positive findings and the most frequently detected analytes are summarized in Table 2. The most frequently detected analytes accounted for 50% of the positive findings. The
ed
distribution of variation was similar in the two segments and there was no significant difference between the relative differences in the two segments (Wilcoxon’s rank test, P > 0.05). As seen from the percentage differences plots of the
ce pt
most frequently occurring drugs methadone and cocaine, the relative differences were evenly distributed around the zero line in both hair segments (Fig. 1 and Fig. 2).
The analytical coefficient of variation was less than 13% for the frequently detected analytes. As opposed to the analytical variability, the pre-analytical variation and thus the total uncertainty depended on the analytes. The pre-
Ac
analytical variation ranged from 41 to 69% for most of the compounds resulting in total uncertainties of 42–70%. For the benzodiazepines bromazepam and diazepam, the pre-analytical variation values were slightly lower, with average values of 27 and 38%, respectively. The total uncertainties for these two compounds were 29 and 40%, respectively. The pre-analytical variation was in general 3–7 fold larger than the analytical variation and thus the most essential component of variation due to the principle of squared addition. The calibrator variations were estimated to be 0.5– 5% for the frequently detected analytes and constituted only a small part of the total uncertainty budget. The 95%-uncertainty interval was calculated for the median values for the frequently detected analytes with and without the pre-analytical variation included in the estimation of the total uncertainty. If the pre-analytical variation
7 Page 10 of 22
was not included in the estimation of the total uncertainty, the 95%-uncertainty interval was underestimated with a factor of 3–7 depending on the analyte. For example, for a diazepam result of 0.17 ng/mg, the estimated 95%uncertainty interval ranged from 0.13 to 0.21 ng/mg if the total uncertainty was only derived from the analytical and calibrator variations, while the estimated 95%-uncertainty interval ranged from 0.034 to 0.31 ng/mg if the pre-
ip t
analytical variation was taken into account. Thus, for diazepam the result was underestimated by a factor 3 without the contribution of pre-analytical variation.
cr
Although the components of variation depended on the analyte, the uncertainties for all the positive findings were also evaluated. An average analytical variation of all the analytes included in the method was estimated to be 10%
us
based on the long-term precisions. The total uncertainty for 203 positive determinations was 48–52% and as for the
an
individual analytes the pre-analytical variation accounted for the main part of the total uncertainty.
4. Discussion
M
The validated method was shown to be specific, accurate, and precise across a wide linear range of concentrations as a result of using deuterated analogues as internal standards and by using matrix-matched standards for calibration.
ed
The validation results are similar to the validation results obtained with an ultra-performance liquid chromatography– time-of-flight mass spectrometry (UHPLC–TOF-MS) method for hair analysis previously developed in our laboratory
ce pt
[8]. The former method was able to quantify 52 common pharmaceuticals and drugs of abuse in hair simultaneously, but the sensitivity of amphetamine, methamphetamine, MDA, MDMA, and morphine did not fulfill the cut-off values recommended by the SoHT for quantitative methods [11]. The sensitivity of the present method exceeds that of the
Ac
former method by a factor of 10 or more for these compounds and thus, the established LLOQ values for the common drugs of abuse were all lower than the recommended SoHT cut-off values. Both the sensitivity and the wide linear concentrations range were found suitable for the determination of the targets analytes in hair samples from chronic drug users. If the method should be used for measurement of a single-dose, an up concentration step could easily be implemented to improve the sensitivity. The high pre-analytical variation may be associated with inconsistent sampling and segmentation of the hair samples [11,19]. It is normally assumed that the hair is collected directly next to the scalp, but this is difficult to achieve in practice, especially when the hair is oily or filtered as in autopsy cases. It has previously been demonstrated
8 Page 11 of 22
that between 0.1 and 3.0 cm of hair with an average of 0.8 ± 0.1 cm remained on the scalp after collecting of the hair samples, depending on the sampling procedure and the length of the hair [19]. This large uncertainty associated with insufficient sampling could significantly affect the total uncertainty of the method especially in segmental hair analysis.
ip t
Besides the experimental uncertainties, the biological variation associated with variability in hair growth rate, drug incorporation rate, hair pigmentation, and blood supply to the hair follicle may also contribute to the total variation
cr
[12–15]. The hair grows in three phases and at any time a substantial proportion of hair will not be in the growing phase, leading to irregular hair growth [14]. The hair growth rate is most homogenous on the head at the posterior
us
vertex region, where approximately 15% of the hairs are in the resting phase and the remaining 85% are in their growing phase [20]. It is therefore recommended that the sample is collected from the posterior vertex region as close
an
to the scalp as possible to minimize the uncertainty associated with variable hair growth [11]. Although, the hair
most predominant component of the total variation.
M
samples used in this study were collected according to the recommendations, the pre-analytical variation was the
The intra-individual variability of incorporated drug due to pigmentation has been investigated in white and grey
ed
hair collected from the same donor [21,22]. In two cases, the results showed that the zolpidem concentration was more than 100-fold higher in the black hair compared to the white hair. These findings are consistent with other
ce pt
pigmentation studies where levels of amphetamine, cocaine, and methamphetamine were found to be significantly higher in pigmented hair than in white hair [23,24]. Thus, persons with plain pigmented hair would have more consistent incorporation of drug than persons with grey hair. Also inter-individual variability in the amount of incorporated drug has been evaluated by administration of single-dosages of benzodiazepines to different volunteers
Ac
followed by segmental hair analysis [21,25–27]. The detected levels of benzodiazepines in the segment representing the exposure period were similar in Chinese subjects with black hair, while the levels deviated more in subjects with different hair types.
Reports on variations between replicates of authentic hair results are sparse. In two studies, the relative deviations were less than 20% for most of the detected analytes [8,28]. The replicates were not genuine determinations of hair samples collected at different sites from the same subject, but replicate measurements of the same sample of finely cut hair, thus excluding the pre-analytical variation. The obtained relative deviations in these two studies are
9 Page 12 of 22
consistent with the analytical coefficients of variation assessed in this study, which were based on replicate measurements of homogenous cut hair.
5. Conclusion
ip t
A specific and reproducible multi-target UHPLC–MS/MS method has been developed for the determination of 31 common pharmaceuticals and drugs of abuse in human hair. The method was fully validated and successfully applied
cr
for the quantification of pharmaceuticals and drugs of abuse in segmented hair samples from 25 autopsy cases. The total uncertainty was estimated to approximately 29–70% from genuine duplicate measurements of hair samples
us
depending on the analytes. The pre-analytical variation was 3–7 fold larger than the analytical variation and thus the dominant component of variation. The study demonstrated the importance of including the pre-analytical variation
an
when evaluating the total uncertainty and the width of the 95%-uncertainty interval for authentic hair samples. Due to this large uncertainty associated with hair analysis, precautions should be taken when evaluating quantitative results
Ac
ce pt
ed
M
in authentic hair and when comparing hair results within and between subjects.
10 Page 13 of 22
Figure legends Fig. 1. Plots of percentage differences between methadone concentrations in two bundles of hair in segment 1 (top)
ip t
and segment 2 (bottom). The long-dashed line indicates the averages differences, and the short-dashed line indicates the 95%-uncertainty interval (± 2 CVT).
cr
Fig. 2. Plots of percentage differences between cocaine concentrations in two bundles of hair in segment 1 (top) and segment 2 (bottom). The long-dashed line indicates the averages differences, and the short-dashed line indicates the
Ac
ce pt
ed
M
an
us
95%-uncertainty interval (± 2 CVT).
11 Page 14 of 22
References
ip t
[1] F. Musshoff, K. Lachenmeier, J. Trafkowski, B. Madea, F. Nauck, U. Stamer, Determination of opioid analgesics in hair samples using liquid chromatography/tandem mass spectrometry and application to patients under
cr
palliative care, Ther. Drug. Monit. 29 (2007) 655–661.
us
[2] M. Chèze, G. Duffort, M. Deveaux, G. Pepin, Hair analysis by liquid chromatography–tandem mass spectrometry in toxicological investigation of drug-facilitated crimes: Report of 128 cases over the period
an
June 2003 May 2004 in metropolitan Paris, Forensic Sci. Int. 153 (2005) 3–10.
[3] P. Kintz, A. Tracqui, P. Mangin, Detection of Drugs in Human Hair for Clinical and Forensic Applications, Int. J.
M
Legal Med. 105 (1992) 1–4.
ed
[4] R. Kronstrand, I. Nystrom, M. Forsman, K. Kall, Hair analysis for drugs in driver's license regranting. A Swedish pilot study, Forensic Sci. Int. 196 (2010) 55–58.
ce pt
[5] F. Musshoff, K. Lachenmeier, H. Wollersen, D. Lichtermann, B. Madea, Opiate concentrations in hair from subjects in a controlled heroin-maintenance program and from opiate-associated fatalities, J. Anal. Toxicol. 29 (2005) 345–352.
Ac
[6] M. Laloup, M.D.R. Fernandez, G. De Boeck, M. Wood, V. Maes, N. Samyn, Validation of a liquid chromatography–tandem mass spectrometry method for the simultaneous determination of 26 benzodiazepines and metabolites, zolpidem and zopiclone, in blood, urine, and hair, J. Anal. Toxicol. 29 (2005) 616–626.
[7] J. Kim, S. Lee, S. In, H. Choi, H. Chung, Validation of a simultaneous analytical method for the detection of 27 benzodiazepines and metabolites and zolpidem in hair using LC–MS/MS and its application to human and rat hair, J. Chromatogr. B 879 (2011) 878–886.
12 Page 15 of 22
[8] M.K. Nielsen, S.S. Johansen, P.W. Dalsgaard, K. Linnet, Simultaneous screening and quantification of 52 common pharmaceuticals and drugs of abuse in hair using UPLC–TOF-MS, Forensic Sci. Int. 196 (2010) 85–92.
[9] E. Lendoiro, Ó. Quintela, A. de Castro, A. Cruz, M. Ópez-Rivadulla, M. Concheiro, Target screening and confirmation of 35 licit and illicit drugs and metabolites in hair by LC–MSMS, Forensic Sci. Int. 217 (2012)
ip t
207–215.
cr
[10] K.Y. Rust, M.R. Baumgartner, N. Meggiolaro, T. Kraemer, Detection and validated quantification of 21
us
benzodiazepines and 3 “z-drugs” in human hair by LC–MS/MS, Forensic Sci. Int. 215 (2012) 64–72.
[11] G.A.A. Cooper, R. Kronstrand, P. Kintz, Society of Hair Testing guidelines for drug testing in hair, Forensic Sci.
an
Int. 218 (2012) 20–24.
M
[12] E.J. Cone, Mechanisms of drug incorporation into hair, Ther. Drug. Monit. 18 (1996) 438-443.
[13] F. Pragst, M.A. Balikova, State of the art in hair analysis for detection of drug and alcohol abuse, Clin. Chim.
ed
Acta. 370 (2006) 17–49.
[14] H. Sachs, Theoretical limits of the evaluation of drug concentrations in hair due to irregular hair growth,
ce pt
Forensic Sci. Int. 70 (1995) 53–61.
[15] G.L. Henderson, M.R. Harkey, C.H. Zhou, R.T. Jones, P. Jacob, Incorporation of isotopically labeled cocaine and metabolites into human hair: 1. Dose-response relationships, J. Anal. Toxicol. 20 (1996) 1–12.
Ac
[16] M. Bjørk, K. Simonsen, D. Andersen, P. Dalsgaard, S. Sigurôardóttir, K. Linnet, B. Rasmussen, Quantification of 31 illicit and medicinal drugs and metabolites in whole blood by fully automated solid-phase extraction and ultra-performance liquid chromatography–tandem mass spectrometry, Anal. Bioanal. Chem. 405 (2013) 2607–2617.
[17] F.T. Peters, O.H. Drummer, F. Musshoff, Validation of new methods, Forensic Sci. Int. 165 (2007) 216–224.
13 Page 16 of 22
[18] K. Linnet, J.C. Boyd, Selection and analytical evaluation of methods-with statistical techniques, in: C.A. Burtis, th
E.R. Ashwood, D.E. Bruns (Eds), Tietz textbook of clinical chemistry and molecular diagnostics, 5
ed.,
Elsevier/Saunders, St. Louis, 2012, pp. 7–48.
[19] M.A. LeBeau, M.A. Montgomery, J.D. Brewer, The role of variations in growth rate and sample collection on
cr
[20] M.R. Harkey, Anatomy and physiology of hair, Forensic Sci. Int. 63 (1993) 9–18.
ip t
interpreting results of segmental analyses of hair, Forensic Sci. Int. 210 (2011) 110–116.
us
[21] M. Villain, M. Chèze, A. Tracqui, B. Ludes, P. Kintz, Windows of detection of zolpidem in urine and hair: application to two drug facilitated sexual assaults, Forensic Sci. Int. 143 (2004) 157–161.
an
[22] H. Miyaguchi, Determination of zolpidem in human hair by micropulverized extraction based on the evaluation of relative extraction efficiency of seven psychoactive drugs from an incurred human hair
M
specimen, J. Chromatogr. A 1293 (2013) 28–35.
ed
[23] R. Kronstrand, J. Ahlner, N. Dizdar, G. Larson, Quantitative analysis of desmethylselegiline, methamphetamine, and amphetamine in hair and plasma from Parkinson patients on long-term selegiline
ce pt
medication, J. Anal. Toxicol. 27 (2003) 135–141.
[24] R.W. Reid, F.L. OConnor, A.G. Deakin, D.M. Ivery, J.W. Crayton, Cocaine and metabolites in human graying hair: Pigmentary relationship, J. Toxicol-Clin. Toxic. 34 (1996) 685–690.
Ac
[25] P. Xiang, Q.R. Sun, B.H. Shen, P. Chen, W. Liu, M. Shen, Segmental hair analysis using liquid chromatography– tandem mass spectrometry after a single dose of benzodiazepines, Forensic Sci. Int. 204 (2011) 19–26.
[26] M. Villain, M. Chèze, A. Tracqui, B. Ludes, P. Kintz, Testing for zopiclone in hair application to drug-facilitated crimes, Forensic Sci. Int. 145 (2004) 117–121.
[27] A. Negrusz, C.M. Moore, K.B. Hinkel, T.L. Stockham, M. Verma, M.J. Strong, P.G. Janicak, Deposition of 7aminoflunitrazepam and flunitrazepam in hair after a single dose of Rohypnol (R), J. Forensic Sci. 46 (2001) 1143–1151.
14 Page 17 of 22
[28] H. Miyaguchi, H. Takahashi, T. Ohashi, K. Mawatari, Y.T. Iwata, H. Inoue, T. Kitamori, Rapid analysis of methamphetamine in hair by micropulverized extraction and microchip-based competitive ELISA, Forensic
Ac
ce pt
ed
M
an
us
cr
ip t
Sci. Int. 184 (2009) 1–5.
15 Page 18 of 22
Ac ce p
te
d
M
an
us
cr
ip t
Figure 1 methadone.eps
Page 19 of 22
Ac ce p
te
d
M
an
us
cr
ip t
Figure 2 cocaine.eps
Page 20 of 22
Table 1. Validation data. Linear range
LLOQ
ng/mg
Accuracy %
Levels above LLOQ CV %
Accuracy %
CV %
QC1
QC2
us
Analyte
cr
ip t
Table
ng/mg
CV %
ng/mg
CV %
Ac c
ep te
d
M
an
Alprazolam 0.025-2.5 110 13 94–114 10–14 1.2 6 3.5 12 Amphetamine 0.025-25 106 9 100–108 8–14 0.59 7 1.5 11 Benzoylecgonine 0.025-25 94 9 97–105 4–9 0.83 8 1.5 5 Bromazepam 0.025-25 96 12 97–114 6–8 1.7 10 5.1 10 Chlordiazepoxide 0.025-10 99 22 101–106 10–18 0.371 16 2.9 25 Clonazepam 0.025-10 95 15 94–106 4–12 1.4 5 4.7 9 7-Aminoclonazepam 0.025-25 98 7 94–102 5–10 0.55 5 1.8 5 Cocaine 0.050-10 102 6 96–104 4–11 1.8 10 0.74 11 Codeine 0.050-10 98 14 92–110 5–15 0.11 11 0.771 11 Diazepam 0.025-10 91 17 96–104 4–14 2.1 14 6.9 11 Nordiazepam 0.050-10 95 13 102–113 12–14 2.4 10 6.9 10 EDDP 0.025-25 105 17 93–104 3–17 0.30 5 1.1 7 Flunitrazepam 0.025-25 90 18 96–106 5–10 1.2 7 4.2 5 7-Aminoflunitrazepam 0.025-25 98 5 95–103 4–10 0.58 3 1.8 4 Ketamine 0.025-10 93 15 97–115 8–14 0.33 9 0.95 16 Ketobemidone 0.025-10 94 9 88–110 5–12 0.19 14 0.62 15 MDA 0.025-25 84 18 93–100 3–10 0.77 7 2.2 8 MDMA 0.025-10 100 19 96–109 5–19 0.73 11 2.11 13 Methamphetamine 0.025-10 93 16 89–108 5–13 0.54 14 1.41 21 Methadone 0.025-25 93 12 91–101 3–8 0.60 9 2.2 7 Morphine 0.050-25 97 21 98–113 5–11 0.077 11 0.62 6 6-Monoacetylmorphine 0.050-25 93 7 96–110 5–10 0.067 9 0.59 7 Nitrazepam 0.025-10 93 12 94–106 4–12 1.5 6 4.9 8 7-Aminonitrazepam 0.025-25 90 8 97–106 4–9 0.75 4 2.3 5 Oxazepam 0.025-25 98 9 101–113 5–13 1.7 10 5.6 9 Oxycodone 0.050-25 98 18 87–108 5–23 0.093 22 0.71 17 Temazepam 0.025-10 89 13 100–108 5–10 1.9 8 6.0 11 Tramadol 0.025-25 106 8 85–102 2–16 0.14 9 0.46 11 O-desmethyltramadol 0.025-25 95 9 95–101 5–8 0.073 6 0.51 6 Zolpidem 0.025-10 92 12 96–107 7–12 0.68 5 2.2 7 Zopiclone 0.025-25 97 17 91–104 7–16 0.37 11 1.7 12 1 One outlier removed. 2Mean recovery (RE) and coefficient of variation (CV) obtained from 10 different drug-free hair samples spiked at 0.05 ng/mg.
Absolute recovery
Recovery2
%
RE %
CV %
86 78 82 43 56 58 70 67 79 42 55 92 70 85 80 102 73 78 74 57 64 84 59 74 60 79 64 101 106 95 61
87 123 107 118 118 101 104 111 146 96 89 102 102 106 112 97 101 113 104 103 105 101 117 110 112 96 83 108 105 120 76
6 11 4 10 12 9 5 3 10 7 12 2 4 4 12 10 4 7 15 13 15 7 5 4 10 17 6 6 3 8 23
Page 21 of 22
Table
Table 2. Components of variation for frequently detected analytes in 25 autopsy cases. Segment
N
Median ng/mg
Range ng/mg
CVPA %
CVA %
CVcal %
CVT %
All positive findings
S1 S2
203 212
0.45 0.48
0.025–67 0.028–55
511 471
101
21
521 481
Amphetamine
S1 S2
8 7
0.27 0.48
0.029–2.0 0.043–2.5
67 69
9
0.6
67 70
Bromazepam
S1 S2
10 9
0.69 0.70
0.089–2.0 0.060–2.6
26 27
10
5
29 29
Clonazepam
S1 S2
6 6
0.31 0.26
0.098–0.93 0.043–0.98
52 53
7
2
53 53
Cocaine
S1 S2
11 11
1.9 4.3
0.083–67 0.14–37
51 60
11
2
53 61
Codeine
S1 S2
8 7
0.60 0.28
0.092–7.9 0.071–7.9
59 42
11
0.5
60 43
Diazepam
S1 S2
9 9
0.17 0.12
0.025–2.1 0.040–2.0
38 37
13
2
40 39
Methadone
S1 S2
24 24
9.0 8.6
0.33–65 0.24–55
47 41
2
48 42
Morphine
S1 S2
9 9
0.23 0.18
0.31–3.3 0.097–2.6
Tramadol
S1 S2
6 6
2.2 1.6
0.064–23 0.046–17
Zopiclone
S1 S2
8 8
0.55 0.28
0.039–3.1 0.61–2.4
cr
us
8
64 54
8
5
64 55
65 44
10
0.6
66 45
49 51
12
2
51 52
an
1
ip t
Compound
Ac
ce pt
ed
M
An average coefficient of variation was calculated for all the analytes included in the method.
Page 22 of 22
S0379-0738(13)00474-X http://dx.doi.org/doi:10.1016/j.forsciint.2013.10.029 FSI 7398
To appear in:
FSI
Received date: Revised date: Accepted date:
13-7-2013 21-10-2013 22-10-2013
Please cite this article as: M.K.K. Nielsen, S.S. Johansen, K. Linnet, Pre-analytical and analytical variation of drug determination in segmented hair using ultra-performance liquid chromatographyndashtandem mass spectrometry, Forensic Science International (2013), http://dx.doi.org/10.1016/j.forsciint.2013.10.029 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Acknowledgments
Acknowledgement
Ac
ce pt
ed
M
an
us
cr
ip t
We would like to thank Jytte Lundsby Jensen for her skillful technical assistance with the laboratory work.
Page 1 of 22
Cover letter.docx
To the journal Forensic Science International
10. July 2013
ip t
We hereby submit the paper “Pre-analytical and analytical variation of drug determination in segmented hair using ultra-performance liquid chromatography–tandem mass spectrometry” for publication in the journal Forensic Science International.
cr
Sincerely
us
Marie Katrine Klose Nielsen M. Sc. Environmental Chemistry
M
Kristian Linnet Professor, MD, DMSci, Head of Section
an
Sys Stybe Johansen Associate Professor, M. Sc. Pharmacy, PhD.
Ac
ce pt
ed
Section of Forensic Chemistry Department of Forensic Medicine Faculty of Health Sciences, University of Copenhagen Frederik V´s vej11, DK-2100 Copenhagen Ø E-mail: [email protected] TEL: +45 35 32 61 03 FAX: +45 35 32 60 85
Page 2 of 22
Title page.docx
Pre-analytical and analytical variation of drug determination in segmented hair using ultra-performance liquid chromatography–
a
a
cr
a,*
ip t
tandem mass spectrometry
Marie Katrine Klose Nielsen , Sys Stybe Johansen and Kristian Linnet .
Section of Forensic Chemistry, Department of Forensic Medicine, Faculty of Health Sciences, University of
us
a
*
an
Copenhagen, Frederik V´s vej 11, DK-2100, Denmark.
Corresponding author at: Section of Forensic Chemistry, Department of Forensic Medicine, Faculty of Health
Ac
ce pt
ed
E-mail address: [email protected]
M
Sciences, University of Copenhagen, Frederik V´s vej 11, DK-2100, Denmark. Tel.: + 45 3532 6312; fax: +45 3532 6085.
Page 3 of 22
*Manuscript (without author details) Click here to view linked References
Pre-analytical and analytical variation of drug determination in segmented hair using ultra-performance liquid chromatography–
ip t
tandem mass spectrometry Abstract
cr
Assessment of total uncertainty of analytical methods for the measurements of drugs in human hair has mainly
us
been derived from the analytical variation. However, in hair analysis several other sources of uncertainty will contribute to the total uncertainty. Particularly, in segmental hair analysis pre-analytical variations associated with the
an
sampling and segmentation may be significant factors in the assessment of the total uncertainty budget. The aim of this study was to develop and validate a method for the analysis of 31 common drugs in hair using ultra-performance
M
liquid chromatography–tandem mass spectrometry (UHPLC–MS/MS) with focus on the assessment of both the preanalytical and analytical sampling variations. The validated method was specific, accurate (80-120%), and precise (CV ≤ 20%) across a wide linear concentration range from 0.025–25 ng/mg for most compounds. The analytical variation
ed
was estimated to be less than 15% for almost all compounds. The method was successfully applied to 25 segmented hair specimens from deceased drug addicts showing a broad pattern of poly-drug use. The pre-analytical sampling
ce pt
variation was estimated from the genuine duplicate measurements of two bundles of hair collected from each subject after subtraction of the analytical component. For the most frequently detected analytes, the pre-analytical variation was estimated to be 26–69%. Thus, the pre-analytical variation was 3–7 fold larger than the analytical variation (7–
Ac
13%) and hence the dominant component in the total variation (29–70%). The present study demonstrated the importance of including the pre-analytical variation in the assessment of the total uncertainty budget and in the setting of the 95%-uncertainty interval (± 2 CVT). Excluding the pre-analytical sampling variation could significantly affect the interpretation of results from segmental hair analysis.
Keywords: Segmental hair analysis, pre-analytical variation, uncertainty, drugs of abuse, pharmaceuticals, drug addicts
1 Page 4 of 22
1. Introduction Hair analysis has gained increasing importance in recent year for the analysis of drugs in therapy compliance control, drug-facilitated sexual assaults, driving license re-granting, chronic drug abuse intoxication and post-mortem toxicology [1–5]. In contrast to other biological matrices analysis of hair allows retrospective investigation of drug
ip t
consumption and evaluation of historic drug use when using segmental hair. These advantages are offset by shortcomings of external contamination leading to false-positive hair results, as well as low drug concentrations,
cr
which require more sensitive analytical methods for both screening and quantification.
Several multi-target liquid chromatography–tandem mass spectrometry methods (HPLC or UHPLC–MS/MS) have
us
been reported for drug determination in human hair [6–10]. The validation of analytical methods for hair analysis has mainly focused on the evaluation of specificity, sensitivity, trueness, and precision of the target analytes. However, it
an
is also important to account for sources of uncertainty from hair sampling, segmentation and washing procedures, etc. Especially, in segmental hair analysis, the precision and accuracy of the method is highly dependent on both the
M
sampling and segmentation procedure [11]. Biological variation associated with the variability in hair growth, the variations in drug incorporation rates, horizontal washout and deposition by diffusion from sweat and sebum may also
ed
contribute to the total uncertainty [11–15]. Another factor that could contribute to the overall variation is the amount of incorporated drug, which depends on several things such as the melamine content, blood flow, blood
ce pt
concentration, as well as the acid dissociation constant (pK a) and the lipophilicity of the drug [13]. The aim of this study was to develop and validate an ultra-performance liquid chromatography–tandem mass spectrometry (UHPLC–MS/MS) method for the analysis of 31 common drugs in human hair with focus on the
Ac
assessment of both the pre-analytical and analytical variations. For this purpose, we have adapted a recently presented method for blood analysis of 31 common pharmaceuticals and drugs of abuse for hair analysis [16].
2. Materials and methods 2.1 Hair samples
Authentic hair from 25 deceased drug addicts was routinely collected during forensic autopsies by trained forensic technicians. Hair samples of approximately 1 cm in diameter were collected with a scissor from the posterior vertex
2 Page 5 of 22
region as close to the scalp as possible and wrapped in tinfoil. The hair samples were stored at room temperature until analysis. Blank hair samples used for negative controls, spiked calibration standards and quality control samples were obtained from drug-free healthy volunteers. Different types of drug-free hair (blond, dark brown, straight and curly)
ip t
were used for the investigations of matrix interferences and specificity. Quality control samples were prepared by soaking 2 g drug-free hair cut in 2–3 mm pieces with a mixture of
removed with four times of 40 ml MeOH, and the spiked hair was dried overnight.
cr
reference standards (20 mg/L) in DMSO (50%) for 10 days with regular shaking. The standard solutions were briefly
us
An external hair control sample was obtained from Arvecon GmbH (Walldorf, Germany) and a certified proficiency
an
test was obtained from the Society of Hair Testing (SoHT).
2.2 Sample preparation of authentic hair samples
M
Two bundles of hair collected from the same case were analyzed as genuine duplicate measurements. All hair samples were carefully aligned and cut into two segments; S1 (5 mm) and S2 (10 mm), where S1 was the proximal
ed
segment (root). About 10 mg of each segment was weighted into 2 mL hard tissue grinding tubes (Precellys MK28-R, Bertin Technologies, France). The hair samples were washed one time with 2-propanol (1 mL), two times with purified
ce pt
water (1 mL) and finally one time with 2-propanol (1 mL). The mixtures were shaken 5 min with the first three washing solutions and 1 min with the last washing solution. The washing solutions were removed after 3 min centrifugation at 3600 rpm at room temperature. The last washing water was analyzed to test for external contamination. The decontaminated hair samples were dried overnight, weighed, and mixed with six stainless steel beads before addition
Ac
of 500 µL extraction media (25:29:24 (v/v) methanol:acetonitrile:2mM ammonium formate (pH 5.3)) and 25 µL IS solution (0.4 mg/L). The hair samples were pulverized in a Precellys 24 tissue homogenizer (Bertin Technologies, France) for three times, with 2 x 30 sec bursts at 5600 rpm. The pulverized hair samples were extracted in an oven at 37:C for 18 h and filtered with MiniUni Prep filter vials (Whatman Inc., New Jersey) as described by Nielsen et al. [8].
2.3 Liquid chromatography–tandem mass spectrometry The hair extracts were analyzed with an UHPLC–MS/MS method previously developed in our laboratory for routine analysis of whole blood samples with small modifications [16]. Cathine and cathinone were replaced with EDDP and
3 Page 6 of 22
oxycodone. The MRM transitions for EDDP were m/z 278→234 and m/z 278→249 as the quantitative and confirmative traces, respectively. The cone voltage was set to 50 V and the collision energies were set to 34 and 22 eV for MRM 1 and MRM 2, respectively. For oxycodone, MRM 1 and MRM 2 were m/z 316→241 and m/z 316→298. The cone voltage was set to 30 V and the collision energies were set to 28 and 10 eV, respectively.
ip t
The analytes in authentic samples were identified by the retention times and the ion ratio of two characteristic
cr
MRM transitions. Tolerance was set to ±1% for the retention time and ± 30% for the ion ratio.
2.4 Method validation
document EN ISO/IEC 17025:2000) and as described by Peters et al. [17].
us
The method was fully validated according to The Danish Accreditation and Metrology Fund (DANAK, guidance
an
The specificities of the methods were investigated in different hair types by analyzing 10 drug-free hair samples of different origins. The same 10 drug-free hair samples were also spiked at two levels: 0.05 and 0.5 ng/mg, and the
M
recoveries and coefficient of variation (CV%) were calculated for each hair sample. Post-column infusion experiments were performed for 10 selected compounds eluting at different time areas:
ed
amphetamine, codeine, ketamine, ketobemidone, MDMA, methamphetamine, oxycodone, chlordiazepoxide, nitrazepam, zopiclone, and their deuterated internal standards. The compounds were divided into four groups based
ce pt
on their retention time and masses. The compounds were injected continuously into the mass spectrometer postcolumn using a syringe pump and a T-piece connected to the UHPLC-system. The syringe pump delivered a constant flow of 10 µL/min standard solution (0.05 mg/L), and the UHPLC-system delivered a constant flow of 0.4 mL/min mobile phase. The baseline responses were monitored following the injection of 10 µL extracts of nine drug-free hair
Ac
samples from the autosampler. These baseline responses were compared to the baseline responses following injection of a blank mobile phase. The absolute recoveries were determined by comparing the absolute peak area for each analyte in fortified hair at 5 ng/mg with the absolute peak area for each analyte in a pure standard solution at similar level in 16 different series. The linear range was investigated by analyzing matrix-extracted calibrators at seven concentration levels from 0.025 to 25 ng/mg. The analysis was repeated in four different series. The calibration curves were obtained by plotting the peak area of the analytes to IS versus the theoretical concentration. Linear regression with 1/x weighting was 2
used, and the linearity was evaluated from the residual plots and the squared correlation coefficient (r ). The linearity
4 Page 7 of 22
range was defined as the range from the lower limit of quantification (LLOQ) to the highest level where the criteria of both precision (CV ≤ 15%) and accuracy (85–115%) were met. The LLOQ was defined as the lowest concentration where CV was ≤ 20% and accuracy was within 80–120%. The evaluation of precision and accuracy were performed at seven concentration levels: (0.025; 0.050; 0.10; 0.50;
ip t
2.5; 10 and 25 ng/mg). The determinations were based on four replicates performed on four days (N=16). The longterm precision was determined from measurements of low and high quality control samples analyzed in 17 series.
cr
The total uncertainty (CVT) was derived from the individual components of variations [18]:
us
The analytical uncertainty (CVA) was derived as an average value of the long-term precision from low and high quality control samples. The variation of the calibrator (CVcal) was estimated from the uncertainty of the purity of the
an
compound and the uncertainty of the preparation of calibrator solution. The sampling variation (CV) was derived from
ed
relative differences of duplicate measurements:
M
measurements of two bundles of hair collected from each subject on the basis of the standard deviation of the
where rdi was the paired difference between measurements of the ith sample divided by the mean of the two measurements. The pre-analytical variation (CVPA) was derived from the sampling variation and the analytical variation
ce pt
from the following relation:
Ac
A 95%-uncertainty interval for authentic hair samples was estimated as the measured value ± 2 CV T.
3. Results
3.1 Validation of the method
The validation data are summarized in Table 1. The method was found to be selective for all the tested analytes since no interfering peaks were observed in the traces of the analytes or the internal standards when analyzing 10 drug-free hair samples except for morphine where a small peak was detected below LLOQ in one sample. When the same 10 drug-free hair samples were spiked at two different levels the average recoveries ranged between 80 and
5 Page 8 of 22
120% for almost all compounds. The coefficient of variation between the recoveries obtained from 10 different types of hair was less than 15% for the majority of the analytes. The post-column infusion experiment showed some ion suppression or enhancement for all the selected compounds, but the suppression or enhancement was in the same order of magnitude as their deuterated internal
ip t
standards. The only compound which was not corrected by the chosen internal standard was ketobemidone, but since the variations between the different hair types was small, the use of matrix-matched standards will minimize the
cr
effect of suppression from endogenous hair matrix compared to a pure standard solution. A deuterated internal standard for ketobemidone is preferable to correct for suppression, but a deuterated analogue was not available at
us
the time the method was validated.
Interference from co-eluting analytes and from other possible drugs in forensic cases has previously been
an
investigated for the chromatographic system by Bjørk et al. [16]. No significant interferences from co-eluting analytes or from 167 tested drugs such as other benzodiazepines, analgesics, antidepressants, antipsychotics, β-blockers,
M
narcotics, and stimulants contributed to positive blank values in the traces of the target analytes. For all targets the average absolute recovery ranged from 42 to 106%. The recoveries for the internal standard
ed
were of the same order of magnitude as the corresponding analyte. Linearity was observed from 0.025 to 10 or 25 2
ng/mg for all compounds except alprazolam with squared correlation coefficients (r ) greater than 0.994. Deviations
ce pt
from the fitted curve were less than 15% for levels above LLOQ and less than 20% at LLOQ for most compounds. A value for LLOQ of 0.025 ng/mg was obtained for all compounds except for cocaine, codeine, morphine, 6monoacetylmorphine, oxycodone, and nordiazepam, where the obtained LLOQ values were 0.050 ng/mg. Cases with
system.
Ac
values above the upper limit of quantification were diluted with extraction media before injection into the UHPLC-
For all analytes the criteria for precision and accuracy were fulfilled at seven concentration levels except for a few compounds, where CV was slightly above the limit for one concentration level. Chlordiazepoxide, morphine, and oxycodone showed CVs up to 23% at one level, while the accuracy was within 85–115% at all levels. The long-term precisions were below 15% at both levels except for chlordiazepoxide (≤25%), ketamine (≤16%), methamphetamine (≤21%), and oxycodone (≤22%). The accuracy of the method was further evaluated for some analytes by analyzing an external quality control sample from Arvecon GmbH in 15 different series. The average accuracies ranged between 83
6 Page 9 of 22
and 105% for amphetamine, benzoylecgonine, cocaine, codeine, EDDP, methadone, methamphetamine, 6monoacetylmorphine, and morphine. The accuracy of the method was also tested for some analytes (amphetamine, benzoylecgonine, cocaine, codeine, MDA, MDMA, methadone, methamphetamine, 6-monoacetylmorphine, morphine and tramadol) by participating in a
ip t
quality proficiency program provided by the Society of Hair Testing (SoHT). All the analytes passed the test with zscores less than 2 (z = (observed value-median of participants)/SD).
cr
The method was successfully applied to 25 authentic hair specimens from deceased drug addicts. Thirty of the target compounds were detected in one or more cases and several drug groups were found in each case, indicating a
us
broad poly-drug use. Most of the drugs were found in both hair segments and there were no significant differences
an
between the mean concentration levels in the two segments according to a Wilcoxon’s rank test (P > 0.05).
3.2 Evaluation of components of variations
M
The components of variation for all the positive findings and the most frequently detected analytes are summarized in Table 2. The most frequently detected analytes accounted for 50% of the positive findings. The
ed
distribution of variation was similar in the two segments and there was no significant difference between the relative differences in the two segments (Wilcoxon’s rank test, P > 0.05). As seen from the percentage differences plots of the
ce pt
most frequently occurring drugs methadone and cocaine, the relative differences were evenly distributed around the zero line in both hair segments (Fig. 1 and Fig. 2).
The analytical coefficient of variation was less than 13% for the frequently detected analytes. As opposed to the analytical variability, the pre-analytical variation and thus the total uncertainty depended on the analytes. The pre-
Ac
analytical variation ranged from 41 to 69% for most of the compounds resulting in total uncertainties of 42–70%. For the benzodiazepines bromazepam and diazepam, the pre-analytical variation values were slightly lower, with average values of 27 and 38%, respectively. The total uncertainties for these two compounds were 29 and 40%, respectively. The pre-analytical variation was in general 3–7 fold larger than the analytical variation and thus the most essential component of variation due to the principle of squared addition. The calibrator variations were estimated to be 0.5– 5% for the frequently detected analytes and constituted only a small part of the total uncertainty budget. The 95%-uncertainty interval was calculated for the median values for the frequently detected analytes with and without the pre-analytical variation included in the estimation of the total uncertainty. If the pre-analytical variation
7 Page 10 of 22
was not included in the estimation of the total uncertainty, the 95%-uncertainty interval was underestimated with a factor of 3–7 depending on the analyte. For example, for a diazepam result of 0.17 ng/mg, the estimated 95%uncertainty interval ranged from 0.13 to 0.21 ng/mg if the total uncertainty was only derived from the analytical and calibrator variations, while the estimated 95%-uncertainty interval ranged from 0.034 to 0.31 ng/mg if the pre-
ip t
analytical variation was taken into account. Thus, for diazepam the result was underestimated by a factor 3 without the contribution of pre-analytical variation.
cr
Although the components of variation depended on the analyte, the uncertainties for all the positive findings were also evaluated. An average analytical variation of all the analytes included in the method was estimated to be 10%
us
based on the long-term precisions. The total uncertainty for 203 positive determinations was 48–52% and as for the
an
individual analytes the pre-analytical variation accounted for the main part of the total uncertainty.
4. Discussion
M
The validated method was shown to be specific, accurate, and precise across a wide linear range of concentrations as a result of using deuterated analogues as internal standards and by using matrix-matched standards for calibration.
ed
The validation results are similar to the validation results obtained with an ultra-performance liquid chromatography– time-of-flight mass spectrometry (UHPLC–TOF-MS) method for hair analysis previously developed in our laboratory
ce pt
[8]. The former method was able to quantify 52 common pharmaceuticals and drugs of abuse in hair simultaneously, but the sensitivity of amphetamine, methamphetamine, MDA, MDMA, and morphine did not fulfill the cut-off values recommended by the SoHT for quantitative methods [11]. The sensitivity of the present method exceeds that of the
Ac
former method by a factor of 10 or more for these compounds and thus, the established LLOQ values for the common drugs of abuse were all lower than the recommended SoHT cut-off values. Both the sensitivity and the wide linear concentrations range were found suitable for the determination of the targets analytes in hair samples from chronic drug users. If the method should be used for measurement of a single-dose, an up concentration step could easily be implemented to improve the sensitivity. The high pre-analytical variation may be associated with inconsistent sampling and segmentation of the hair samples [11,19]. It is normally assumed that the hair is collected directly next to the scalp, but this is difficult to achieve in practice, especially when the hair is oily or filtered as in autopsy cases. It has previously been demonstrated
8 Page 11 of 22
that between 0.1 and 3.0 cm of hair with an average of 0.8 ± 0.1 cm remained on the scalp after collecting of the hair samples, depending on the sampling procedure and the length of the hair [19]. This large uncertainty associated with insufficient sampling could significantly affect the total uncertainty of the method especially in segmental hair analysis.
ip t
Besides the experimental uncertainties, the biological variation associated with variability in hair growth rate, drug incorporation rate, hair pigmentation, and blood supply to the hair follicle may also contribute to the total variation
cr
[12–15]. The hair grows in three phases and at any time a substantial proportion of hair will not be in the growing phase, leading to irregular hair growth [14]. The hair growth rate is most homogenous on the head at the posterior
us
vertex region, where approximately 15% of the hairs are in the resting phase and the remaining 85% are in their growing phase [20]. It is therefore recommended that the sample is collected from the posterior vertex region as close
an
to the scalp as possible to minimize the uncertainty associated with variable hair growth [11]. Although, the hair
most predominant component of the total variation.
M
samples used in this study were collected according to the recommendations, the pre-analytical variation was the
The intra-individual variability of incorporated drug due to pigmentation has been investigated in white and grey
ed
hair collected from the same donor [21,22]. In two cases, the results showed that the zolpidem concentration was more than 100-fold higher in the black hair compared to the white hair. These findings are consistent with other
ce pt
pigmentation studies where levels of amphetamine, cocaine, and methamphetamine were found to be significantly higher in pigmented hair than in white hair [23,24]. Thus, persons with plain pigmented hair would have more consistent incorporation of drug than persons with grey hair. Also inter-individual variability in the amount of incorporated drug has been evaluated by administration of single-dosages of benzodiazepines to different volunteers
Ac
followed by segmental hair analysis [21,25–27]. The detected levels of benzodiazepines in the segment representing the exposure period were similar in Chinese subjects with black hair, while the levels deviated more in subjects with different hair types.
Reports on variations between replicates of authentic hair results are sparse. In two studies, the relative deviations were less than 20% for most of the detected analytes [8,28]. The replicates were not genuine determinations of hair samples collected at different sites from the same subject, but replicate measurements of the same sample of finely cut hair, thus excluding the pre-analytical variation. The obtained relative deviations in these two studies are
9 Page 12 of 22
consistent with the analytical coefficients of variation assessed in this study, which were based on replicate measurements of homogenous cut hair.
5. Conclusion
ip t
A specific and reproducible multi-target UHPLC–MS/MS method has been developed for the determination of 31 common pharmaceuticals and drugs of abuse in human hair. The method was fully validated and successfully applied
cr
for the quantification of pharmaceuticals and drugs of abuse in segmented hair samples from 25 autopsy cases. The total uncertainty was estimated to approximately 29–70% from genuine duplicate measurements of hair samples
us
depending on the analytes. The pre-analytical variation was 3–7 fold larger than the analytical variation and thus the dominant component of variation. The study demonstrated the importance of including the pre-analytical variation
an
when evaluating the total uncertainty and the width of the 95%-uncertainty interval for authentic hair samples. Due to this large uncertainty associated with hair analysis, precautions should be taken when evaluating quantitative results
Ac
ce pt
ed
M
in authentic hair and when comparing hair results within and between subjects.
10 Page 13 of 22
Figure legends Fig. 1. Plots of percentage differences between methadone concentrations in two bundles of hair in segment 1 (top)
ip t
and segment 2 (bottom). The long-dashed line indicates the averages differences, and the short-dashed line indicates the 95%-uncertainty interval (± 2 CVT).
cr
Fig. 2. Plots of percentage differences between cocaine concentrations in two bundles of hair in segment 1 (top) and segment 2 (bottom). The long-dashed line indicates the averages differences, and the short-dashed line indicates the
Ac
ce pt
ed
M
an
us
95%-uncertainty interval (± 2 CVT).
11 Page 14 of 22
References
ip t
[1] F. Musshoff, K. Lachenmeier, J. Trafkowski, B. Madea, F. Nauck, U. Stamer, Determination of opioid analgesics in hair samples using liquid chromatography/tandem mass spectrometry and application to patients under
cr
palliative care, Ther. Drug. Monit. 29 (2007) 655–661.
us
[2] M. Chèze, G. Duffort, M. Deveaux, G. Pepin, Hair analysis by liquid chromatography–tandem mass spectrometry in toxicological investigation of drug-facilitated crimes: Report of 128 cases over the period
an
June 2003 May 2004 in metropolitan Paris, Forensic Sci. Int. 153 (2005) 3–10.
[3] P. Kintz, A. Tracqui, P. Mangin, Detection of Drugs in Human Hair for Clinical and Forensic Applications, Int. J.
M
Legal Med. 105 (1992) 1–4.
ed
[4] R. Kronstrand, I. Nystrom, M. Forsman, K. Kall, Hair analysis for drugs in driver's license regranting. A Swedish pilot study, Forensic Sci. Int. 196 (2010) 55–58.
ce pt
[5] F. Musshoff, K. Lachenmeier, H. Wollersen, D. Lichtermann, B. Madea, Opiate concentrations in hair from subjects in a controlled heroin-maintenance program and from opiate-associated fatalities, J. Anal. Toxicol. 29 (2005) 345–352.
Ac
[6] M. Laloup, M.D.R. Fernandez, G. De Boeck, M. Wood, V. Maes, N. Samyn, Validation of a liquid chromatography–tandem mass spectrometry method for the simultaneous determination of 26 benzodiazepines and metabolites, zolpidem and zopiclone, in blood, urine, and hair, J. Anal. Toxicol. 29 (2005) 616–626.
[7] J. Kim, S. Lee, S. In, H. Choi, H. Chung, Validation of a simultaneous analytical method for the detection of 27 benzodiazepines and metabolites and zolpidem in hair using LC–MS/MS and its application to human and rat hair, J. Chromatogr. B 879 (2011) 878–886.
12 Page 15 of 22
[8] M.K. Nielsen, S.S. Johansen, P.W. Dalsgaard, K. Linnet, Simultaneous screening and quantification of 52 common pharmaceuticals and drugs of abuse in hair using UPLC–TOF-MS, Forensic Sci. Int. 196 (2010) 85–92.
[9] E. Lendoiro, Ó. Quintela, A. de Castro, A. Cruz, M. Ópez-Rivadulla, M. Concheiro, Target screening and confirmation of 35 licit and illicit drugs and metabolites in hair by LC–MSMS, Forensic Sci. Int. 217 (2012)
ip t
207–215.
cr
[10] K.Y. Rust, M.R. Baumgartner, N. Meggiolaro, T. Kraemer, Detection and validated quantification of 21
us
benzodiazepines and 3 “z-drugs” in human hair by LC–MS/MS, Forensic Sci. Int. 215 (2012) 64–72.
[11] G.A.A. Cooper, R. Kronstrand, P. Kintz, Society of Hair Testing guidelines for drug testing in hair, Forensic Sci.
an
Int. 218 (2012) 20–24.
M
[12] E.J. Cone, Mechanisms of drug incorporation into hair, Ther. Drug. Monit. 18 (1996) 438-443.
[13] F. Pragst, M.A. Balikova, State of the art in hair analysis for detection of drug and alcohol abuse, Clin. Chim.
ed
Acta. 370 (2006) 17–49.
[14] H. Sachs, Theoretical limits of the evaluation of drug concentrations in hair due to irregular hair growth,
ce pt
Forensic Sci. Int. 70 (1995) 53–61.
[15] G.L. Henderson, M.R. Harkey, C.H. Zhou, R.T. Jones, P. Jacob, Incorporation of isotopically labeled cocaine and metabolites into human hair: 1. Dose-response relationships, J. Anal. Toxicol. 20 (1996) 1–12.
Ac
[16] M. Bjørk, K. Simonsen, D. Andersen, P. Dalsgaard, S. Sigurôardóttir, K. Linnet, B. Rasmussen, Quantification of 31 illicit and medicinal drugs and metabolites in whole blood by fully automated solid-phase extraction and ultra-performance liquid chromatography–tandem mass spectrometry, Anal. Bioanal. Chem. 405 (2013) 2607–2617.
[17] F.T. Peters, O.H. Drummer, F. Musshoff, Validation of new methods, Forensic Sci. Int. 165 (2007) 216–224.
13 Page 16 of 22
[18] K. Linnet, J.C. Boyd, Selection and analytical evaluation of methods-with statistical techniques, in: C.A. Burtis, th
E.R. Ashwood, D.E. Bruns (Eds), Tietz textbook of clinical chemistry and molecular diagnostics, 5
ed.,
Elsevier/Saunders, St. Louis, 2012, pp. 7–48.
[19] M.A. LeBeau, M.A. Montgomery, J.D. Brewer, The role of variations in growth rate and sample collection on
cr
[20] M.R. Harkey, Anatomy and physiology of hair, Forensic Sci. Int. 63 (1993) 9–18.
ip t
interpreting results of segmental analyses of hair, Forensic Sci. Int. 210 (2011) 110–116.
us
[21] M. Villain, M. Chèze, A. Tracqui, B. Ludes, P. Kintz, Windows of detection of zolpidem in urine and hair: application to two drug facilitated sexual assaults, Forensic Sci. Int. 143 (2004) 157–161.
an
[22] H. Miyaguchi, Determination of zolpidem in human hair by micropulverized extraction based on the evaluation of relative extraction efficiency of seven psychoactive drugs from an incurred human hair
M
specimen, J. Chromatogr. A 1293 (2013) 28–35.
ed
[23] R. Kronstrand, J. Ahlner, N. Dizdar, G. Larson, Quantitative analysis of desmethylselegiline, methamphetamine, and amphetamine in hair and plasma from Parkinson patients on long-term selegiline
ce pt
medication, J. Anal. Toxicol. 27 (2003) 135–141.
[24] R.W. Reid, F.L. OConnor, A.G. Deakin, D.M. Ivery, J.W. Crayton, Cocaine and metabolites in human graying hair: Pigmentary relationship, J. Toxicol-Clin. Toxic. 34 (1996) 685–690.
Ac
[25] P. Xiang, Q.R. Sun, B.H. Shen, P. Chen, W. Liu, M. Shen, Segmental hair analysis using liquid chromatography– tandem mass spectrometry after a single dose of benzodiazepines, Forensic Sci. Int. 204 (2011) 19–26.
[26] M. Villain, M. Chèze, A. Tracqui, B. Ludes, P. Kintz, Testing for zopiclone in hair application to drug-facilitated crimes, Forensic Sci. Int. 145 (2004) 117–121.
[27] A. Negrusz, C.M. Moore, K.B. Hinkel, T.L. Stockham, M. Verma, M.J. Strong, P.G. Janicak, Deposition of 7aminoflunitrazepam and flunitrazepam in hair after a single dose of Rohypnol (R), J. Forensic Sci. 46 (2001) 1143–1151.
14 Page 17 of 22
[28] H. Miyaguchi, H. Takahashi, T. Ohashi, K. Mawatari, Y.T. Iwata, H. Inoue, T. Kitamori, Rapid analysis of methamphetamine in hair by micropulverized extraction and microchip-based competitive ELISA, Forensic
Ac
ce pt
ed
M
an
us
cr
ip t
Sci. Int. 184 (2009) 1–5.
15 Page 18 of 22
Ac ce p
te
d
M
an
us
cr
ip t
Figure 1 methadone.eps
Page 19 of 22
Ac ce p
te
d
M
an
us
cr
ip t
Figure 2 cocaine.eps
Page 20 of 22
Table 1. Validation data. Linear range
LLOQ
ng/mg
Accuracy %
Levels above LLOQ CV %
Accuracy %
CV %
QC1
QC2
us
Analyte
cr
ip t
Table
ng/mg
CV %
ng/mg
CV %
Ac c
ep te
d
M
an
Alprazolam 0.025-2.5 110 13 94–114 10–14 1.2 6 3.5 12 Amphetamine 0.025-25 106 9 100–108 8–14 0.59 7 1.5 11 Benzoylecgonine 0.025-25 94 9 97–105 4–9 0.83 8 1.5 5 Bromazepam 0.025-25 96 12 97–114 6–8 1.7 10 5.1 10 Chlordiazepoxide 0.025-10 99 22 101–106 10–18 0.371 16 2.9 25 Clonazepam 0.025-10 95 15 94–106 4–12 1.4 5 4.7 9 7-Aminoclonazepam 0.025-25 98 7 94–102 5–10 0.55 5 1.8 5 Cocaine 0.050-10 102 6 96–104 4–11 1.8 10 0.74 11 Codeine 0.050-10 98 14 92–110 5–15 0.11 11 0.771 11 Diazepam 0.025-10 91 17 96–104 4–14 2.1 14 6.9 11 Nordiazepam 0.050-10 95 13 102–113 12–14 2.4 10 6.9 10 EDDP 0.025-25 105 17 93–104 3–17 0.30 5 1.1 7 Flunitrazepam 0.025-25 90 18 96–106 5–10 1.2 7 4.2 5 7-Aminoflunitrazepam 0.025-25 98 5 95–103 4–10 0.58 3 1.8 4 Ketamine 0.025-10 93 15 97–115 8–14 0.33 9 0.95 16 Ketobemidone 0.025-10 94 9 88–110 5–12 0.19 14 0.62 15 MDA 0.025-25 84 18 93–100 3–10 0.77 7 2.2 8 MDMA 0.025-10 100 19 96–109 5–19 0.73 11 2.11 13 Methamphetamine 0.025-10 93 16 89–108 5–13 0.54 14 1.41 21 Methadone 0.025-25 93 12 91–101 3–8 0.60 9 2.2 7 Morphine 0.050-25 97 21 98–113 5–11 0.077 11 0.62 6 6-Monoacetylmorphine 0.050-25 93 7 96–110 5–10 0.067 9 0.59 7 Nitrazepam 0.025-10 93 12 94–106 4–12 1.5 6 4.9 8 7-Aminonitrazepam 0.025-25 90 8 97–106 4–9 0.75 4 2.3 5 Oxazepam 0.025-25 98 9 101–113 5–13 1.7 10 5.6 9 Oxycodone 0.050-25 98 18 87–108 5–23 0.093 22 0.71 17 Temazepam 0.025-10 89 13 100–108 5–10 1.9 8 6.0 11 Tramadol 0.025-25 106 8 85–102 2–16 0.14 9 0.46 11 O-desmethyltramadol 0.025-25 95 9 95–101 5–8 0.073 6 0.51 6 Zolpidem 0.025-10 92 12 96–107 7–12 0.68 5 2.2 7 Zopiclone 0.025-25 97 17 91–104 7–16 0.37 11 1.7 12 1 One outlier removed. 2Mean recovery (RE) and coefficient of variation (CV) obtained from 10 different drug-free hair samples spiked at 0.05 ng/mg.
Absolute recovery
Recovery2
%
RE %
CV %
86 78 82 43 56 58 70 67 79 42 55 92 70 85 80 102 73 78 74 57 64 84 59 74 60 79 64 101 106 95 61
87 123 107 118 118 101 104 111 146 96 89 102 102 106 112 97 101 113 104 103 105 101 117 110 112 96 83 108 105 120 76
6 11 4 10 12 9 5 3 10 7 12 2 4 4 12 10 4 7 15 13 15 7 5 4 10 17 6 6 3 8 23
Page 21 of 22
Table
Table 2. Components of variation for frequently detected analytes in 25 autopsy cases. Segment
N
Median ng/mg
Range ng/mg
CVPA %
CVA %
CVcal %
CVT %
All positive findings
S1 S2
203 212
0.45 0.48
0.025–67 0.028–55
511 471
101
21
521 481
Amphetamine
S1 S2
8 7
0.27 0.48
0.029–2.0 0.043–2.5
67 69
9
0.6
67 70
Bromazepam
S1 S2
10 9
0.69 0.70
0.089–2.0 0.060–2.6
26 27
10
5
29 29
Clonazepam
S1 S2
6 6
0.31 0.26
0.098–0.93 0.043–0.98
52 53
7
2
53 53
Cocaine
S1 S2
11 11
1.9 4.3
0.083–67 0.14–37
51 60
11
2
53 61
Codeine
S1 S2
8 7
0.60 0.28
0.092–7.9 0.071–7.9
59 42
11
0.5
60 43
Diazepam
S1 S2
9 9
0.17 0.12
0.025–2.1 0.040–2.0
38 37
13
2
40 39
Methadone
S1 S2
24 24
9.0 8.6
0.33–65 0.24–55
47 41
2
48 42
Morphine
S1 S2
9 9
0.23 0.18
0.31–3.3 0.097–2.6
Tramadol
S1 S2
6 6
2.2 1.6
0.064–23 0.046–17
Zopiclone
S1 S2
8 8
0.55 0.28
0.039–3.1 0.61–2.4
cr
us
8
64 54
8
5
64 55
65 44
10
0.6
66 45
49 51
12
2
51 52
an
1
ip t
Compound
Ac
ce pt
ed
M
An average coefficient of variation was calculated for all the analytes included in the method.
Page 22 of 22
