Egyptian Journal of Forensic Sciences (2014) 4, 45–49
Contents lists available at ScienceDirect
Egyptian Journal of Forensic Sciences journal homepage: www.ejfs.org
A recommended procedure for establishing the source level relationships between heroin case samples of unknown origins Kar-Weng Chan *, Nurul Dalila Nik Mat, Mohamad Faiez Abdul Rahim, Maimonah Sulaiman Department of Chemistry Malaysia, Ministry of Science, Technology and Innovation (MOSTI), Petaling Jaya 46661, Malaysia Received 15 June 2013; revised 23 December 2013; accepted 11 February 2014 Available online 20 March 2014
KEYWORDS Heroin profiling; Impurities; Sample classification
Abstract A recent concern of how to reliably establish the source level relationships of heroin case samples is addressed in this paper. Twenty-two trafficking heroin case samples of unknown origins seized from two major regions (Kuala Lumpur and Penang) in Malaysia were studied. A procedure containing six major steps was followed to analyze and classify these samples. Subsequently, with the aid of statistical control samples, reliability of the clustering result was assessed. The final outcome reveals that the samples seized from the two regions in 2013 had highly likely originated from two different sources. Hence, the six-step procedure is sufficient for any chemist who attempts to assess the relative source level relationships of heroin samples. ª 2014 Production and hosting by Elsevier B.V. on behalf of Forensic Medicine Authority.
1. Introduction Despite the high influx of various synthetic drugs into the black market, heroin remains the most frequently abused sub-stance in Malaysia. This semi-synthetic drug is processed from morphine, one of the active compounds present in the opium latex. The resulting heroin can be diluted/cut with many other substances, such as caffeine and paracetamol, two commonly used diluents in Malaysia. Uncut heroin substances * Corresponding author. Tel.: +60 123627609. E-mail address:
[email protected] (K.-W. Chan). Peer review under responsibility of Forensic Medicine Authority.
Production and hosting by Elsevier
usually contain a high amount of heroin (HR) and quantifiable amounts of codeine (CD), morphine (MP), acetylcodeine (AC) and monoacetylmorphines (MM). It was demonstrated that most heroin samples in Malaysia still contained the aforementioned compounds at detectable levels even though they had been 90% cut.1 In some cases, papaverine, noscapine and thebaine can also be detected. However, these three compounds can become masked very easily if the heroin substance is diluted with other substances. As a result, it is established that only the five formerly mentioned opium alkaloids can be effectively used as indicators for heroin profiling since they are readily measurable in the local heroin samples. Opium alkaloids and neutral/acidic manufacturing impurities present in the heroin samples have long been used to establish sample relationships at the origin and source levels. As suggested by Besacier et al.2, opium alkaloids that are present as major components are the important indicators that one
2090-536X ª 2014 Production and hosting by Elsevier B.V. on behalf of Forensic Medicine Authority. http://dx.doi.org/10.1016/j.ejfs.2014.02.001
46 Table 1
K.-W. Chan et al. Recommended procedure for determining the source level relationships between heroin case samples of unknown origins.
No.
Step
Description
Task previously completed or to be undertaken in this study
1.
Method optimization and validation
The analytical method (e.g., gas chromatographic method) must be fit for the profiling purpose. For example, it should meet the acceptance criteria set by the international bodies or the laboratory undertaking the task
2.
Statistical Validation
This must be done by using samples of known origins or known relationships to assess the validity of the preferred clustering technique. For example, it should inform which data pretreatment and distance measure that are able to provide the best clustering results
3.
Collection of statistical control sample(s)
4.
Collection of case samples
Some samples should be collected to serve as statistical controls. The sample should be at least 2-year temporally spaced from the case samples to be profiled Case samples are collected for analysis
A gas chromatography-flame ionization detection method has been developed and proved stable for Malaysian heroin samples in this laboratory.1,3–10 Note: Previous studies employed concentrations (equivalent to area ratios) for step 1. This study, However, employed peak areas that were corrected by the IS. Hierarchical cluster analysis (HCA) using Ward linkage and Manhattan distance presented promising clustering results for simulated samples of known relationships when GC data heroin were interpreted in 11 normalized parameters (Nselected), namely AC/HR, AC/MM, AC/ (MM + HR), AC/(MP + MM + HR), MM/ HR, (CD + MP)/(MM + HR), HR/MM, (CD + AC)/(MP + MM + HR), (CD + MP + MM + HR)/AC, HR/ (CD + MP + AC + MM) and (MP + MM + HR)/(CD + AC), which were subsequently standardized (S) to achieve 11 Nselected + S variables. Details of this step can be found in the original study4 To be undertaken in this study A statistical control sample was collected in 2010 from Kuala Lumpur
5.
Chemical analysis
6.
Statistical analysis
Statistical control samples and case samples should be analyzed by the same instrumental method determined in step 1 All data obtained should be statistically analyzed using the pretreatment and statistical technique determined in step 2
should profile because they can reliably be used to correlate with other findings derived from trace manufacturing impurities and occluded solvents. Determination of the major components is very straightforward and it involves less sample preparation. Besides, most of these major components are readily extractable with universal solvents. So, much profiling effort can be eliminated as compared with trace manufacturing impurities as the latter requires a tedious liquid–liquid extraction procedure. Hence, five major components present in the illicit heroin samples were used in this study to explain a recommended procedure for establishing the source level relationships of heroin samples with unknown backgrounds. 2. Recommended procedure A number of profiling methods for illicit heroin have been established to determine impurities present in the samples. In the current status, some of these studies have focused on method development and validation, statistical validation and some basic steps for classifying the illicit drug samples.1,3–10 However, there have not been any clear-cut, stepby-step guidelines available to help forensic chemists to determine sample relationships between drug seizures/samples,
To be undertaken in this study 22 case samples seized in 2013 from Kuala Lumpur and Penang were collected To be undertaken in this study
To be undertaken in this study
whether source or street level. In such attempts, one may feel nervous because of three reasons: (1) it is not some kind of routine usually performed in the laboratory; (2) the chemist does not even know how and where to start the profiling work; and (3) the chemist is not sure of the degree of certainty in the result obtained because there lacks some form of quality control (QC) for heroin profiling. Remedially, this paper seeks to put forward a set of practical guidelines to the profiling and determination of the source level relationships between heroin samples of unknown origins using the above-cited five opium alkaloids. One major concern is that HR is easily converted to 6-MM or MP under favorable conditions. Instead of attesting to the stability of these alkaloids, this paper utilizes quotients derived from the five major components to address the issue of stability. With proper pretreatment, the quotients in the form of normalized–standardized variables (Nselected + S) proved excellent for clustering 216 simulated heroin samples of known relationships in a previous study.4 Table 1 summarizes the major steps necessary for classifying heroin case samples through unsupervised pattern recognition (where samples of known origins are not available). The subsequent discussion will demonstrate how the procedure should be followed in order to arrive at a reliable conclusion.
A recommended procedure for establishing the source level relationships between heroin case samples of unknown origins
47
weighed in a dissolution vessel to which 10 mL of IS solution containing 0.18 mg/mL 2,2,2 triphenyl acetophenone in 1:9 methanol:chloroform was added. Subsequently, each aliquot was injected in duplicate into the GC system.
3. Materials and method 3.1. Standards and solvents 2,2,2 Triphenyl acetophenone (internal standard, IS) was purchased from Aldrich Chemical Company. HPLC grade methanol and analytical reagent grade chloroform, respectively manufactured by Fisher Scientific and Merck, were used.
3.4.2. Statistical control samples Five statistical control samples taken from a heroin case sample (marked ‘B’) seized from Kuala Lumpur in 2010 were prepared separately in individual flasks, each weighing 90 mg. The samples were dissolved in 10 mL of IS solution respectively like case samples.
3.2. Gas chromatography–flame ionization detector (GC–FID)3 Analysis was achieved by using an HP6890 N GC-FID preinstalled with a J&W HP Ultra 2 (length 25 m, i.d. 0.20 mm, film thickness 0.33 lm) capillary column. Chromatographic separation was accomplished by holding the oven temperature at 240 C for 1 min and heating up to 270 C at the rate of 12 C/min. The oven was then held for 8–10 min at this temperature. Injector and detector temperatures were both set at 290 C with a split ratio of 40:1 and injection volume of 1 lL.
3.5. Statistical analysis All GC data of the 22 case samples were obtained in peak areas which had been processed against the calibrated peak areas given off by sample A. They were pretreated to achieve 11 Nselected + S variables and statistically analyzed with Minitab 15 in a single data set. 4. Results and discussion
3.3. Calibration
3.4. Sample preparation 3.4.1. Case samples
7.070
3.842
1.034
2.410
Fifteen trafficking samples of illicit heroin seized in 2013 from Penang and 12 from Kuala Lumpur (including its neighboring region of Selangor) were collected and analyzed. For each sample, approximately 90 mg homogenized substance was
8.611
In terms of QC, all the samples were analyzed and corrected by the response of the IS as well as the calibration of sample A. Next, the stability of the instrument was assessed by checking on the percentage error of each analyte. All target peaks exhibited by the analytical control sample obtained errors within ±5% except for MP which was approximately ±20%. For profiling purposes, this laboratory could tolerate this large error for MP since it is, as discussed in the previous literature,3 usually unstable in a large quantity of non-polar solvent, such as chloroform. Furthermore, the large error could be due to the matrix effect from the sample itself. Fig. 1 illustrates the positions of the target peaks on a chromatogram for a sample seized from Kuala Lumpur. The use of the five opium alkaloids is practical to provide information regarding the relationships between samples, particularly at the source/production level since the target compounds are directly linked to its source/production batch. Thus, data of all the case samples together with the five
A previously characterized case sample (marked ‘A’) which contained CD (0.59%), MP (4.64%), AC (3.40%), MM (20.11%) and HR (34.67%) was used to calibrate the instrument. The calibration procedure was carried out using peak area (which was corrected by IS through one-point calibration) instead of concentration, determined from a 90 mg substance of sample A. This sample was also inserted between every 20 runs as an analytical control sample to check the performance of the instrument throughout the course of analysis.
HR
MM 6.172
MP
6.892
AC
9.919
IS
Figure 1
11.996
9.492 9.734
8.676 8.791
7.902
6.728
7.233 7.423
6.402
5.664 5.000 5.069
3.932 4.083 4.347
2.204 2.326 2.549
1.861 2.105
CD
Positions of five chosen opium alkaloids on a chromatogram (KL-7).
48
K.-W. Chan et al.
Similarity
-283.36
-155.58
-27.79
Penang (2013)
Figure 2
Cases
Kuala Lumpur (2013)
CTRL 3
CTRL 2 CTRL 4 CTRL 5
KL8 KL9 KL4 KL5 KL7 CTRL 1
KL6 KL3 KL10
PP1 PP8 PP4 PP11 PP9 PP12 PP15 PP14 PP2 PP3 PP10 PP13 PP6 PP5 PP7 KL1 KL11 KL2 KL12
100.00
KL (2010)
A dendogram showing the clusters of Penang (2013), Kuala Lumpur (2013) and statistical controls (2010).
statistical control samples were pretreated to obtain the 11 above-mentioned Nselected + S variables. As demonstrated in the previous study,4 the variables functioned effectively in grouping all the related sample units into their respective distribution links on a dendogram, when HCA (Ward linkage and Manhattan distance) was employed. From that study, it is crucial to note that only the said variables and the chosen linkage-distance with HCA are able to minimize the withinbatch variation and maximize the between-batch variation in the samples of interest. Therefore, statistical validation is important to ensure the data are manipulated in a statistically sound manner. Specifically, the variables also take into consideration the inter-conversion of the compounds (e.g., HR is decomposed to MM). The sum of morphine content (MP + MM + HR) was used for constructing certain ratios out of the 11 Nselected parameters in order to compensate for the decomposition effect. The pretreated data of all the case samples and statistical control samples were analyzed by HCA (Ward–Manhattan) and the clustering outcome is illustrated in Fig. 2. To ensure the validity and reliability of the results, the integration of the statistical controls is crucial. It is assumed that if these controls can be successfully clustered into a single group on the same dendogram, the overall result shall be reliable.11 According to Fig. 2, all the statistical controls have successfully been clustered in one group (indicated as ‘KL (2010)’) without any failure. This means that the clustering result is statistically valid and could be used for reliable interpretation. The dendogram suggests two sources/batches respectively responsible for the heroin samples seized in Kuala Lumpur and Penang. The samples seized in the respective regions are linked among themselves before they finish the final linkage that links the two clusters representing the two regions. In other words, the samples seized within one particular region had closer agreement in the 11 variables rather than that dis-
played by the samples seized from different regions. Besides, the between-region (KL-2013 versus PP-2013) variability as indicated by the dendogram is definitely greater than the between-year (KL 2013 versus KL 2010) variability. Hence, this again supports that the heroin samples distributed within the two regions were most likely from two different sources. The between-sample variability found within the region cluster is attributable to many reasons. Cutting and decomposition during distribution/storage are among other possible factors causing such variation. If there were many different sources responsible for the mini clusters within the region cluster, the variation in the cultivation and processing conditions in the hypothetical region would have been said to be insignificant as compared with that of the between-region, thus enabling the within-region samples to be clustered back into the region without failure. Overall, the above steps are easy to follow if steps 1 and 2 have been in place. 5. Conclusion It is noteworthy that a valid analytical method and preferred statistical procedure must be established in the laboratory before one can embark on the profiling work to estimate sample relationships. The first two steps may be tedious, but they will be worthwhile as they can help one arrive at accurate conclusions. The integration of statistical controls is another important step to detect errors in the final outcome. The controls will not be clustered into their group if the procedure (e.g., the analytical method, the variables chosen, the linkage and distance measure chosen, etc.) is no longer working optimally with the GC data obtained with the analytical method. If errors arise, this signals that one should check on the procedure and re-establish its reliability by verifying steps 1 and 2. More importantly, it is recommended to obtain heroin samples
A recommended procedure for establishing the source level relationships between heroin case samples of unknown origins from different known countries of origin for testing the recommended procedure in a more rigorous manner. Funding None. Conflict of interest None declared. Ethical approval Necessary ethical approval was obtained from the institute ethics committee. References 1. Chan KW, Tan GH, Wong RCS. Investigation of illicit heroin seized in malaysia: physical characteristics and chemical profiling. Aust J Forensic Sci 2012;44:353–69. 2. Besacier F, Chaudron-Thozet H. Chemical profiling of illicit heroin samples. Forensic Sci Rev 1999;11:105–19. 3. Chan KW, Tan GH, Wong RCS. Gas chromatographic method validation for the analysis of major components in illicit heroin. Sci Justice 2012;52:9–16.
49
4. Chan KW, Tan GH, Wong RCS. The profiling of heroin by associating simulated postcut samples with the corresponding precut sample. J Forensic Sci 2013;58:199–207. 5. Chan KW, Tan GH, Wong RCS. Optimization and statistical validation for the determination of trace impurities in street doses of heroin. Anal Lett 2012;45:1156–71. 6. Chan KW, Tan GH, Wong RCS. Chemometric procedures for analyzing trace organic impurities present in street doses of heroin via a constant weight approach. Aust J Forensic Sci 2012;44:299–309. 7. Besacier F, Chaudron-Thozet H, Rousseau-Tsangaris M, Girald J, Lamotte A. Comparative chemical analyses of drug samples: general approach and application to heroin. Forensic Sci Int 1997;85:113–25. 8. Zelkowics A, Magora A, Ravreby MD, Levy R. Analysis of a simulated heroin distribution chain by HPLC. J Forensic Sci 2005;50:849–52. 9. Esseiva P, Dujourdy L, Anglada F, Taroni F, Margot P. A methodology for illicit heroin seizures comparison in a drug intelligence perspective using large databases. Forensic Sci Int 2003;132:139–52. 10. Zhang ZY, Yang JH, Ouyang H, Li ZJ, Chai ZF, Zhu J, et al. Study of trace impurities in heroin by neutron activation analysis. J Radioanal Nucl Chem 2004;262:295–7. 11. Chan KW, Maimonah S. Statistical controls for sample classification. Buletin Kualiti Teknikal 2013;23:38–42.
