EFFECTS OF EXTRACTION PROCEDURE AND GAS CHROMATOGRAPHY TEMPERATURE PROGRAM ON DISCRIMINATION OF MDMA EXHIBITS USING IMPURITY PROFILES By Karlie Marie McManaman A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Forensic Science 2012 ABSTRACT EFFECTS OF EXTRACTION PROCEDURE AND GAS CHROMATOGRAPHY TEMPERATURE PROGRAM ON DISCRIMINATION OF MDMA EXHIBITS USING IMPURITY PROFILES By Karlie Marie McManaman MDMA impurity profiling is used to determine the synthesis route of exhibits, and can be used to link separate exhibits to each other and to clandestine laboratories based on the impurities present. Profiles are typically generated using a liquid-liquid extraction (LLE) procedure and gas chromatography-mass spectrometry (GC-MS). However, headspace solid phase microextraction (HS-SPME) is used in some forensic laboratories as an alternative extraction procedure, and has shown potential benefits. In addition, GC temperature programs vary between forensic laboratories, making comparisons of chromatographic data difficult. The aim of this research was to compare the LLE and HS-SPME procedures, as well as four GC temperature programs found in the literature, to determine their effects on the ability to associate replicates and discriminate exhibits of MDMA. Principal components analysis (PCA) was used to evaluate the data, and it was determined that HS-SPME provided increased discrimination of exhibits over LLE. While LLE provided higher precision of peak areas and, therefore, closer clustering of replicates, HS-SPME allowed for the extraction of several trace impurities, which provided additional discriminatory power over LLE. In the temperature program investigation, one temperature program was more successful than the others at associating and discriminating samples. However, this temperature program was significantly longer than the rest, therefore further investigation is needed to determine the beneficial parameters of this program and create a shorter, more practical temperature program. ACKNOWLEDGEMENTS I would like to first thank my advisor, Dr. Ruth Smith, for her guidance and motivation throughout this research project. Without her, none of this research would have been possible. He enthusiasm for forensic chemistry and her dedication to her students is truly inspiring. I would also like to thank my committee members, Mr. Frank Schehr and Dr. Jeremy Wilson, for their time and feedback on this thesis. I would like to thank the Michigan State Police for providing the MDMA exhibits used in this research. Additionally, I would like to thank, Elaine Dougherty, Kathy Boyer and analysts at the Bridgeport laboratory, who also provided training and guidance during my internship. I would like to thank Dr. Kathryn Severin and the staff at the Michigan State University Mass Spectrometry Facility for the assistance with instrumentation throughout this research. I would like to thank the MSU Forensic Science Program for funding to present this research at the MAFS 2011 Fall Meeting in Chicago, IL; the FSF Lucas Grant for funding for supplies and instrument time; and the NIJ/FSF Forensic Science Student Research Grant for funding for supplies, instrument time, and travel to present this research at the AAFS 2012 Annual Meeting in Atlanta, GA. I would like to thank the forensic chemistry students, especially John McIlroy, Christy Hay, Seth Hogg, Melissa Willard, Emily Riddell, Suzanne Towner, and Drew DeJarnette for their assistance, encouragement, and entertainment throughout my time at MSU. Lastly, I would like to thank my family and friends who have supported me without question throughout all of my endeavors. Their love and encouragement is what motivates me in everything I do. iii TABLE OF CONTENTS 1. Introduction............................................................................................................................1 1.1. MDMA History and Use.…………………….……...………………………………....…1 1.2. MDMA Synthesis …………………………………………………………………....…...2 1.3. Chemical Profiling Using Liquid-Liquid Extraction..........................................................3 1.4. Chemical Profiling Using Headspace Solid Phase Microextraction..................................7 1.5. Gas Chromatography Temperature Programs..................................................................11 1.6. Research Objectives..........................................................................................................13 References.........................................................................................................................15 . 2. Theory.....................................................................................................................................18 2.1. Liquid-Liquid Extraction..................................................................................................18 2.2. Headspace Solid Phase Microextraction..........................................................................18 2.3. Gas Chromatography........................................................................................................22 2.4. Mass Spectrometry...........................................................................................................29 2.5. Data Pretreatment.............................................................................................................33 2.5.1. Retention Time Alignment......................................................................................34 2.5.2. Normalization..........................................................................................................36 2.6. Principal Components Analysis........................................................................................36 References.........................................................................................................................40 3. Materials and Methods..........................................................................................................42 3.1. MDMA Exhibits...............................................................................................................42 3.2. Liquid-Liquid Extraction (LLE) Procedure......................................................................43 3.3. Headspace Solid Phase Microextraction (HS-SPME) Procedure.....................................44 3.4. Gas Chromatography-Mass Spectrometry (GC-MS) Analysis........................................44 3.5. Data Analysis 3.5.1. Total Ion Chromatograms (TICs)............................................................................47 3.5.2. Selected Compounds...............................................................................................49 References.........................................................................................................................51 4. Effect of Gas Chromatography Temperature Program on the Association and Discrimination of MDMA Exhibits Extracted using a Liquid-Liquid Extraction Procedure................................................................................................................................53 4.1. MDMA Exhibits Extracted by LLE and Analyzed using Temperature Program A.........53 4.2. MDMA Exhibits Extracted by LLE and Analyzed using All Temperature Programs .....61 4.3. Association and Discrimination of Exhibits Extracted by LLE based on PCA using Total Ion Chromatograms.................................................................................................62 4.3.1. Association and Discrimination of Exhibits Extracted by LLE and Analyzed using Temperature Program D based on PCA using TICs......................................67 4.3.2. Association and Discrimination of Exhibits Extracted by LLE and Analyzed using Temperature Program A based on PCA using TICs .....................................76 4.4. Association and Discrimination of Exhibits Extracted by LLE based on PCA using Selected Compounds ........................................................................................................82 iv 4.4.1. Association and Discrimination of Exhibits Extracted by LLE and Analyzed using Temperature Program A based on PCA using Selected Compounds............86 4.4.1.1. Comparison of Association and Discrimination of Exhibits Extracted by LLE and Analyzed using Temperature Program A based on TICs and Selected Compounds...............................................................................90 4.4.1.2. Further Investigation of Normalization Procedures on Exhibits Extracted by LLE and Analyzed using Temperature Program A...................................91 4.4.2. Comparison of Association and Discrimination of Exhibits Extracted by LLE and Analyzed using All Temperature Programs based on Selected Compounds ...93 4.5. Summary of Findings using LLE.....................................................................................97 References........................................................................................................................98 5. Effect of Gas Chromatography Temperature Program on the Association and Discrimination of MDMA Exhibits Extracted using a Headspace Solid Phase Microextration Procedure...................................................................................................100 5.1. MDMA Exhibits Extracted by HS-SPME and Analyzed using Temperature Program A.......................................................................................................................101 5.2. MDMA Exhibits Extracted by HS-SPME and Analyzed using Temperature Program D.......................................................................................................................109 5.3. Association and Discrimination of Exhibits Extracted by HS-SPME based on PCA using Selected Compounds....................................................................................109 5.3.1. Association and Discrimination of Exhibits Extracted by HS-SPME and Analyzed using Temperature Program D based on PCA using Selected Compounds............................................................................................................111 5.3.2. Association and Discrimination of Exhibits Extracted by HS-SPME and Analyzed using Temperature Program A based on PCA using Selected Compounds............................................................................................................116 5.4. Comparison of Association and Discrimination of Exhibits Extracted by LLE and HS-SPME based on PCA using Selected Compounds...................................................120 5.4.1. Comparison of Association and Discrimination of Exhibits Analyzed using Temperature Program D........................................................................................121 5.4.2. Comparison of Association and Discrimination of Exhibits Analyzed using Temperature Program A........................................................................................124 5.5. Summary of Findings using HS-SPME..........................................................................124 References.......................................................................................................................126 6. Conclusions & Further Work.............................................................................................128 6.1. Conclusions.....................................................................................................................128 6.1.1. Effect of Temperature Program on MDMA Impurity Profiles.............................129 6.1.2. Effect of Extraction Procedure on MDMA Impurity Profiles...............................131 6.2. Future Work....................................................................................................................132 v LIST OF TABLES Table 3.1: Physical characteristics of MDMA exhibits used in this study...................................43 Table 3.2: GC temperature programs investigated in this research..............................................45 Table 4.1: Selected compounds for data analysis of MDMA exhibits extracted by LLE using Temperature Programs A-D.........................................................................................85 Table 5.1: Selected compounds for data analysis of MDMA exhibits extracted by HS-SPME using Temperature Programs A and D.......................................................................111 vi LIST OF FIGURES Figure 1.1: Structure of MDMA with substitutions on the phenethylamine core circled in green.........................................................................................................................1 Figure 2.1: Headspace solid phase microextraction (HS-SPME) setup .......................................21 Figure 2.2: Components of a gas chromatography (GC) system..................................................23 Figure 2.3: Ideal peak shape (left), fronting peak (center), and tailing peak (right) with corresponding isotherms below..................................................................................28 Figure 2.4: Components of a mass spectrometry (MS) system....................................................30 Figure 2.5: Quadrupole mass analyzer .........................................................................................32 Figure 2.6: Electron multiplier......................................................................................................32 Figure 3.1: MDMA exhibits used in this study (a) T-17 (b) T-27 (c) T-29 (d) T-30 (e) MSU 900-01..........................................................................................................42 Figure 4.1: (a-e) TICs of exhibits after LLE and GC-MS analysis using Temperature Program A.............................................................................................................54-58 Figure 4.2: (a-b) Chemical structures of MDEA and MDDMA...................................................60 Figure 4.3: (a-d) TICs of Exhibit T-17 after LLE and GC-MS using Temperature Programs A-D.......................................................................................................63-66 Figure 4.4: PCA scores plot for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program D...................................................................................68 Figure 4.5: (a-b) Loadings plots for PCs 1 and 2 for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program D............................................69-70 Figure 4.6: (a-f) Effect of mean-centering on compound contribution: TIC showing caffeine peak in Exhibit T-17, mean-centered caffeine peak in Exhibit T-17, loadings for caffeine on PC1 and PC2, negative loading for caffeine on PC1 and PC2 in Exhibit T-17.......................................................................................72-74 Figure 4.7: PCA scores plot for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program A...................................................................................78 Figure 4.8: (a-b) Loadings plots for PCs 1 and 2 for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program A............................................79-80 vii Figure 4.9: Misaligned caffeine peak in replicates of Exhibits T-27 and T-30 after LLE and GC-MS analysis using Temperature Program A. Alignment was performed using a correlation optimized warping algorithm with a warp of 2 and a segment size of 150. Inset: Scores plot obtained for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program A........................................................84 Figure 4.10: PCA scores plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program A. .......87 Figure 4.11: Loadings plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program A.........88 Figure 4.12: PCA scores plot based on selected compounds after logarithmic normalization for five MDMA exhibits extracted by LLE and analyzed by GC-MS using Temperature Program .........................................................................92 Figure 4.13: PCA scores plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program B.........94 Figure 4.14: PCA scores plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program C.........95 Figure 4.15: PCA scores plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program D.........96 Figure 5.1: (a-e) TICs of exhibits after HS-SPME and GC-MS analysis using Temperature Program A....................................................................................102-106 Figure 5.2: TICs of Exhibit T-17 after HS-SPME (top) and LLE (bottom) and GC-MS analysis using Temperature Program A...................................................................108 Figure 5.3: TIC of Exhibit T-17 after HS-SPME and GC-MS analysis using Temperature Program D................................................................................................................110 Figure 5.4: PCA scores plot for five MDMA exhibits after HS-SPME and GC-MS analysis using Temperature Program D...................................................................112 Figure 5.5: PCA loadings plot for five MDMA exhibits after HS-SPME and GC-MS analysis using Temperature Program D...................................................................113 Figure 5.6: PCA scores plot for five MDMA exhibits after HS-SPME and GC-MS analysis using Temperature Program A...................................................................117 Figure 5.7: PCA loadings plot for five MDMA exhibits after HS-SPME and GC-MS analysis using Temperature Program D...................................................................118 viii Figure 5.8: PCA scores plots based on selected compounds for five MDMA exhibits extracted using LLE (left) and HS-SPME (right) and analyzed by GC-MS using Temperature Program D.................................................................................122 Figure 5.9: PCA scores plots based on selected compounds for five MDMA exhibits extracted using LLE (left) and HS-SPME (right) and analyzed by GC-MS using Temperature Program A.................................................................................123 ix KEY TO ABBREVIATIONS CAR Carboxen CHAMP Collaborative Harmonization of Methods for Profiling of Amphetamine Type Stimulants COW Correlation optimized warping DC Direct current DVB Divinylbenzene EI Electron ionization GC Gas chromatography GC-MS Gas chromatography-mass spectrometry HCA Hierarchical cluster analysis HS-SPME Headspace solid phase microextraction LLE Liquid-liquid extraction LR Likelihood ratio MA Methamphetamine MDDMA 3,4-methylenedioxydimethylamphetamine MD-DMB 3,4-methylenedioxy-N,N-dimethylbenzylamine MDEA 3,4-methylenedioxy-N-ethylamphetamine MDMA 3,4-methylenedioxymethamphetamine MDP2P 3,4-methylenedioxyphenyl-2-propanone MDP2P-OH 3,4-methylenedioxyphenyl-2-propanol MS Mass spectrometry NIST National Institute of Standards and Technology NRIPS Japan’s method for profiling methamphetamine x ONCB Thailand’s method for profiling methamphetamine P2P phenyl-2-propanone PC1 First principal component PC2 Second principal component PCA Principal components analysis PDMS Polydimethylsiloxane PDMS/DVB Polydimethylsiloxane/divinylbenzene PPMC Pearson product moment correlation RF Radio frequency RSD Relative standard deviation SPME Solid phase microextraction TIC Total ion chromatogram xi Chapter 1 Introduction 1.1 MDMA History and Use 3,4-methylenedioxymethamphetamine (MDMA) is a stimulant and hallucinogen that became federally regulated under the Controlled Substances Act in 1985, after gaining popularity 1 as the active ingredient in “ecstasy” tablets. MDMA (Figure 1.1) is a phenethylamine with a methylenedioxy substitution on the aromatic ring, which produces hallucinogenic effects; a methyl group substitution on the nitrogen atom, which doubles the potency of the drug; and a methyl group substitution on the α-carbon, which stimulates the central nervous system and 2 suppresses appetite. α-carbon methyl group methylenedioxy Figure 1.1: Structure of MDMA with substitutions on the phenethylamine core circled in green. For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this thesis. MDMA is listed as a Schedule I hallucinogenic compound, which means it has a high potential for abuse and no current medical use. Although the drug was originally created for 1 pharmaceutical development in Germany, it became a substance of abuse in the 1970s after 1 psychiatrists began treating patients with this drug, even though it had not been tested by the US Food and Drug Administration. The increase in abuse of MDMA throughout the 1970s and ’80s and lack of verified medical use prompted the government to schedule this drug in the Controlled Substances Act. Although MDMA is now regulated, use of illicit substances is still a growing trend in this country, especially among young adults. Surveys such as the Monitoring the Future survey 3 provide statistics on current and past abuse of illegal substances, including MDMA. Even though the majority of students who participated in the 2010 Monitoring the Future Survey admit to being aware of the harmfulness of MDMA, it continues to be abused by youths. An estimated 1.1 million people tried MDMA for the first time in 2009, most of them middle school through college age, which was a significant increase from the year before. This survey also revealed that th there has been a significant increase in MDMA use for all age groups surveyed (8 , 10 th 12 grades) since the survey five years prior, with 3.3% of 8 th grade, 6.4% of 10 th th and grade and th 7.3% of 12 grade students reporting taking the substance at least once in their lives. 1.2 MDMA Synthesis MDMA tablets are manufactured in clandestine laboratories throughout this country and in many European nations. Each lab produces tablets that are slightly different based on the method used to synthesize the MDMA and the processing techniques used to produce the final tablet. The synthesis method used depends on the ingredients available to the clandestine chemist. The most common method of MDMA production is through the reductive amination of 3,4-methylenedioxyphenyl-2-propanone (MDP2P), and three different variations of the reductive 2 4 amination method are used: the high-pressure, cold, or aluminum amalgam methods. The highpressure method uses hydrogen under increased pressure with platinum as a catalyst. In the cold method, sodium borohydride is the reducing agent and the synthesis takes place at cooler temperatures by storing the reaction mixture in a freezer to prevent reduction of MDP2P. In the aluminum amalgam method, aluminum foil and mercury chloride are used as the reducing 5 agent. Since MDP2P itself is now a List 1 Chemical of Concern in the United States, this compound is also produced illegally. The most common methods of MDP2P production are oxidation of isosafrole in acid, and reductive amination of piperonal using substances such as 4 MDP-2-nitropropene or methylamine. With all of the variations in synthesis methods, the MDMA produced has obvious differences. The aim of chemical profiling MDMA is to identify impurities, byproducts, and additives in the tablets. This chemical profile will vary among tablets because of the differences in the ingredients available and the synthesis method used. Having these profiles allows authorities to identify the synthesis method, which can be used to link tablets back to a common 4-14 clandestine laboratory and potentially to a common production batch. 1.3 Chemical Profiling Using Liquid-Liquid Extraction Researchers have investigated chemical profiling of illicit substances to determine synthetic route, aiming to identify links among exhibits, for many years. Typically, a liquidliquid extraction (LLE) procedure is used to extract the compounds and the resulting extract is analyzed by gas chromatography-mass spectrometry (GC-MS). The LLE procedure has been thoroughly investigated and optimized for numerous illicit compounds, including MDMA and 3 methamphetamine. Van Deursen et al. optimized a LLE procedure specifically for MDMA 6 profiling in 2006. In this investigation, the authors compared extraction solvents, such as alkanes and toluene; phosphate buffers of different concentrations (0.066 M and 0.33 M) and pH (1.0, 5.0, 7.0 and 8.1); and filtering membranes such as regenerated cellulose and PTFE, with varying pore sizes. Based on the abundance of impurities and the repeatability of the extraction, it was ultimately determined that toluene should be used as the extraction solvent, with a 0.33 M phosphate buffer at pH 7.0, and a regenerated cellulose membrane with a pore size of 0.45 µm. This method was then used as the standard extraction procedure in the Collaborative Harmonization of Methods for Profiling of Amphetamine Type Stimulants (CHAMP) project in 7 Europe. The aim of the CHAMP project was to standardize MDMA profiling procedures to allow for better future comparisons of the resulting impurity profiles among laboratories. Various other studies have used LLE and GC-MS to generate chemical profiles of MDMA tablets with the aim of identifying the synthesis route used. Gimeno et al. related many 8 of the impurities present to specific synthesis routes. For example, the presence of 3,4methylenedioxy-N,N-dimethylbenzylamine indicated synthesis of MDMA via the reductive amination of piperonal by dimethylamine, while 3,4-methylenedioxy-N-ethyl-N- methylbenzylamine indicated reductive amination of piperonal by ethylmethylamine, and 3,4methylenedioxy-N-methylbenzylamine indicated reductive amination of piperonal by methylamine. In another study investigating synthetic route determination, Swist et al. manufactured MDP2P, the most popular precursor of MDMA, using methods commonly used in clandestine laboratories. 4 Using LLE and GC-MS, different chemical markers for each method were 4 identified in the synthesized MDP2P. Then, the MDP2P was used to synthesize MDMA via reductive amination using sodium borohydride as the reducing agent. The MDMA was subjected to LLE and the resulting extract was analyzed by GC-MS to generate chemical profiles. These profiles were then used to successfully determine the likely method used to synthesize the MDP2P precursor. These studies were helpful in identifying the impurities that can be used to determine synthesis methods, but there was no attempt to link MDMA exhibits to common sources based on the impurities present. Cheng et al. used statistical procedures, along with recent knowledge of route specific 9 impurities, to relate MDMA exhibits based on the impurities present. In one study, 89 ecstasy tablets were analyzed using LLE and GC-MS and the profiles were compared using hierarchical cluster analysis (HCA), which clusters samples together based on chemical similarities. Using HCA, the tablets were separated into four groups at a similarity index of 0.5. The tablets were grouped based on differences in the synthesis route, which was verified by the identification of route specific impurities in the corresponding chemical profiles. To further investigate the use of HCA to group tablets based on chemical profiles, four test samples were prepared from one of the afore-mentioned tablets which was of high purity, 9 containing nearly 100% MDMA. Samples from the tablet were cut with two different cutting agents in different proportions, reducing the purity of the MDMA to 50% and 25%. The samples were extracted using LLE and analyzed by GC-MS to generate the profiles, which were then assessed using HCA. Test samples containing the same cutting agents were closely clustered, indicating that the additives present have a greater impact on clustering than the purity of the sample. All four test samples were clustered in the same group as the corresponding tablet, meaning that the original source of the samples was still identifiable. This study was a step 5 forward in MDMA chemical profiling, but it only linked exhibits by common synthesis methods and not necessarily common clandestine laboratories, which would be the next step in aiding police identify drug trafficking patterns. In order to link different exhibits by possible production laboratories, the statistical methods of analysis need to be investigated further, and a larger population of samples from around the world should be used. A collaborative effort of several European forensic laboratories sought to identify links among MDMA exhibits based on chemical profiles and physical characteristics of the tablets. 6 Using the extraction procedure optimized by van Deursen et al., laboratories in four different countries extracted the same 26 MDMA exhibits, as well as an additional 80 samples collected in each of their own countries, using LLE. All extracts were analyzed by GC-MS using a 7 temperature program from The Netherlands Forensic Institute. Thirty-two common compounds were identified in the 26 exhibits, and the eight compounds that were most variable in abundance were selected as the most discriminating compounds. Pearson product moment correlation (PPMC) coefficients and principal components analysis (PCA) were used to assess association and discrimination of tablets based on all 32 compounds, as well as the eight most discriminating compounds. The PCA scores plots were assessed visually, and PPMC coefficients, which were modified to fit a scale of 0-100, were given a threshold of 2.69 or less to indicate that samples were statistically similar with a false positive rate of 2% or less. Of the 26 exhibits, all exhibits were discriminated using all 32 compounds, while 99% of the exhibits were associated and discriminated correctly using only the eight selected compounds. The same statistical procedures were then applied to the 80 exhibits collected in the respective countries and links were discovered between separate exhibits based on consistencies in their profiles and modified Pearson values. 6 The above study continued by investigating additional statistical procedures to link exhibits based on physical characteristics (e.g. diameter, thickness and weight), as well as 10 percent purity of MDMA in the tablets. Using both PPMC coefficients and Euclidean distances, a likelihood ratio (LR) was calculated to determine the likelihood that two tablets originated from the same production batch. The ability to associate and discriminate tablets based on each physical characteristic individually, as well as all three characteristics, was assessed. Using all three physical characteristics to relate samples to each other provided higher LRs than each characteristic individually. While the three physical characteristics gave LRs that were determined to be acceptable, the discriminatory power was greatly improved when percent purity was added, increasing the LRs nearly 10-fold in some cases. These collaborative works are very useful to associate and discriminate MDMA exhibits seized in different countries. However, no similar collaborative works have been instigated in the U.S., which is necessary to better understand drug-trafficking patterns throughout this country. 1.4 Chemical Profiling Using Headspace Solid Phase Microextraction While LLE is the most common extraction method used in impurity profiling, there are some drawbacks to this procedure. The extraction is lengthy because of the number of steps involved, a large mass of sample is necessary, often equivalent to up to one entire tablet, and there are several potentially harmful chemicals used. 11 Due to these limitations, an alternative to this procedure is headspace solid phase microextraction (HS-SPME). Compared to the typical LLE method, HS-SPME is less time consuming in sample preparation, requires less sample (as little as ¼ tablet or less) per extraction, is a relatively non-destructive method, and is less 7 hazardous since no chemicals are used, yet it has been demonstrated to be just as effective for impurity extraction. Recent literature shows great potential for linking MDMA tablets when HS-SPME is used to extract the compounds. Bonadio et al. compared HS-SPME to LLE in terms of the identity and number of compounds extracted from the same MDMA tablets. 11 While the number of compounds extracted was relatively similar for both extraction procedures (31 compounds for HS-SPME and 32 for LLE), the extraction of other compounds, such as lubricants, only occurred in HS-SPME because these compounds are removed during the filtration step in the LLE procedure. While these other compounds do not give information on the synthesis route, they are used in the tablet production process and are important in linking tablets back to a specific clandestine laboratory. In the comparison study by Bonadio et al., the ability of the HS-SPME and LLE procedures to correctly associate and discriminate tablets was assessed. Four datasets were evaluated, composed of either all of the selected reproducible compounds (31 for HS-SPME and 32 for LLE), or the most discriminating reproducible compounds (10 for HS-SPME and 8 for LLE). The authors optimized and discussed normalization procedures, and the data were analyzed using PPMC coefficients, again modified to a scale of 0-100. Each method was efficient in associating and discriminating tablets correctly, regardless of the number of compounds used. After observing the potential benefits of using HS-SPME for MDMA chemical profiling, Bonadio et al. developed and optimized the extraction procedure specifically for this use. 12 Four different types of HS-SPME fiber coatings were tested: polydimethylsiloxane (PDMS), carboxen (CAR)/PDMS, PDMS/divinylbenzene (DVB), and DVB/CAR/PDMS. Different masses (10, 40, 8 and 100 mg) of homogenized MDMA tablets were extracted at different temperatures (60, 80, 90 and 100 °C) for different times (15, 30, 45, and 60 min). Of the four fibers, PDMS/DVB had the strongest affinity for the eight target compounds, which were selected based on repeatability and discriminating power. Ideally, the smallest mass of sample should be used, as long as it is enough to allow the less abundant compounds to be extracted. However, using only 10 mg of sample, some of the target compounds were only extracted in low abundance or were not extracted at all. Since there was no appreciable difference in the number or abundance of compounds extracted using 40 mg or 100 mg of sample, the lower mass was deemed optimal. At high temperatures (e.g. 100 °C) and/or with longer extraction times (e.g. 60 min), the fiber was overloaded, causing many compounds to co-elute and hence, making identification more difficult. However, using low extraction temperatures (60 °C or less), many less volatile compounds were not extracted at sufficiently high concentration to be detected. Based on these considerations, the optimal HS-SPME procedure used a PDMS/DVB fiber and 40 mg of sample, with an extraction temperature of 80 °C for 15 min. The optimization of the HS-SPME extraction procedure is very beneficial for future work and can be used as a standard for extracting ecstasy tablets for comparison. Researchers have also compared HS-SPME and LLE procedures for chemical profiling of other controlled substances. Kuwayama et al. used both extraction procedures to profile 69 methamphetamine samples seized throughout Japan and 42 samples from Thailand. 13 The LLE extracts were analyzed by GC-MS while the HS-SPME extracts were analyzed by GC with a flame ionization detector. The resulting chromatograms were compared using HCA to statistically assess similarities among samples. The HS-SPME procedure was more successful than LLE in distinguishing high purity samples from one another. This is mainly because 9 compounds present at lower abundance are easily visible in the HS-SPME profiles, but are not observed in the profiles obtained with LLE because LLE is more efficient at extracting the controlled substance (in this case, methamphetamine) than impurities. Pre-concentration of impurities on the SPME fiber also allows for more impurities to be extracted at higher abundances than with LLE. The combination of LLE and HS-SPME profiles allowed for the successful discrimination of all exhibits using HCA. Lee et al. also compared the LLE and HS-SPME procedures for chemical profiling of methamphetamine using 48 exhibits seized in Korea. 14 In this study, the authors determined that less volatile compounds, such as a methamphetamine dimer, were only extracted using LLE and were not observed in exhibits extracted using HS-SPME. However, several volatile components, such as diphenylketone and many unknown compounds, were clearly separated in the HS-SPME chromatograms, while in the LLE chromatograms they were often masked by the broad amphetamine peak which resulted from a degradation of methamphetamine with LLE but not HS-SPME. HCA was performed on the chromatograms of the 48 exhibits obtained using LLE and 14 again on the chromatograms obtained using HS-SPME. The samples were clustered into five groups irrespective of extraction procedure; however, the exhibits within each group showed some variation depending on extraction procedure. In these cases, the exhibits were subclassified such that the identities matched between the two procedures. For example, one of the HS-SPME clusters contained 10 methamphetamine exhibits. When these were cross-matched, nine of the exhibits were in the same LLE cluster, while one exhibit was in a different LLE cluster. Similar to the findings reported by Kuwayama et al., it was determined that the combination of LLE and HS-SPME chemical profiles provides more accurate discrimination of 10 13,14 exhibits than either extraction procedure alone. While this may be true for methamphetamine samples, similar studies comparing extraction procedures should be conducted for MDMA. Illicit MDMA tablets are typically not as high purity as methamphetamine, and the extra compounds, along with the less abundant controlled substance, in the tablets may allow for better discrimination using one extraction procedure over the other. 1.5 Gas Chromatography Temperature Programs All of the afore-mentioned studies were good resources to identify many of the compounds found in MDMA tablets; however, the profiles from studies like these are not readily comparable to each other because different instrument parameters and oven temperature programs were used to analyze the extracts. This makes comparisons difficult because the same compounds will elute at different retention times, depending on the temperature program. To make comparisons of impurity profiles among labs more valid, a standardized GC temperature program for MDMA impurity profiling should be investigated. While slower ramp rates (e.g. 5 °C/min) and longer hold times (e.g. 15 minutes) may provide better separation of compounds with similar retention times, these conditions may also cause broadening of peaks due to the longer time the compounds spend in the column. Broadened peaks are not ideal in chromatography since they can mask lower abundance peaks eluting at similar retention times. Baerncopf et al. studied the effect of different GC temperature programs on the ability to associate and discriminate five different diesel samples. 15 In this study, six GC temperature programs, ranging from 15 to 113 minutes, with both one- and two-step ramps were adapted from the literature. The resulting total ion chromatograms were then subjected to analysis using PPMC coefficients and PCA. Association of replicates of the same diesel with discrimination of 11 the five different diesels was observed, irrespective of temperature program. While this study focused on diesel, similar data analysis procedures can be applied to the chemical profiles of MDMA tablets to investigate the effect of GC temperature program on association and discrimination of the tablets. Kuwayama et al. compared GC instrument parameters, as well as extraction procedures, for methamphetamine profiling. 13 Methamphetamine samples seized in Japan and Thailand were extracted using LLE procedures and then analyzed using two different GC methods commonly used in these countries. While both GC methods used very similar temperature programs, there were other notable differences in the methods. The method used in Japan (referred to as NRIPS) used an injection temperature that was 40 °C lower than the method used in Thailand (referred to as ONCB). Additionally, the NRIPS method used a transfer line temperature that was 20 °C higher and a final hold time that was 5 min shorter than the ONCB method. There were also differences in the flow rates and columns used for the analysis. In the NRIPS method, the carrier gas was maintained at a constant flow rate of 2 mL/min, while in the ONCB method, the carrier gas was maintained at constant pressure. While both methods used non-polar stationary phases, there were differences in the phase thickness (1.0 µm in the NRIPS method compared to 0.33 µm in the ONCB method). Phase thickness affects mass transfer of compounds between the carrier gas and stationary phase. Hence, this can lead to differences in retention time of compounds, and could allow for co-elution of compounds. After analyzing liquid-liquid extracts of the methamphetamine samples using each GC program, the NRIPS method used in Japan demonstrated improved chromatographic separation and greater success in detecting compounds, both of which are essential in chemical profiling. 12 However, along with different GC methods, the actual LLE procedures used were also 13 different. Although the buffer and solvent used in each case were similar, the sample concentration was nearly double in the ONCB method compared to the NRIPS method. The authors noted a large methamphetamine peak with the ONCB method that made it difficult to analyze minor impurities in high purity samples because of co-elution. This is likely a result of the increased concentration, not the GC parameters. Therefore, to truly evaluate the GC methods, the same extraction procedure should have been performed. A similar study comparing temperature programs and extraction procedures, with statistical analysis of the results, should be conducted for MDMA exhibits. A study such as this is essential in establishing a standard procedure for MDMA profiling. 1.6 Research Objectives The objectives of this research are as follows: (1) Use statistical procedures to investigate the effect of extraction procedures on association and discrimination of MDMA exhibits based on chemical profiles. (2) Use statistical procedures to investigate the effect of GC temperature programs on association and discrimination of MDMA exhibits based on chemical profiles. To address the first objective, impurities from five different MDMA exhibits will be 6,12 extracted using LLE and HS-SPME procedures previously optimized in the literature. The extracts will then be analyzed by GC-MS, using a temperature program available in the literature. Principal components analysis will be used to assess the effect of the extraction procedure on the ability to associate samples from the same exhibit and differentiate samples 13 from different exhibits. The second objective will be addressed by analyzing MDMA extracts, obtained using both LLE and HS-SPME, using different GC temperature programs, each taken 4,6,8,12 from the literature. Again, the effect of each program on the ability to associate samples from the same exhibit and discriminate exhibits will be investigated using PCA. Determining the effects of extraction procedure and temperature program on the chemical profiles is an essential step in creating a standard method for MDMA profiling. Currently, MDMA profiles are not comparable between forensic laboratories due to the different extraction and analysis procedures used. Establishing a standard extraction procedure and GC temperature program will allow for comparisons of MDMA profiles between forensic laboratories and will greatly benefit law enforcement in their ability to link exhibits to one another and to clandestine laboratories. 14 REFERENCES 15 REFERENCES 1. National Institute on Drug Abuse. Research Reports: MDMA (Ecstasy) Abuse. National Institutes of Health. Bethesda, MD. March, 2006. Available at: http://www.drugabuse.gov/publications/research-reports/mdma-ecstasy-abuse. (Accessed March, 2011) 2. Smith, FP, Siegel JA, editors. Handbook of Forensic Drug Analysis. Burlington, MA: Elsevier Academic Press, 2005. 3. National Institute on Drug Abuse DrugFacts: MDMA (Ecstasy). National Institutes of Health. Bethesda, MD. Revised December, 2010. Available at: http://www.drugabuse.gov/Infofacts/ecstasy.html. (Accessed March, 2011) 4. Swist M, Wilamowski J, Zuba D, Kochana J, Parczewski A. Determination of synthesis route of 1-(3-4-methylenedioxyphenyl)-2-propanone (MDP2P) based on impurity profiles of MDMA. Forensic Sci Int 2005;149:181-192. 5. Koper C, van den Boom C, Wiarda W, Schrader M, de Joode P, van der Peijl G, Bolck A. Elemental analysis of 3,4-methylenedioxymethamphetamine (MDMA): A tool to determine the synthesis method and trace links. Forensic Sci Int 2007;171:171-179. 6. Van Deursen M, Poortman-van der Meer A. Organic impurity profiling of 3,4methylenedioxymethamphetamine (MDMA) tablets seized in the Netherlands. Sci Justice 2006;46:135-152. 7. Weyermann C, Marquis R, Delaporte C, Esseiva P, Lock E, Aalberg L, Bonzenko Jr. J, Dieckmann S, Dujourdy L, Zrcek F. Drug intelligence based on MDMA tablets data I. Organic impurities profiling. Forensic Sci Int 2008;177:11-16. 8. Gimeno P, Besacier F, Chaudron-Thozet H, Girard J, Lamotte A. A contribution to the chemical profiling of 3,4-methylenedioxymethamphetamine (MDMA) tablets. Forensic Sci Int 2002;127:1-44. 9. Cheng J, Chan M, Chan T, Hung M. Impurity profiling of ecstasy tablets seized in Hong Kong by gas chromatography-mass spectrometry. Forensic Sci Int 2006;162:87-94. 10. Bolck A, Weyermann C, Dujourdy L, Esseiva P, van den Berg J. Different likelihood ratio approaches to evaluate the strength of evidence of MDMA tablet comparison. Forensic Sci Int 2009;191:42-51. 16 11. Bonadio F, Margot P, Delémont O, Esseiva P. Headspace solid-phase microextraction (HS-SPME) and liquid-liquid extraction (LLE): Comparison of the performance in classification of ecstasy tablets (Part 2). Forensic Sci Int 2008;182:52-56. 12. Bonadio F, Margot P, Delémont O, Esseiva P. Optimization of HS-SPME/GC-MS analysis and its use in the profiling of illicit ecstasy tablets (Part 1). Forensic Sci Int 2009;187:73-80. 13. Kuwayama K, Inoue H, Phorachata J, Kongpatnitiroj K, Puthaviriyakorn V, Tsujikawa K, Miyaguchi H, Kanamori T, Iwata Y, Kamo N, Kishi T. Comparison and classification of methamphetamine seized in Japan and Thailand using gas chromatography with liquid-liquid extraction and solid-phase microextraction. Forensic Sci Int 2008;175:8592. 14. Lee J, Park Y, Yang W, Chung H, Choi W, Inoue H, Kuwayama K, Park J. Crossexamination of liquid-liquid extraction (LLE) and solid-phase microextraction (SPME) methods for impurity profiling of methamphetamine. Forensic Sci Int 2012;215:175-178. 15. Baerncopf J, McGuffin V, Waddell Smith R. Effect of gas chromatography temperature program on the association and discrimination of diesel samples. J Forensic Sci 2010;55:185-192. 17 Chapter 2 Theory 2.1 Liquid-Liquid Extraction Extraction procedures such as liquid-liquid extraction (LLE) are used to separate 1 compounds in a complex sample based on their solubilities. In LLE, a water-based solvent and an organic solvent are typically used because these two liquids are immiscible. The sample is first dissolved in an aqueous solvent, and then extracted into an organic solvent, such as toluene, which can be used for analysis by gas chromatography-mass spectrometry (GC-MS). Since the two liquids do not mix, two distinct layers are formed, allowing for a clean separation of the desired organic layer from the aqueous layer. Only substances which are soluble in the organic solvent are extracted into that layer. Therefore, solvents must be chosen for each sample based on the compounds desired to be extracted. Often, the aqueous solvent is a buffer, such as the phosphate buffer that was used in this 1 research. Buffers act to keep a constant pH by resisting change when either an acid or base is added. In this research, a buffer with a neutral pH of 7.0 was used to correct for the influence of any compounds that may have been either slightly basic or acidic. Weakly basic compounds are extracted slightly more efficiently at a higher pH, but a pH above 7 leads to an excess of MDMA 2 being extracted. Therefore, a neutral pH is preferable for the extraction of MDMA tablets. 2.2 Headspace Solid Phase Microextraction Solvent-free sampling methods are often beneficial due to their ease of use, low cost, and 3 reduced safety hazards. Solid phase microextraction (SPME) is a type of solvent-free extraction that was invented in the early 1990s by Janusz Pawliszyn. This sampling method is used to 18 minimize sample preparation and avoid destruction of the sample. In SPME, a fiber coated in a sorbent material adsorbs the compounds until an equilibrium is reached between the sample and the fiber coating. The coating can be solid, liquid, or a combination of the two. Equation 2.1 describes this equilibrium to give the number of moles, n, of compound that is extracted from the sample onto the fiber coating: 3 Equation 2.1 where Kfs is the distribution constant for the fiber coating/sample matrix, which is the ratio of equilibrium concentration of compound on the fiber (Cf) to compound in the sample (Cs); Vf is the volume of the fiber coating; Vs is the volume of the sample; and C0 is the initial concentration of compound in the sample. There are several types of fiber coatings available, varying in material and thickness, and 3 each coating leads to a different Kfs. Therefore, both the composition and thickness of the coating must be chosen specifically for each application. For example, a thicker coating (e.g. 100 µm) means a greater coating volume and, therefore, increased extraction of compound, but it also leads to longer extraction and desorption times. Thin coatings are more effective at extracting low volatility compounds, while thicker coatings are used for extracting high volatility compounds. The most practical and common coating is a liquid polydimethylsiloxane (PDMS), which is popular because of its ability to withstand high temperatures (up to 300 °C), as well as its ability to extract compounds with wide ranges of volatility and polarity. This is achieved by altering the thickness of the coating or combining it with another coating. Divinylbenzene (DVB) is often used with PDMS as a solid/liquid combination coating because of its uniform structure, 19 which allows for a wider range of molecular weights of compounds to be adsorbed. In this research, headspace SPME (HS-SPME) was used to extract impurities from the MDMA tablets. Using the headspace approach, the sample is often heated in a vial to increase vapor pressure and volatility of the compounds, therefore increasing the concentration of 3 compounds in the headspace where they can be directly extracted onto the fiber coating. Using this method, volatile compounds pass through the air before reaching the fiber, thus protecting the fiber from contact with the sample matrix and reducing the amount of cleaning of the fiber after desorption. Figure 2.1 depicts the HS-SPME extraction setup used in this research, with a sample in a closed vial and the fiber inserted through the cap of the vial. The compounds move from the sample into the headspace and are extracted from the headspace onto the fiber. When using HS-SPME, Equation 2.1 is modified to account for the compounds in the headspace, as shown in Equation 2.2. 3 Equation 2.2 Kfs, which represented the distribution constant between the fiber and sample matrix in Equation 2.1, now requires two terms: Kfh and Khs. Kfh is the distribution constant for the fiber and headspace, and is defined as the ratio of equilibrium concentration of compound on the fiber (Cf) to compound in the headspace (Ch), and Khs is the distribution constant for the headspace and sample. Vh refers to the volume of the headspace. 20 fiber holder fiber sample vial headspace, moving analytes sample Figure 2.1: Headspace solid phase microextraction (HS-SPME) setup. Alterations to the basic extraction procedure may be made to increase the number of moles of compound extracted onto the fiber. Dissolving the sample in a buffer can change the distribution constants between the sample and headspace or fiber, and the pH of the buffer can be 3 altered to determine the optimal distribution ratios. Oftentimes, altering the pH can increase the extraction of certain compounds. For example, increasing the pH to allow for more extraction of slightly basic compounds. Changing the temperature of the sample can also affect the distribution ratios, but without using any chemicals. Heating the sample vial in a water bath, for example, increases the volatility of the compounds, and therefore increases the concentration of compounds in the headspace, leading to more compounds being extracted onto the fiber. However, since increasing the temperature increases the concentration of compounds extracted, it can cause overloading of the GC column and can cause co-elution of compounds. While there are many advantages to HS-SPME, there are several potential disadvantages. Pre-concentrating the fiber allows for the extraction of trace compounds, but it also increases the concentration of high abundance compounds, which again can cause overloading of the GC 21 column. Therefore, the extraction time and temperature must be individually optimized for different sample types to determine the parameters that form a compromise between extracting trace compounds and overloading abundant compounds. Overall, HS-SPME is a less reproducible extraction procedure than LLE because of the variability in the concentration of compounds extracted onto the fiber. This can cause difficulties during data analysis, as peaks in the chromatograms do not always align, and the precision of peak areas between replicates is, in general, lower than those obtained using LLE. 2.3 Gas Chromatography Chromatography is a separation technique that uses two phases: a stationary phase which is bonded to a vessel, such as a column, and a mobile phase that flows over the stationary phase. 1 As the sample mixture moves through the column, individual compounds are separated over time based on the amount of interactions of each compound with the stationary phase. There are a variety of chromatography procedures, such as thin layer chromatography, liquid chromatography, and gas chromatography. In forensic laboratories, gas chromatography (GC) is the most common procedure used for a variety of applications, such as drug analysis and fire debris analysis. A schematic of a typical GC instrument is shown in Figure 2.2. A cylinder containing the carrier gas is connected to the GC, where the gas combines with the sample from the injector 1 port and flows onto the column. The sample and gas are carried through the column, which contains the stationary phase and is housed inside an oven. The compounds in the sample are separated in the column and reach the detector at different times. A chromatogram is shown on the monitor, which displays a peak for each compound as it elutes from the column. 22 injector port monitor detector column oven carrier gas tank Figure 2.2: Components of a gas chromatography (GC) system. For analysis, the sample mixture is usually dissolved in a liquid, and the solution is 1 injected into the heated injector port of the GC. Injection volumes are typically 1 µL, and, for very concentrated samples, much of this volume is discarded to prevent overloading the column. This is known as a split injection and a user-set split ratio determines how much of the sample goes onto the GC column and how much is discarded. For example, using a 50:1 split ratio, 50 parts of the sample/carrier gas mixture are discarded for every 1 part that is carried onto the GC column. In contrast, using a splitless injection, all the injected sample is transferred onto the column for analysis. Since the samples in this research contained several trace impurities, a splitless injection was used. 23 The temperature of the injector port is selected by the user, and needs to be sufficiently high (typically ~ 250 °C) to completely volatilize the sample to ensure that all of the sample is in 1 the gas phase and can enter the GC column. In the injector port, the carrier gas (i.e., the mobile phase) carries the sample onto the GC column. The carrier gas must be of high or ultra-high purity so that no additional compounds from the gas are observed in the sample. Ultra-high purity helium, hydrogen, or nitrogen are common choices because, as well as being high purity, they are light and inert, and, therefore, will not react with the sample. Helium is the preferred gas when a mass spectrometer is used as the detector in GC because of its stability, its high viscosity and diffusivity that allow for efficient chromatographic separation, as well as its inertness and high purity as described above. The flow rate of the carrier gas is also set by the user, and is generally 1 mL/min when a mass spectrometer is used as the detector. GC columns are typically open tubular columns, which are fused silica with a polyimide coating. The inner walls are coated with the stationary phase, such as polyethylene glycol or 1 dimethyl polysiloxane. The length of a GC column used for drug analysis is typically around 30 m, the internal diameter around 0.25 mm, and the stationary phase thickness around 0.25 µm. Stationary phases range in polarity, and are chosen specifically for the samples of interest. Combination stationary phases are available to account for a range of polar compounds. For example, in this research, the stationary phase consisted of 5% diphenyl and 95% dimethyl polysiloxane. This is a low polarity stationary phase, but it is more polar than 100% dimethyl polysiloxane. This stationary phase is recommended for drugs of abuse because of its slight polarity, which makes it efficient at analyzing the controlled substances, which are slightly polar. This column is also an ultra-low bleed column, meaning there is limited deterioration of the column observed in the chromatograms at high temperatures. 24 The GC column in housed inside an oven, the temperature of which can be changed 1 during the analysis. A compound will elute from the column (and reach the detector) when the oven temperature is near the boiling point of the compound. The time taken for the compound to pass through the column and reach the detector is known as the retention time of the compound. The temperature program may be run isothermal, meaning that the temperature remains constant throughout the analysis. This allows for the better separation of compounds that have similar boiling points (other factors effecting retention time will be discussed later). However, isothermal analysis leads to longer retention times since separation of compounds is based purely on interactions with the stationary phase and is not aided by temperature. For sample mixtures containing compounds with a wide range of boiling points, the oven temperature can be varied during the analysis. Typically, the oven temperature is ramped from a low initial temperature to a high final temperature. Faster ramp rates (e.g. 20 °C/min) result in a shorter overall analysis, but may not allow sufficient resolution between compounds that have similar retention times. If the sample contains some compounds that have similar boiling points and hence, would have similar retention times, then a slower ramp rate (e.g. 5 °C/min) may be used or the oven may be held at a specific temperature during the temperature program. In addition to boiling point, the retention time of a compound is affected by its 1 interactions with the stationary phase. While the sample mixture is being carried through the column, compounds in the mobile phase interact with both the surface (via adsorption interactions) and the bulk (via absorption interactions) of the stationary phase. The extent of the interactions determines the retention time of the compound. Compounds that have extensive interactions will have longer retention times than those compounds that have less interaction with the stationary phase. 25 Ideally, in an efficient chromatography system, each compound in the sample mixture is resolved from the others, and has a Gaussian-shaped peak in the resulting chromatogram. However, due to the column interactions, some band broadening occurs, resulting in less efficient separations. The efficiency of a chromatography system is defined in the van Deemter equation (Equation 2.3), in which H is the theoretical plate height. 1 More efficient chromatography systems have a lower H value. Therefore, to improve efficiency, the three terms in the van Deemter equation should be minimized. Equation 2.3 The A term in Equation 2.3 describes the ability of the mobile phase to take multiple paths through the stationary phase. However, this term is not applicable in GC since the stationary phase is a liquid, rather than a solid. The B term in the equation is the molecular diffusion term. Diffusion occurs because of the compounds' tendency to move from an area of high concentration to low concentration. The sample enters the column as a narrow band, but as it travels through the column it spreads from its highly concentrated band into areas of the mobile phase that have a low concentration of the sample. B accounts for this phenomenon, and can be improved by increasing the flow rate of the mobile phase, u, allowing less time for the band to spread. However, increasing the flow rate also increases the C term, which accounts for mass transfer of the compounds. As the mobile phase moves across the stationary phase and interactions take place, some compounds prolong interactions with the column while others continue to move, resulting in band broadening. Re-establishing equilibrium between the mobile and stationary phases takes place as the mobile phase flows through the column, but requires time and is, therefore, more efficient with a slower flow rate. However, the C term can also be 26 minimized by decreasing the stationary phase thickness, which allows compounds to diffuse faster from the stationary phase back into the mobile phase. While the above factors can lead to broadening of peaks, there are other factors that can 1 also distort the Gaussian shape of peaks. Figure 2.3 displays the ideal symmetrical peak shape, as well as a fronting peak and a tailing peak. Isotherms, which display the concentration of sample in the stationary phase (Cs) versus the concentration in the mobile phase (Cm), can be seen below the corresponding peak shapes. Fronting peaks are caused by overloading the column, so that Cs is greater than Cm. If a sample is too concentrated, more interactions take place with the stationary phase so that the stationary phase may become overloaded with a compound. This leaves little compound in the mobile phase while a large concentration of the compound lags behind in the stationary phase. The compound then elutes gradually from the column, but ends abruptly. Tailing peaks, on the other hand, occur when Cs is less than Cm. These are caused by hydrogen bonding at active sites in the stationary phase to polar compounds in the sample. As the column degrades, OH groups are exposed in the stationary phase, which may bind to some of a polar compound, causing it to trail behind the rest. As the compound elutes from the column, the end of the peak is a more gradual decrease in concentration than is ideal since some of the compound sticks to the column longer. 27 Fronting Tailing Abundance Abundance Abundance Gaussian Time Time Cs Cs Cs Time Cm Cm Cm Figure 2.3: Ideal peak shape (left), fronting peak (center), and tailing peak (right) with corresponding isotherms below. 1 The results of GC analysis are displayed in the form of a chromatogram. These graphs display retention time along the x-axis and abundance along the y-axis. Retention times for compounds are reproducible when the same type of column and same instrument parameters (including temperature program, injector port parameters, and flow rate) are used. Therefore, reference standards analyzed under the same conditions, ideally, using the same instrument and column, can be used to identify compounds by comparing retention times. However, multiple compounds can elute at very similar retention times, and it may not be possible to distinguish certain compounds based on retention time alone. Therefore, to definitively determine the unique identity of a compound, mass spectrometers are often used as the detector for the GC system. 28 2.4 Mass Spectrometry Mass spectrometry (MS) is a type of detector commonly used with gas chromatography, especially in forensic laboratories for the identification of controlled substances. Figure 2.4 depicts a schematic representation of a mass spectrometer. As compounds elute from the GC column, which is at atmospheric pressure, they enter the ion source of the mass spectrometer, which is under vacuum to prevent ions from being deflected by collisions with gaseous 1 molecules. The MS is typically pumped down in stages: first, a rough pump reduces the -2 pressure to 10 -10 -4 torr, then a turbomolecular or diffusion pump is used to reduce it even farther, the approximately 10 -5 torr. Sample molecules enter the ion source from the GC. Here, the molecules are ionized and the ions are then transferred to the analyzer, where they are separated by their mass to charge ratio (m/z). Ions within a specified range of m/z then reach the detector, and a mass spectrum is generated based on the m/z present for each compound. 29 ion source mass analyzer ion detector vacuum GC data acquisition Figure 2.4: Components of a mass spectrometry (MS) system. In the ion source of the MS, compounds are ionized and fragmented. A common method of ionization, and the method used in this research, is electron ionization (EI). The EI source contains a filament which emits electrons with, typically, 70 eV energy, that travel in spiral pathways toward a collection plate. Electrons of 70 eV are commonly used in EI since this is in excess of the energy needed to break most organic bonds (typically 4-20 eV required). As the sample molecules enter the ion source, they interact with these electrons. The sample molecules absorb some of the energy, which causes them to lose an electron, thus becoming ionized. This process forms a molecular ion, designated M+•. Since 70 eV electrons provide excess energy, following ionization, there is sufficient excess energy to cause fragmentation of the molecular ion. A positive repeller plate then pushes the ions toward focus plates, which apply a high voltage (~1,000 – 10,000 V) to focus the ions into a narrow beam and accelerate them into the analyzer. 30 While the molecular ion is important in determining the molecular mass of a compound, fragments are important in determining structural information, which can be useful in determining the identity of the original compound. Due to the excessive fragmentation that typically occurs in EI, this ionization mechanism is termed a ‘hard’ ionization method and is commonly used where structural information is necessary. There are several types of mass analyzers available, but perhaps the most commonly used in GC-MS instruments is the quadrupole mass analyzer (Figure 2.5), which was also used in this 1 research. The quadrupole consists of four metal rods, arranged as two opposite pairs, to which an oscillating radio frequency (RF) and a direct current (DC) voltage are applied. Ions are accelerated in to one end of the quadrupole, where their path is altered by the ratio of the RF and DC voltages. For a given RF/DC ratio, only ions within a narrow m/z range have a stable trajectory and reach the detector at the opposite end. All other ions are neutralized by colliding with the rods, and then pumped away by the vacuum system. The RF and DC can be altered, while keeping the ratio constant, to allow ions of different m/z to reach the detector. During analysis, the RF/DC is scanned for a specific range of m/z (e.g. 50-500 m/z) depending on the molecular weights of the compounds of interest. 31 Figure 2.5: Quadrupole mass analyzer. Ions of stable trajectory pass through the quadrupole to the detector, which, in this research, was an electron multiplier. Figure 2.6 depicts an electron multiplier, which contains a series of dynodes, arranged in a horn-like funnel, that release electrons when struck by ions. The electrons released by the first impact then hit the opposite side of the funnel, where more electrons are released and hit the opposite side and so on. The original signal is multiplied by ~105 when it reaches the anode where the current is detected. secondary electrons signal out ion path Figure 2.6: Electron multiplier. 32 The results of analysis by MS are in the form of a mass spectrum, which displays m/z along the x-axis and abundance, as determined by the amount of current detected by the electron 1 multiplier, along the y-axis. For ions produced by EI, the charge (z) is typically equal to 1, therefore, in the resulting mass spectrum, the x-axis is equivalent to the ion mass. Mass spectra are used to identify compounds based on the molecular ion as well as fragments, from which structural elucidation is possible. Fragmentation of a compound occurs in a reproducible manner under the same ionization and analysis conditions. Therefore, the original structure of a compound can be determined based on the mass of the molecular ion, as well as the mass of the fragments formed. This also allows for definitive identification of compounds, since each mass spectrum is unique to a specific compound. In controlled substance identification, the mass spectrum of a questioned sample can be compared to the spectrum of a known reference standard collected using the same instrument parameters or to a database library containing reference spectra. The spectra are compared based on the presence and absence of each m/z, as well as the ratios of the ions that are present. When MS is used as the detector for GC, a full mass spectrum is collected for every separated compound in the sample mixture. Therefore, a questioned sample and reference standard can be compared based on both retention time and the mass spectra, allowing definitive identification of the substance in the questioned sample. 2.5 Data Pretreatment Due to variations in the GC-MS instrument during analysis, non-chemical sources of variance can be introduced into the chromatograms of the samples. Examples of such variations include flow rate fluctuation or column degradation leading to retention time drift, dissimilarities 33 in sample injection size imparting differences in abundance between replicates, and deterioration of the GC column at high temperatures causing high background signal. Prior to data analysis, such variance should be removed since this otherwise could be identified as differences among samples. Numerous data pretreatment procedures are available and are selected according to the source of variation to be minimized or eliminated. In this research, the major sources of variance were retention time drift and differences in the volume of sample injected for analysis by GC-MS. Retention time drift is a result of minor column deterioration over time as well as slight 4 differences in operating temperatures and carrier gas flow rates. Differences in the volume of sample injected results in variability in abundance of the same compound among replicates of a 5 given sample. In this research, an alignment algorithm was used to minimize the effects of retention time drift, while normalization was used to account for differences in injection volume. 2.5.1 Retention Time Alignment A correlation optimized warping (COW) algorithm was used in this research to retention time align chromatograms. Using the COW alignment, chromatograms are aligned to a target chromatogram. While there are numerous methods to select or generate a target chromatogram, an average chromatogram was used in this research. To generate the average target chromatogram, the abundances of all chromatograms in the dataset are averaged at each retention time. The average abundances are then plotted to create a chromatogram representative of all samples. The first step in the COW alignment is to split the target and sample chromatograms into 4 segments. The segment size refers to the number of datapoints within each segment. This is a 34 user-defined parameter and is selected based on the number of datapoints within the entire chromatogram and the approximate number of datapoints across each peak. The segment size should be greater than the number of points across any single peak. Each chromatogram is then individually aligned to the target chromatogram by shifting, 4 or warping, a certain number of datapoints within each segment. The warp is also a user-defined parameter and typical warp sizes are 1-10, depending on the size of the peaks. The warp refers to the number of datapoints that can be added or subtracted for each peak. The COW algorithm begins at the end of the chromatogram and aligns one segment at a 6 time. For each segment, correlation coefficients are calculated for each warp (i.e. for a warp of 10, any number of datapoints up to 10 may be added or subtracted at each segment) and the warp that results in the highest coefficient is considered the optimal alignment of that segment. However, coefficients for all warps are stored. Then, the algorithm moves to the next segment and repeats. Once each segment has been aligned, a global correlation coefficient is calculated to determine the optimal alignment of the chromatogram to the target. This entire process is repeated for all chromatograms in the dataset. 6 The algorithm determines the alignment based on peak shape, not height or area. This means that the apex of the peak could be aligned to the leading or tailing edge of a peak in the sample yet still produce a high correlation coefficient. Therefore, visual assessment of aligned chromatograms is recommended, and the optimal parameters are somewhat determined on a trial and error basis. This type of alignment is beneficial for datasets containing samples with varying abundances of the same compounds since the peaks are aligned even if the heights differ. For this reason, normalization is often a necessary step after alignment to account for differences in 35 abundance of the same compound, especially among sample replicates. 2.5.2 Normalization There are several methods of normalization reported in the literature, but the methods used in this research included total area and maximum peak normalization, as well as investigations of logarithmic, square root, and fourth root normalizations. This data pretreatment procedure is used to account for differences in abundance which are due to differences in injection volume, as well as differences in detector response. Normalization works by dividing each of the chromatograms in a dataset by a constant, such as the maximum abundance or the total area of a chromatogram, therefore putting them all 5 on the same scale. In maximum peak normalization, for example, the highest abundance in each chromatogram is determined and the average maximum across the dataset calculated. Then, the abundance at each retention time is divided by the maximum abundance of that chromatogram, and multiplied by the average maximum to bring it back up to same order of magnitude as the original data. Since, experimentally, these values are not constant across all chromatograms (i.e. maximum abundance not identical for each sample), some procedures may be more efficient at normalizing a specific dataset than others. 2.6 Principal Components Analysis Principal components analysis (PCA) is a statistical procedure used to evaluate multidimensional data, such as chromatograms which contain thousands of datapoints (i.e. variables). Variance in the dataset is accounted for by principal components (PCs), the number of 7 which is equal to or less than the number of variables being investigated. The magnitude of 36 each PC describes the percentage of variance that is accounted for by that particular PC, and is referred to as the eigenvalue. Typically, nearly all of the variance in a dataset can be accounted for using the first three to four PCs. PCA uses mean-centered data to determine which variables (i.e. compounds) contribute 8 most to the variance, as well as how these variables interact with each other. Mean-centering the data is a form of scaling, which removes any influence from the magnitude of the data, and ensures that first principal component accounts for the most variance. For chromatographic data, the mean-centered data are calculated by averaging the abundance at each retention time in the dataset. The average abundance is then subtracted from the abundance at the corresponding retention times in each individual sample. Mean-centered data are then used to calculate the covariance matrix for dataset. The covariance of a dataset (Equation 2.4) describes how the dimensions (i.e. datapoints or selected 8 compounds) vary from the mean in similar manners. The covariance can be either positive, meaning that both dimensions increase or decrease simultaneously, or negative, meaning that one dimension increases as the other decreases. The equation for covariance is shown below, with x and y representing variables in these dimensions and n representing the total number of dimensions in the dataset. 7,8 Equation 2.4 The covariance is calculated for all variables in the dataset and is displayed in the covariance matrix , which is n × n in dimension. For example, for a dataset of three dimensions (x, y, and z), the covariance matrix would be 3 × 3, as seen below: 37 8 The diagonal of the matrix represents the covariance of each dimension with itself, or the variance, and the matrix is symmetrical around the diagonal; that is, the covariance between x and z is equivalent to the covariance between z and x. From the covariance matrix, eigenvectors and eigenvalues are then calculated. 7 Eigenvectors are vectors which can be multiplied by the covariance matrix to yield a vector that is a multiple of the original. Eigenvalues are the values by which the original vectors were multiplied. A dataset of n dimensions has n eigenvectors, all orthogonal to one another, and each eigenvector has a corresponding eigenvalue. The first eigenvector, also known as the first principal component (PC1), accounts for the most variance, and the amount of variance accounted for decreases with each succeeding PC. PCA can be used to provide a visual representation of the data in a graph of two PCs, referred to as a scores plot. 7,8 Any PCs can be plotted on a scores plot, but since the first two PCs account for the most variance, a scores plot typically displays PC1 and PC2 on the x-axis and y-axis, respectively. To generate the score on PC1 for a given sample, the loadings at each retention time are first calculated by multiplying the mean-centered data by the eigenvector for PC1. The score for the sample (i.e. one chromatogram) on PC1 is the sum of these loadings at all datapoints (i.e. retention times). Scores for the sample on other PCs are calculated in a similar 7 manner, using the appropriate eigenvector. On the scores plot, chemically similar samples cluster closely and chemically dissimilar samples are spread across the plot, thus allowing a visual distinction of the samples. 38 Loadings plots can also be generated, and used to explain positioning of the samples on the scores plot. 7,8 In this research, for example, the loadings at each retention time for PC1 and PC2 were individually plotted against retention time. Within the loadings plots some compounds load negatively due to mean-centering the data. Samples containing a high abundance of a compound that is weighted positively in the loadings plot for PC1, for example, will therefore be positioned positively on PC1 in the scores plot. Additionally, since the loading plots can be plotted against retention time, the resulting plots resemble chromatograms. As a result, the compounds in the loadings plots, which contribute most to the variance being described, can be identified based on retention time. 39 REFERENCES 40 REFERENCES 1. Harris D. Quantitative Chemical Analysis, Seventh Edition. New York, NY: W. H. Freeman and Company, 2007. 2. van Deursen M, Poortman-van der Meer A. Organic impurity profiling of 3,4methylenedioxymethamphetamine (MDMA) tablets seized in the Netherlands. Sci Justice 2006;46:135-152. 3. Pawliszyn J. Solid Phase Microextraction Theory and Practice. New York, NY: WileyVCH, Inc., 1997. 4. Tomasi G, van den Berg F, Andersson C. Correlation Optimized Warping and Dynamic Time Warping as Preprocessing Methods for Chromatographic Data. J Chemometrics 2004;18:231-241. 5. Beebe K, Pell R, Seasholtz M. Chemometrics: A Practical Guide. New York, NY: John Wiley & Sons, Inc., 1998. 6. Nielsen N, Carstensen J, Smedsgaard, J. Aligning of Single and Multiple Wavelength Chromatographic Profiles for Chemometric Data Analysis using Correlation Optimised Warping. J Chromatogr A 1998;805:17-35. 7. Brereton R. Applied Chemometrics for Scientists. West Sussex, England: John Wiley & Sons Ltd., 2007. 8. Smith L. A Tutorial on Principal Components Analysis. 2002. Available at: http://www.sccg.sk/~haladova/principal_components.pdf. (Accessed June, 2012). 41 Chapter 3 Materials & Methods 3.1 MDMA Exhibits Five MDMA exhibits were received from various Michigan State Police laboratories for this study. In this research, "exhibit" refers to batches of tablets which were seized together and thus assigned an identifier by the police. Figure 1 shows photographs of one representative tablet from each of the five exhibits, and the physical characteristics of each exhibit are outlined in Table 1, listed by their police-assigned identifiers. Seven tablets from each exhibit were homogenized with a mortar and pestle, and samples were taken from these homogenized batches for subsequent extraction and analysis. (a) (b) (d) (c) (e) Figure 3.1: MDMA exhibits used in this study (a) T-17 (b) T-27 (c) T-29 (d) T-30 (e) MSU 900-01. 42 Table 3.1: Physical characteristics of MDMA exhibits used in this study. Exhibit Shape Edging Color Imprint T-17 T-27 T-29 T-30 MSU 900-01 round round round round round beveled beveled flat beveled beveled blue pink yellow green green/purple Omega heart waving man "ME" alligator Average Mass (n=7) 242.3 mg 267.9 mg 261.6 mg 262.7 mg 267.4 mg 3.2 Liquid-Liquid Extraction (LLE) Procedure The LLE procedure was adapted from the parameters previously optimized by van 1 Deursen et al. A phosphate buffer with pH 7.00 was prepared by combining 6.83 g of KH2PO4 (Mallinckrodt, Paris, KY; lot # 7100 KXSS) and 27.45 g of Na2HPO4·7H2O (Jade Scientific, Canton, MI; lot # J0407) in a 500 mL volumetric flask and filling to the mark with high purity water (Burdick & Jackson Laboratories, Inc., Muskegon, MI; lot # AK865). For each exhibit, 200 mg of homogenized sample were dissolved in 4.0 mL of buffer, vortexed for ten seconds, sonicated for ten minutes, and centrifuged for eight minutes. Then, 400 µL of toluene (Fluka, St. Louis, MO; lot # SSZBA229A), which contained 0.02 mg/mL eicosane (Aldrich, St. Louis, MO; lot # 09801LD) as an internal standard, were added to the buffer mixture. The sample was then vortexed for two minutes, inverted ten times by hand, and centrifuged for ten minutes. Finally, the toluene layer was removed, taking care not to remove any of the aqueous layer, and placed into a vial for subsequent analysis. 43 3.3 Headspace Solid Phase Microextraction (HS-SPME) Procedure The HS-SPME procedure followed the parameters previously optimized by Bonadio 2 et al. A 65 µm polydimethylsiloxane/divinylbenzene (PDMS/DVB) 23 gauge fiber (Supelco, St. Louis, MO) was conditioned in the GC inlet for 30 minutes at 250 °C, as recommended by the manufacturer. A 4 mL vial with a screw cap and silicone septum (Supelco) containing 40 mg of homogenized exhibit were heated in a water bath for 15 minutes at 80 °C. The silicone septum was then removed and the fiber was inserted into the vial for an additional 15 minutes at 80 °C to 2 extract the volatile compounds. The fiber was then retracted and removed from the vial for subsequent analysis. 3.4 Gas Chromatography-Mass Spectrometry (GC-MS) Analysis For liquid-liquid extracts, an Agilent 6890N with 5973 inert MSD (Agilent Technologies, Inc., Santa Clara, CA) was used. The carrier gas was ultrahigh purity helium with a nominal flow rate of 1 mL/min. A 4mm id tapered inlet liner with glass wool (Restek, Bellefonte, PA), an 11 mm septum (Agilent Technologies, Inc.), and a crossbond 5% diphenyl/95% dimethyl polysiloxane column (Rxi-5ms, 30 m × 0.25 mm × 0.25 µm) (Restek) were used for this study. The injection port was kept constant at 250 °C in splitless mode, and the transfer line was equivalent to the final oven temperature. A 1 µL aliquot of each extract was manually injected using a 1 µL syringe (Hamilton, Reno, NV). Four oven temperature programs, found in the literature, were investigated in this research and are detailed in Table 2. The ion source was kept at 230 °C for each temperature program, and the mass spectrometer was operated in electron ionization mode (70 eV) at 3.58 scans/sec, with a mass scan range of 50-500 m/z. Each extract was analyzed in replicate (n=5) by each of the four temperature programs. 44 Table 3.2 GC temperature programs investigated in this research. 2 Initial Hold Time (min) Ramp Rate (°C/min) Final Temp. (°C) Final Hold Time (min) Total Time (min) 60 5 3 10 150 2 36 170 0 10 1 A Solvent Delay (min) 2 Temperature Program Initial Temp. (°C) 260 3 4 D 3 1 8 300 10 37.25 50 9 1 10 150 5.5 39.5 280 10 5 150 12 15 3 C 90 10 B 300 10 50 5 1 53 1. Please see reference 2. 2. Please see reference 1. 3. Please see reference 3. 4. Please see reference 4. Approximately 4-6 weeks after the initial extracts were analyzed, exhibits T-17, T-30, and MSU 900-01 were extracted again using the same LLE procedure as described above. Two replicates of each extract were analyzed again by each of the four temperature programs, using the same instrument and column described above. This was done to see if replicates of the extracts analyzed at the later time could be associated to replicates of the corresponding extracts analyzed earlier, or if factors such as retention time drift or irreproducibility of the extraction procedure would prevent association. The time lapse between samples is better representative of real-world cases, where different tablets from the same original exhibit could be submitted and analyzed months apart. 45 For HS-SPME, an Agilent 6890N with 5975B inert XL MS (Agilent Technologies, Inc.) was used. The same GC-MS parameters as LLE were used with the following exceptions: the injection port was in splitless mode with a purge at 1.0 min, and narrow bore liner (0.75 mm i.d., Sigma-Aldrich, CO) with a Merlin Microseal (Merlin Instrument Company, Half Moon Bay, CA) were used in place of the traditional liner and septum. The Merlin Microseal was necessary to seal the injection port as the SPME fiber remained in the injector throughout the analysis, while the narrow bore liner increases gas flow through the injector, thus improving desorption efficiency of analytes from the fiber. The abovementioned column was transferred to this instrument for HS-SPME analysis so that exactly the same column was used for all analyses. For HS-SPME, three separate extracts, using three separate 40 mg aliquots of homogenized sample, were prepared for each exhibit. Due to limited instrument availability, all extracts were analyzed by two temperature programs: A and D (Table 2). Between each extraction, fiber blanks were performed to assure that the fiber was clean. The following changes were made to the temperature program for the blanks: initial and final temperatures equivalent to those of the program being used but with no initial hold, ramp of 40 °C/min, 3 min final hold, 50:1 split injection mode and 1 min solvent delay. Several blanks were often needed to completely desorb the MDMA and caffeine from the fiber. Because a newer model mass spectrometer was used for analysis of the HS-SPME extracts compared to the LLE extracts, it was necessary to compare sensitivity between the two detectors. To do this, one exhibit (T-27) was again extracted using the LLE procedure and analyzed in triplicate by temperature program A on the newer mass spectrometer. The data collected was then compared to the data collected for the same exhibit analyzed on the older 46 spectrometer. No differences in sensitivity were observed in the resulting data and hence, the chromatograms from HS-SPME and LLE were deemed comparable. 3.5 Data Analysis 3.5.1 Total Ion Chromatograms (TICs) For LLE, total ion chromatograms (TICs) were generated in the instrument software (MSD ChemStation E.01.01.335; Agilent Technologies, Inc., Palo Alto, CA) and provisional identifications of compounds in each exhibit were made based on mass spectral searches of the NIST database (Version 2.0, National Institute of Standards and Technology, Gaithersburg, MD). The TICs were truncated just before the peak corresponding to n-hexadecanoic acid. This was necessary to remove the fatty acid lubricant peaks which were not compatible with the stationary phase used and hence, exhibited poor chromatography. The TICs were then arranged into four separate datasets, corresponding to each of the four GC temperature programs. That is, for each temperature program, a dataset was generated that contained all replicates of each extract of each exhibit (n=25 for exhibits T-27 and T-29, n=31 for exhibits T-17, T-30 and MSU 900-01). An average target chromatogram, necessary for retention time alignment, was created for each dataset using one TIC, selected at random, from each exhibit. The abundances of the five TICs were averaged at each retention time using Microsoft Excel (Microsoft Office 2007, Redmond, WA), and these average abundances then formed the target chromatogram for that dataset. The datasets were retention time aligned to their respective target chromatogram using a correlation optimized warping (COW) algorithm (LineUp GUI, Version 3.01, Infometrix, Inc., Bothell, WA). All combinations of the following warp and segment sizes were investigated to 47 determine the optimal alignment for each dataset: warp = 2, 4, 6, and 8; segment = 25, 75, and 150. The aligned TICs were initially evaluated by visual examination of the overlaid TICs. The combinations of warp and segment size that seemed to visually improve datasets compared to the unaligned TICs were also assessed using principal components analysis (PCA) (Matlab, R2010b, The Mathworks, Natick, MA). The resulting eigenvalues, eigenvectors and scores were then plotted in Microsoft Excel to generate scores and loadings plots. The alignment parameters were evaluated based on association of all replicates from the same exhibit in the scores plots, with discrimination of different exhibits. The loadings plots, which were used to identify the compounds contributing most to the variance, were also assessed to identify any misalignments in the data, which were evidenced by derivative-shaped peaks. The optimal alignment parameters were chosen based on closest clustering of replicates and greatest separation of exhibits in the scores plots, as well as minimal misalignments observed in the loadings plots. Each aligned dataset was then normalized to account for changes in abundance that may have resulted from differences in injection volume. Normalization was performed for replicates of each exhibit so that the differences in abundance among exhibits would not be compromised. Two normalization procedures were investigated using Microsoft Excel: maximum peak normalization and total area normalization. For maximum peak normalization, the average abundance of the maximum peak in replicates of an exhibit was calculated. The abundance at each retention time for a given replicate was divided by the maximum abundance in its respective TIC, then multiplied by the average maximum abundance of the corresponding exhibit. Total area normalization was performed in a similar manner, but using the sum of the abundances at all retention times instead of the maximum abundance. The normalization 48 procedures were assessed for each dataset using PCA as described above, and the optimal normalization was chosen based on close clustering of replicates and greatest distinction of exhibits. The optimally aligned and normalized datasets were then compared based on visual assessment of PCA scores plots. This was done to determine the optimal temperature program for associating replicates of an exhibit while allowing greatest discrimination of exhibits. 3.5.2 Selected Compounds Despite retention time alignment and normalization, some peaks in each TIC, particularly peaks that varied greatly in abundance among exhibits, were misaligned. To overcome this problem, data analysis was also performed using selected compounds rather than the full TICs. Compounds present in each of the five exhibits and at a threshold above the baseline were chosen and peak areas were integrated using ChemStation software (Agilent Technologies). Principal components analysis was then performed based on integrated peak areas for the selected compounds in the TICs generated from the LLE procedure. Total ion chromatograms generated from HS-SPME were only analyzed based on selected compounds since the poor peak shape of many compounds within the TICs resulted in major misalignments. The resulting datasets were composed of peak areas of the selected compounds for each temperature program and extraction procedure combination, giving a total of six datasets. These datasets were normalized by total area within each exhibit as described for the TICs. The datasets were then subjected to PCA so that the ability of each temperature program to associate replicates and discriminate exhibits based on selected compounds could be evaluated. The 49 extraction procedures were then compared using scores plots of the same temperature program to determine which procedure allowed better association and discrimination of exhibits. 50 REFERENCES 51 REFERENCES 1. Van Deursen M, Poortman-van der Meer A. Organic impurity profiling of 3,4methylenedioxymethamphetamine (MDMA) tablets seized in the Netherlands. Sci Justice 2006;46:135-152. 2. Bonadio F, Margot P, Delémont O, Esseiva P. Optimization of HS-SPME/GC-MS analysis and its use in the profiling of illicit ecstasy tablets (Part 1). Forensic Sci Int 2009;187:73-80. 3. Swist M, Wilamowski J, Zuba D, Kochana J, Parczewski A. Determination of synthesis route of 1-(3-4-methylenedioxyphenyl)-2-propanone (MDP2P) based on impurity profiles of MDMA. Forensic Sci Int 2005;149:181-192. 4. Gimeno P, Besacier F, Chaudron-Thozet H, Girard J, Lamotte A. A contribution to the chemical profiling of 3,4-methylenedioxymethamphetamine (MDMA) tablets. Forensic Sci Int 2002;127:1-44. 52 Chapter 4 Effect of Gas Chromatography Temperature Program on the Association and Discrimination of MDMA Exhibits Extracted using a Liquid-Liquid Extraction Procedure Five MDMA exhibits were extracted using a liquid-liquid extraction (LLE) procedure and analyzed by gas chromatography-mass spectrometry (GC-MS). For GC-MS analysis, four different temperature programs were used. This chapter compares each temperature program based on chromatographic separation and the compounds identified. Total ion chromatograms (TICs) were subjected to principal components analysis (PCA) to investigate the ability of each program to associate replicates of the same MDMA exhibit and discriminate different exhibits. While there were no major differences between temperature programs in chromatographic efficiency or compounds detected, the association and discrimination of the exhibits observed with PCA was noticeably different. The differences observed are explained in detail in the subsequent sections. Following PCA of TICs, selected compounds in each exhibit were also used for PCA of each temperature program, and the improvement in association and discrimination is also discussed. 4.1 MDMA Exhibits Extracted by LLE and Analyzed using Temperature Program A An exemplar TIC of each exhibit analyzed using Temperature Program A is shown in Figure 4.1. This temperature program resulted in baseline separation and narrow Gaussian shaped peaks for most compounds, with the exception of caffeine which will be discussed later. The efficient chromatography was most likely a result of the three temperature ramps included in this program. Slow ramps allowed for better separation of compounds when necessary, and faster ramps minimized band broadening when co-elution was not a concern. This temperature program also had the shortest total analysis time, which is more practical for implementation in forensic laboratories. 53 MDEA/MDDMA 1.00E+06 caffeine methylenedioxyphenyl-2-propanol (MDP2P-OH) MDMA methamphetamine N-formyl methamphetamine benzaldehyde isothiocyanic acid lidocaine 0.00E+00 5 10 15 20 Retention Time (min) Figure 4.1: (a) TIC of Exhibit T-17 after LLE and GC-MS analysis using Temperature Program A. 54 25 N-formyl MDMA Figure 4.1 Cont'd ketamine 6.00E+06 MDMA caffeine methylenedioxyphenyl-2-propanol (MDP2P-OH) methamphetamine isothiocyanic acid benzaldehyde N-formyl MDMA N-formyl methamphetamine MDMA MDEA/ MDDMA diphenhydramine acetaminophen 0.00E+00 5 10 15 20 Retention Time (min) Figure 4.1: (b) TIC of Exhibit T-27 after LLE and GC-MS analysis using Temperature Program A. 55 25 Figure 4.1 Cont'd MDMA 4.00E+06 methylenedioxyphenyl-2-propanol (MDP2P-OH) N-formyl methamphetamine isothiocyanic acid methamphetamine benzaldehyde MDEA/ MDDMA N-formyl MDMA diphenhydramine 0.00E+00 5 10 15 20 Retention Time (min) Figure 4.1: (c) TIC of Exhibit T-29 after LLE and GC-MS analysis using Temperature Program A. 56 25 Figure 4.1 Cont'd caffeine 2.00E+06 N-formyl MDMA methylenedioxyphenyl-2-propanol (MDP2P-OH) MDMA N-formyl methamphetamine lidocaine MDEA/ MDDMA methamphetamine isothiocyanic acid benzaldehyde diphenhydramine 0.00E+00 5 10 15 20 Retention Time (min) Figure 4.1: (d) TIC of Exhibit T-30 after LLE and GC-MS analysis using Temperature Program A. 57 25 Figure 4.1 Cont'd caffeine 1.50E+06 MDMA methylenedioxyphenyl-2-propanol (MDP2P-OH) N-formyl methamphetamine methamphetamine isothiocyanic acid benzaldehyde MDEA/ MDDMA lidocaine 0.00E+00 5 10 15 20 Retention Time (min) Figure 4.1: (e) TIC of Exhibit MSU 900-01 after LLE and GC-MS analysis using Temperature Program A. 58 25 N-formyl MDMA Each of the MDMA exhibits was composed of the same major compounds, such as methamphetamine, MDP2P-OH, and MDMA, in varying abundances. Provisional identifications of the major compounds based on mass spectral searches of the NIST library are labeled on the TICs of each exhibit in Figure 4.1. Exhibits T-17, T-27, T-30 and MSU 900-01 each contained caffeine as a prominent cutting agent, while Exhibit T-29 contained no caffeine. Exhibit T-27 predominantly consisted of ketamine, but also contained acetaminophen as an adulterant. Diphenhydramine was identified in Exhibits T-27, T-29, and T-30; lidocaine in Exhibits T-17, T-30, and MSU 900-01; and procaine in Exhibits T-30 and MSU 900-01. It should be noted that procaine is not shown in Figure 4.1 since the TICs were truncated prior to the irreproducible fatty acid peaks, some of which eluted earlier than procaine. Each exhibit also contained a compound that eluted at 18.6 minutes in this temperature 6 program. In Exhibit T-17, this compound was at high abundance (~ 4.6 × 10 ), meaning it was likely an adulterant added to the tablet. In the four remaining exhibits, this compound was 3 present at low abundance (~ 4.0 × 10 ) and was more likely an impurity or contamination that was introduced during one of the manufacturing steps. This compound was identified as either 3,4-methylenedioxy-N-ethylamphetamine (MDEA) or methylenedioxydimethylamphetamine (MDDMA) based on mass spectral library searches. Figure 4.2 shows the chemical structures of 1 these two compounds, which are both amphetamines with a molecular weight of 207 amu. The two compounds differ only by the substitution on the nitrogen atom: in MDEA, the nitrogen contains an ethyl group whereas in MDDMA, the nitrogen contains two methyl groups. The mass spectra of amphetamines often look very similar and, in cases such as this, the identity of the compound can only be confirmed by comparing retention times of the unknown to 59 appropriate reference standards. However, definitive identification of compounds was not the goal of this research; therefore, this compound will be referred to as "MDEA/MDDMA" throughout the remainder of this discussion. (a) (b) 1 Figure 4.2: Chemical structures of (a) MDEA and (b) MDDMA Several exhibits also contained impurities that resulted from the synthesis of MDMA. Piperonal, which was present in all five exhibits, could have been used to directly synthesize 2 MDMA, or could be a byproduct of MDP2P synthesis. Exhibits T-17 and T-29 both contained 3,4-methylenedioxy-N,N-dimethylbenzylamine (MD-DMB), which indicated reductive amination of piperonal by dimethylamine as the probable synthesis route for MDMA. 2 Reductive amination of piperonal was also indicated in Exhibits T-30 and MSU 900-01, but 60 likely via the nitropropene route based on the presence of (3,4-methylenedioxyphenyl)-2propanone-2-oxime. This impurity results from the reduction of nitroisosafrole, which was the likely synthesis method for the MDP2P precursor of MDMA in these two exhibits. 2 Based on the compounds present in each of the exhibits, and the respective abundances, it is unlikely that any of these five exhibits were produced in the same clandestine laboratory. Although the MDMA in Exhibits T-17 and T-29 appeared to have been produced by a similar method, differences in adulterants present in these exhibits indicated that the tablets were likely produced in different laboratories. Similarly, while the MDP2P precursor in Exhibits T-30 and MSU 900-01 was possibly produced by the same method, the differences in abundances of compounds such as MDP2P-OH and N-formyl MDMA indicated differences in the steps taken to produce the MDMA. 4.2 MDMA Exhibits Extracted by LLE and Analyzed using All Temperature Programs Each of the compounds identified in the five exhibits was observed in all replicates using all four temperature programs, with the exception of tetradecanoic acid, which is one of many fatty acids and may have been used in the tabletting process. This compound, which was only identified in Exhibits T-17 and T-30, was not observed initially, but was visible in replicates of these exhibits analyzed at a later date. This is because tetradecanoic acid is a very polar compound, and likely created active sites on the column and formed hydrogen bonds, adhering to the column instead of eluting. Once there was sufficient build-up of this compound on the column, the additional tetradecanoic acid in the exhibits passed through the column and was visible in the chromatograms of the replicates analyzed later. 61 Figure 4.3 shows Exhibit T-17 analyzed by each temperature program. While the same compounds were identified in each of the four temperature programs, the ratios of some of these compounds differed. For example, the ratio of the two main adulterants in this exhibit, MDEA/MDDMA and caffeine, varied by temperature program. The differences in abundance were a result of differences in temperature ramps, and trends such as this were observed in each of the exhibits. The variability in compound ratios observed between temperature programs makes comparing MDMA profiles challenging: chromatograms of the same exhibit can look very different when analyzed by different temperature programs and links among different exhibits may be overlooked. 4.3 Association and Discrimination of Exhibits Extracted by LLE based on PCA using Total Ion Chromatograms Principal components analysis was conducted separately on the datasets from each temperature program; that is, TICs for all exhibits analyzed using Temperature Program A were treated as one dataset and subjected to PCA and so on for the remaining three programs. Prior to PCA, chromatograms in each dataset were aligned to the corresponding average TIC, and then normalized. From visual assessment of the resulting scores plots, the five exhibits were only fully distinguished using Temperature Program D: for the remaining three programs, some overlap of exhibits was observed. Two temperature programs are discussed in more detail in the following sections: Temperature Program D, which resulted in the greatest discrimination of exhibits, and Temperature Program A, which resulted in the least discrimination of exhibits. 62 Figure 4.3 Cont'd MDEA/MDDMA 1.00E+06 caffeine methylenedioxyphenyl-2-propanol (MDP2P-OH) MDMA methamphetamine N-formyl methamphetamine benzaldehyde isothiocyanic acid lidocaine 0.00E+00 5 10 15 20 Retention Time (min) Figure 4.3: (a) TIC of Exhibit T-17 after LLE and GC-MS analysis using Temperature Program A. 63 25 N-formyl MDMA Figure 4.3 Cont'd MDEA/MDDMA 5.00E+06 caffeine MDMA MDP2P-OH N-formyl MDMA N-formyl methamphetamine methamphetamine isothiocyanic acid tetradecanoic acid 0.00E+00 5 10 15 Retention Time (min) Figure 4.3: (b) TIC of Exhibit T-17 after LLE and GC-MS analysis using Temperature Program B. 64 lidocaine Figure 4.3 Cont'd 1.50E+06 caffeine MDEA/MDDMA MDP2P-OH MDMA methamphetamine N-formyl methamphetamine benzaldehyde phenol isothiocyanic acid lidocaine 0.00E+00 5 10 15 Retention Time (min) Figure 4.3: (c) TIC of Exhibit T-17 after LLE and GC-MS analysis using Temperature Program C. 65 20 N-formyl MDMA Figure 4.3 Cont'd caffeine 4.00E+06 MDEA/MDDMA MDP2P-OH MDMA N-formyl MDMA N-formyl methamphetamine methamphetamine lidocaine isothiocyanic acid tetradecanoic acid 0.00E+00 10 15 20 25 30 Retention Time (min) Figure 4.3: (d) TIC of Exhibit T-17 after LLE and GC-MS analysis using Temperature Program D. 66 35 4.3.1 Association and Discrimination of Exhibits Extracted by LLE and Analyzed using Temperature Program D based on PCA using TICs The scores plot generated for PCA of the five exhibits analyzed using Temperature Program D is shown in Figure 4.4. The first two principal components (PC1 and PC2, respectively) accounted for 81.79% of the variance among the exhibits. Replicates of each exhibit were clustered closely and the five exhibits were clearly distinguished. Exhibits T-30 and MSU 900-01 were positioned positively on PC1 while Exhibits T-17, T-27, and T-29 were all positioned negatively on this PC. Greater discrimination of the exhibits was apparent on PC2, in which Exhibits T-17 and T-29 were positioned negatively, Exhibits T-30 and MSU 900-01 were close to zero, and Exhibit T-27 was positioned positively. The positioning of each exhibit in the scores plot could be explained with reference to the loadings plots for PC1 and PC2 (Figure 4.5). From Figure 4.5 (a), the compounds contributing to the variance described by PC1 were MDP2P-OH, N-formyl MA, caffeine, and N-formyl MDMA, which were all weighted positively, and MDMA, MDEA/MDDMA, and ketamine, which were all weighted negatively. Of these compounds, caffeine and ketamine had the greatest weighting, and, hence, contributed most to the variance. From Figure 4.5 (b), MDMA, caffeine, ketamine, and N-formyl MDMA were all weighted positively on PC2, while methamphetamine, MDP2P-OH, and MDEA/MDDMA were all weighted negatively. On this PC, ketamine had the greatest weighting and, therefore, the greatest contribution to the variance described by PC2. Based on the loadings plots, positioning of all exhibits in the scores plot (Figure 4.4) could be explained by the abundance of the afore-mentioned compounds. Exhibit T-17 contained more methamphetamine and MDEA/MDDMA than any other exhibit, and although the peak with the highest abundance in this exhibit was caffeine, this compound was less abundant in Exhibit T-17 than in the other three exhibits in which it was present. In the scores plot, Exhibit 67 Figure 4.5 Cont'd PC2 (19.17%) 2.00E+07 -2.00E+07 2.50E+07 -1.50E+07 PC1 (62.62%) Figure 4.4: PCA scores plot for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program D. 68 Figure 4.5 Cont'd caffeine 2.00E-01 N-formyl MDMA 1.00E-01 N-formyl methamphetamine MDP2P-OH 0.00E+00 10 15 methamphetamine 20 25 30 35 MDEA/MDDMA MDMA ketamine -1.00E-01 Retention Time (min) Figure 4.5: (a) Loadings plots for PC1 for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program D. 69 Figure 4.5 Cont'd ketamine 2.00E-01 caffeine 1.00E-01 N-formyl MDMA MDMA N-formyl methamphetamine 0.00E+00 10 -1.00E-01 15 methamphetamine 20 25 MDP2P-OH 30 35 MDEA/MDDMA Retention Time (min) Figure 4.5: (b) Loadings plots for PC2 for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program D. 70 T-17 was positioned negatively on both PC1 and PC2. This was due to the high abundances of methamphetamine and MDEA/MDDMA, which were both weighted negatively on both PCs. Since caffeine was weighted positively on both PC1 and PC2 and should have influenced this exhibit to position more positively, the mean-centered data were considered. To mean-center the data, the mean abundance across all exhibits at each retention time was first calculated. Then, in the TIC of each exhibit, the mean abundance was subtracted from the abundance at the corresponding retention time. The mean-centered data for each exhibit were then multiplied by the eigenvector for the respective PC, and the product was summed to generate the score for the exhibit in the scores plot. While there was a large caffeine peak in Exhibit T-17 (Figure 4.6 (a)), the abundance in this exhibit was less than the mean abundance of caffeine across all exhibits. Therefore, when the mean caffeine abundance was subtracted, caffeine displayed a negative contribution in the mean-centered data for Exhibit T-17 (Figure 4.6 (b)). This negative contribution was then multiplied by the respective eigenvector, which was positive at this retention time on both PCs, as seen in the loadings plots (Figure 4.6 (c) and (d)). Therefore, this compound had an overall negative contribution to the scores for the Exhibit T-17 on both PCs (Figure 4.6 (e) and (f)). Exhibit T-27 was positioned negatively on PC1 and most positively on PC2 in the scores plot (Figure 4.4). The TICs from this exhibit were dominated by ketamine, caffeine, MDMA and N-formyl MDMA. On PC1, MDMA and ketamine were weighted negatively, while caffeine and N-formyl MDMA were weighted positively. However, ketamine was present only in this exhibit, and the abundance of MDMA was relatively high compared to other exhibits. As a result, the positioning of Exhibit T-27 in the scores plot was mainly determined by these two compounds. Since both compounds were weighted negatively on PC1 (Figure 4.5 (a)), Exhibit T-27 was 71 Figure 4.6 Cont'd 4.50E+06 caffeine 0.00E+00 35 36 1.00E+06 37 38 Retention Time (min) 35 36 37 -3.00E+06 38 caffeine Figure 4.6: Effect of mean-centering on compound contribution (a) TIC showing caffeine peak in Exhibit T-17 (b) Mean-centered caffeine peak in Exhibit T-17 , 72 Figure 4.6 Cont'd caffeine 2.00E-01 1.00E-01 0.00E+00 35 36 37 38 37 38 -1.00E-01 Retention Time (min) 2.00E-01 caffeine 1.00E-01 0.00E+00 35 36 -1.00E-01 Figure 4.6: Effect of mean-centering on compound contribution (c) Loadings for caffeine on PC1, (d) Loadings for caffeine on PC2, 73 Figure 4.6 Cont'd 1.00E+05 35 36 37 37 -7.00E+05 38 caffeine 1.00E+05 38 Retention Time (min) 35 36 caffeine -4.00E+05 Figure 4.6: Effect of mean-centering on compound contribution (e) Negative loading for caffeine on PC1 in Exhibit T-17, and (f) Negative loading for caffeine on PC2 in Exhibit T-17. 74 positioned negatively on PC1 in the scores plot. On PC2, this exhibit was positioned most positively because the major compounds in this exhibit (MDMA, caffeine, ketamine, and Nformyl MDMA) were all weighted positively in the loadings plot for this PC (Figure 4.5 (b)), thus impacting the positioning of this exhibit more than the negatively-weighted compounds. Exhibit T-29 was composed mainly of MDMA and MDP2P-OH. It was positioned most negatively on PC1 in the scores plot (Figure 4.4) because it had the highest abundance of MDMA of all exhibits. This compound was weighted negatively on PC1 (Figure 4.5 (a)) and the high abundance of MDMA in this exhibit outweighed the small positive contribution of MDP2P-OH. Exhibit T-29 was positioned negatively on PC2 mainly due to the high abundance of MDP2P-OH, which had a high negative weighting on PC2 (Figure 4.5 (b)). The lack of caffeine in this exhibit also contributed to its negative positioning on both PCs. Since no caffeine was present in the exhibit, the mean-centered data for Exhibit T-29 contained negative contributions from this compound. When multiplied by the positive eigenvector for each PC at this retention time, the resulting overall contributions on PC1 and PC2 were negative. Exhibit T-30 had a higher abundance of caffeine and N-formyl MDMA than any other exhibit, and it also contained a high abundance of MDP2P-OH. This exhibit was positioned most positively on PC1 in the scores plot (Figure 4.4), due to the high abundance of caffeine, N-formyl MDMA, and MDP2P-OH, all of which were weighted positively on PC1 (Figure 4.5 (a)). On PC2, Exhibit T-30 was positioned slightly positive, but very close to zero. Although this exhibit contained caffeine and MDMA, which were weighted positively on PC2, it also contained MDP2P-OH, which was weighted negatively. In addition, when the mean-centered data were considered, the below average abundances of MDMA and ketamine led to negative 75 contributions of these positively-weighted compounds. Therefore, the overall positioning of this exhibit in the scores plot was very close to zero. Exhibit MSU 900-01 contained a high abundance of caffeine, nearly as abundant as in Exhibit T-30, and contained a moderate abundance of MDMA compared to the other exhibits. Similar to Exhibit T-30, this exhibit was positioned positively on PC1 in the scores plot and slightly positive, but close to zero, on PC2 (Figure 4.4). On PC1, this exhibit was positive because of the high abundance of caffeine, but it was less positive than Exhibit T-30 because of the higher abundance of MDMA, which was weighted negatively on this PC (Figure 4.5 (a)). To explain the positioning on PC2, the mean-centered data were again considered. Due to the relatively low abundances of MDMA and N-formyl MDMA in MSU 900-01 compared to other exhibits, these two compounds contributed negatively in the mean-centered data. When the negative mean-centered data were multiplied by the respective eigenvectors, which were positive for both compounds, the resulting loadings were negative. These negative contributions offset the positive influence of caffeine on the positioning of this exhibit in the scores plot, therefore positioning it close to zero on PC2. 4.3.2 Association and Discrimination of Exhibits Extracted by LLE and Analyzed using Temperature Program A based on PCA using TICs The scores plot for Temperature Program A, which displayed the most spread among replicates and the most overlap between exhibits, is shown in Figure 4.7. The first two PCs accounted for 68.58% of the total variance among the exhibits. Exhibits T-27 and T-29 were clearly distinguished, while Exhibit MSU 900-01 displayed some overlap with both Exhibits T-17 and T-30. The latter three exhibits also displayed the most spread among replicates, 76 especially on PC2. Exhibit T-27 was the only exhibit to load positively on PC1, along with one replicate of Exhibit T-30, while the remaining exhibits were positioned negatively on this PC, with Exhibits T-29 and T-30 close to zero. On PC2, all replicates of Exhibit T-30 and two replicates of MSU 900-01 loaded positively, while the remaining exhibits were positioned negatively, with Exhibit T-27 closest to zero. The positioning of each exhibit in the scores plot could be explained with reference to the loadings plots for PC1 and PC2 (Figure 4.8). From Figure 4.8 (a), the compounds contributing to the variance described by PC1 were MDMA, caffeine, ketamine, and N-formyl MDMA, which were all weighted positively, and MDEA/MDDMA, and MDP2P-OH, which were both weighted negatively. Of these compounds, ketamine and MDMA had the greatest weighting, contributing most to the variance. On this PC, caffeine contributed both positively and negatively because of retention time misalignments that were observed throughout the TICs and which will be discussed later. From Figure 4.8 (b), Nformyl methamphetamine, caffeine, and N-formyl MDMA were all weighted positively on PC2, while MDP2P-OH, MDMA, and MDEA/MDDMA were all weighted negatively. On this PC, caffeine and MDMA had the greatest weighting and, therefore, the greatest contribution to the variance described by PC2. Exhibit T-27 was the only exhibit that contained ketamine. This compound was weighted positively on PC1 and had the greatest contribution to this PC (Figure 4.8 (a)). As a result, Exhibit T-27 was positioned most positively on PC1 in the scores plot (Figure 4.7). To explain the negative positioning of all other exhibits on PC1, the mean-centered data were again considered. While other compounds also contributed to the negative positioning of the four exhibits on PC1 in the scores plot, the main contribution to the positioning was ketamine. Since the other four exhibits did not contain ketamine, there were large, negative contributions from 77 PC2 (23.33%) 2.00E+07 -2.00E+07 2.50E+07 -1.50E+07 PC1 (45.25%) Figure 4.7: PCA scores plot for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program A. 78 ketamine 2.00E-01 MDMA caffeine 1.00E-01 N-formyl MDMA N-formyl methamphetamine 0.00E+00 5 10 methamphetamine 15 20 25 caffeine MDP2P-OH MDEA/MDDMA -1.00E-01 Retention Time (min) Figure 4.8: (a) Loadings plots for PC1 for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program A. 79 Figure 4.8 Cont'd 2.00E-01 caffeine 1.00E-01 N-formyl MDMA N-formyl methamphetamine 0.00E+00 5 10 15 20 25 ketamine methamphetamine MDP2P-OH MDEA/MDDMA MDMA -1.00E-01 Retention Time (min) Figure 4.8: (b) Loadings plots for PC2 for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program A. 80 this compound in the mean-centered data for each exhibit. Ketamine had a large positive loading on PC1 so that when the mean-centered data were multiplied by the PC1 eigenvector, the result was a negative loading at this retention time. Therefore, Exhibits T-17, T-29, T-30, and MSU 900-01 were all positioned negatively on PC1 in the scores plot. Exhibit T-17, which was composed mainly of MDEA/MDDMA and caffeine, was positioned negatively on PC2. This was due to the high abundance of MDEA/MDDMA, which was weighted negatively on PC2. While there was a high abundance of caffeine in this exhibit, due to misalignments and its average abundance relative to other exhibits, this compound did not influence the positioning of Exhibit T-17 as greatly as MDEA/MDDMA. Exhibit T-27 contained the highest abundances of ketamine, MDMA, and caffeine of all five exhibits. In the scores plot, this exhibit was positioned slightly negative on PC2. On PC2, MDMA was weighted very negatively, and ketamine was also weighted slightly negatively, while caffeine had a strong positive contribution. Overall, the positive contribution of caffeine reduced the negative contributions of MDMA and ketamine, positioning this exhibit slightly negative but close to zero on PC2. Exhibit T-29, which consisted mostly of MDP2P-OH and MDMA, was positioned most negatively on PC2. Both MDP2P-OH and MDMA were weighted negatively on this PC, making this the most negative exhibit on PC2 in the scores plot. Exhibit T-30 contained caffeine, N-formyl MDMA, and MDP2P-OH. It was positioned slightly negatively on PC1, with one replicate loading positively on this PC. Retention time misalignments in the caffeine peak caused both positive and negative contributions in the loadings plot, thus positioning some replicates more positively than others in the scores plot. Since this exhibit was positioned close to zero due to the positive contributions of N-formyl 81 MDMA and caffeine, the misalignment in caffeine caused one replicate to be positioned positively on this PC. While this exhibit was positioned positively on PC2, spread was apparent among replicates. The positive positioning on PC2 was due to the positively loading caffeine and N-formyl MDMA. The spread among replicates could be attributed to the caffeine, which was not as reproducible as other compounds, leading to variability in abundance between replicates. Exhibit MSU 900-01 did not contain a very high abundance of any compounds relative to the other four exhibits: only abundances of MDP2P-OH, caffeine and N-formyl MDMA were slightly above average. This exhibit was positioned slightly negatively on PC2, with two replicates loading positively. This was because of the negative contribution of MDP2P-OH, as well as the misalignment of caffeine, which led to both positive and negative contributions to the positioning of this exhibit, explaining why some replicates were positioned negatively while other replicates were positioned positively. 4.4 Association and Discrimination of Exhibits Extracted by LLE based on PCA using Selected Compounds Retention time misalignments were observed throughout the TICs obtained using Temperature Programs A, B, and C. In each program, the major misalignment occurred for the caffeine peak, as mentioned previously. This compound was observed at significantly different 5 6 abundances (~ 6.0 × 10 - 6.2 × 10 by Temperature Program A) in four of the five exhibits (T-17, T-27, T-30 and MSU 900-01). Chromatographic peak shape for caffeine was very poor, displaying extensive fronting, which resulted from overloading the column with this compound. When compounds in a sample mixture are too concentrated, more interactions take place with the stationary phase, thus increasing the time the compound spends in the system, causing peak broadening. While reducing the amount of sample used in the extraction procedure would reduce 82 the concentration of caffeine, and likely improve chromatography of this specific peak, it would also decrease the abundance of minor compounds to levels below the baseline or limit of detection. In any chromatographic separation, there must be a compromise between good chromatography of high abundance peaks and detection of low abundance peaks. This was taken into account in the optimization studies for this extraction procedure, as reported by van Deursen 3 et al. In this research, no steps were taken to correct the fronting since optimization was not the objective of this research. Rather, this research was intended as a comparative study of previously optimized profiling methods. As a result of the poor chromatography, the caffeine peak could not be retention time aligned satisfactorily (Figure 4.9). Poor alignment impacted PCA, which cannot distinguish nonchemical sources of variance from true, chemical sources of variance. Misaligned compounds often appeared in the loadings plots as two peaks, one weighted positively and the other negatively, adding variance to the data that was not actually a result of differences among the exhibits. This caused more spread among replicates and/or closer association of different exhibits. The misalignments in the caffeine peak observed using Temperature Program A could be used to explain the spread observed among replicates of exhibits in the scores plot (Figure 4.7), as well as the incorrect associations observed between exhibits. For example, in the retentiontime aligned TICs, the caffeine peak in the seven replicates of Exhibit T-30 (green) was not wellaligned (Figure 4.9). In three replicates, the apex of the caffeine peak was aligned with a retention time of 26.36 min, and for two replicates, the apex of the peak was aligned at 26.51 min. The remaining two replicates were not aligned to each other or to the other replicates: 83 2.00E+07 5.00E+06 -2.00E+07 2.50E+07 -1.50E+07 0.00E+00 26.1 26.2 26.3 26.4 26.5 26.6 Retention Time (min) Figure 4.9: Misaligned caffeine peak in replicates of Exhibits T-27 and T-30 after LLE and GC-MS analysis using Temperature Program A. Alignment was performed using a correlation optimized warping algorithm with a warp of 2 and a segment size of 150. Inset: Scores plot obtained for five MDMA exhibits after LLE and GC-MS analysis using Temperature Program A. 84 for one, the apex of the caffeine peak occurred at 26.32 min, and for the other, the apex was at 26.40 min. This same grouping of the Exhibit T-30 replicates was observed in the scores plot (shown as an inset in Figure 4.9), in which three of the replicates were clustered closely, two of the replicates were clustered closely, and the remaining two replicates formed individual clusters. In contrast, the alignment of the caffeine peak in Exhibit T-27 (blue) was acceptable: the apex of the peak in all five replicates was aligned with a retention time of approximately 26.39 min (Figure 4.9). In the scores plot (inset to Figure 4.9), replicates of this exhibit were all closely clustered as one group. In an attempt to overcome problems introduced due to retention time misalignment, the data analysis was repeated, this time using selected compounds rather than the full TIC. When peak areas of compounds were considered, retention time shifts were no longer an issue. The selected compounds used for data analysis, which varied by temperature program, are summarized in Table 4.1. Table 4.1 Selected compounds for data analysis of MDMA exhibits extracted by LLE using Temperature Programs A-D. Temperature Program Compound A B C D benzaldehyde   methamphetamine     isothiocyanic acid     MDP2P-OH     MDMA     N-formyl MA     N-formyl MDMA     Note: Check indicates compound present and used in data analysis. 85 4.4.1 Association and Discrimination of Exhibits Extracted by LLE and Analyzed using Temperature Program A based on PCA using Selected Compounds The PCA scores plot based on selected compounds for the five exhibits analyzed using Temperature Program A is shown in Figure 4.10, in which the first two PCs accounted for 90.85% of the variance. Replicates of Exhibits T-27 and T-29 were closely clustered, while more spread was apparent, especially on PC2, among replicates of each of the other three exhibits. The corresponding loadings plot is shown in Figure 4.11. It should be noted that, since selected compounds were used as the variables rather than the full TIC, the loadings for each PC could not be plotted against retention time as in previous loadings plots. Instead, the weight of each selected compound on both PC1 and PC2 is shown in the form of a scatter plot, with PC1 on the x-axis and PC2 on the y-axis. From the loadings plot, MDMA and MDP2P-OH contributed most to the variance described: MDMA had the highest positive weighting on PC1 while MDP2P-OH had the highest positive weighting on PC2. The remaining compounds were positioned close to zero on both PCs and therefore had less contribution to the positioning of the exhibits in the scores plot. This was particularly true for benzaldehyde and isothiocyanic acid, which were positioned at zero on both PCs in the loadings plot. In the scores plot (Figure 4.10), Exhibit T-17 was positioned negatively on both PC1 and PC2. This exhibit was composed mainly of MDEA/MDDMA and caffeine, neither of which was considered for this analysis. The compounds in this exhibit that were used in this analysis, MDP2P-OH and MDMA, were at relatively low abundances compared to the other exhibits. Therefore, in the mean-centered data, MDMA contributed slightly negatively on PC1 while MDP2P-OH contributed slightly negatively on PC2. When the negative mean-centered data were 86 PC2 (12.19%) 6.00E+07 -1.50E+08 1.50E+08 PC1 (78.66%) -6.00E+07 Figure 4.10: PCA scores plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program A. 87 1.00E+00 benzaldehyde methamphetamine isothiocyanic acid MDP2P-OH PC2 MDMA N-formyl MA N-formyl MDMA -2.00E-01 1.00E+00 PC1 -2.00E-01 Figure 4.11: Loadings plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program A. 88 multiplied by the positive eigenvectors, the results were negative. Accordingly, positioning of this exhibit was slightly negative on both PCs. Spread among replicates was observed on PC2 in this exhibit, as well as Exhibits T-30 and MSU 900-01, due to the variability in abundance of MDP2P-OH and N-formyl MDMA. MDP2P-OH had the greatest influence on positioning on PC2, and variability was observed among replicates, especially in these three exhibits, because this compound eluted at the end of a hold and beginning of a slow temperature ramp in Temperature Program A. This change in temperature programming led to more variability in the instrument, which was visible in the abundances of the peak at this retention time. Exhibit T-27, which contained the highest abundance of MDMA of all five exhibits, was positioned most positively on PC1, and most negatively on PC2, in the scores plot (Figure 4.10). MDMA was weighted very positively on PC1 (Figure 4.11), which positioned this exhibit most positively on PC1. This exhibit also contained the lowest abundance of MDP2P-OH. Therefore, after mean-centering, MDP2P-OH had a negative contribution in this exhibit. When multiplied by the positive eigenvector for MDP2P-OH on PC2, the product was negative at this retention time. This, in combination with the negatively-weighted MDMA, positioned Exhibit T-27 most negatively on PC2 in the scores plot. Exhibit T-29 contained the highest abundance of MDP2P-OH of any of the exhibits, and nearly as much MDMA as Exhibit T-27. In the scores plot, this exhibit was positioned very positively on PC1 and most positively on PC2 (Figure 4.10). The positive positioning on PC1 was due to the large positive weighting of MDMA on this PC, while the positive positioning on PC2 was due to the large positive weighting of MDP2P-OH on this PC (Figure 4.11). Exhibit T-30 contained more N-formyl methamphetamine and N-formyl MDMA than the other four exhibits, but only moderate abundances of each of the other selected compounds. This 89 exhibit was positioned negatively on PC1 in the scores plot (Figure 4.10). This was mainly due to the low abundance of MDMA in the exhibit. After mean-centering, MDMA had a negative contribution, therefore, when multiplied by the eigenvector for PC1, the very positive weighting for MDMA resulted in an overall negative loading. The exhibit was positioned positively on PC2 in the scores plot because of the combination of positive weightings of MDP2P-OH, N-formyl methamphetamine, and N-formyl MDMA on this PC (Figure 4.11). Exhibit MSU 900-01 contained low abundances of each of the selected compounds compared to the other exhibits. Due to the low abundances, each of the compounds contributed negatively after mean-centering. When multiplied by the eigenvector for each PC, overall, there were negative contributions, resulting in the negative positioning of this exhibit on both PCs in the scores plot (Figure 4.10). An exception to this was benzaldehyde, which contributed positively to the mean-centered data. However, benzaldehyde had zero contribution to the eigenvectors for each PC (Figure 4.11) and therefore, did not influence the positioning of the exhibit in the scores plot. 4.4.1.1 Comparison of Association and Discrimination of Exhibits Extracted by LLE and Analyzed using Temperature Program A based on TICs and Selected Compounds While the five exhibits could not be distinguished using PCA of the TICs using Temperature Program A (Figure 4.7), they were all distinguished using PCA of selected compounds (Figure 4.10). This was primarily due to the major misalignments that were observed in the TICs, especially in the caffeine peak. When using selected compounds, retention time alignment were not necessary; therefore, problems associated with misalignments were avoided. Exhibits T-27 and T-29 were clearly separated from the other three exhibits when the TICs were considered (Figure 4.7), mainly due to the presence of ketamine and lack of caffeine, 90 respectively, in these exhibits. Although ketamine and caffeine were not considered in the selected compounds analysis, Exhibits T-27 and T-29 were still clearly separated from the other three exhibits. This was still possible because these two exhibits contained much higher abundance of MDMA than the other three exhibits, and Exhibit T-29 also contained a high abundance of MDP2P-OH. Although there was still spread among replicates of Exhibits T-17, T-30, and MSU 900-01 using selected compounds, these three exhibits were more clearly separated than when TICs were used (Figure 4.10 compared to Figure 4.7). 4.4.1.2 Further Investigation of Normalization Procedures on Exhibits Extracted by LLE and Analyzed using Temperature Program A Based on the poor discrimination of exhibits in the scores plot using Temperature Program A, additional normalization procedures were investigated. These procedures included th square root, 4 root, and logarithmic normalization, since these procedures were also investigated in MDMA research by Bonadio et al. and Weyermann et al. 4-6 The data were normalized using the procedure under investigation, then subjected to PCA, again based only on the selected compounds rather than the full TIC. The remaining normalization procedures were all investigated in a similar manner. The resulting scores plots were visually assessed to determine if there was improvement in association of replicates and discrimination of exhibits using each normalization procedure. No significant improvements over the total area normalization used initially were observed with any of the normalization procedures investigated. Of those procedures investigated, only logarithmic normalization demonstrated successful discrimination of the five exhibits (Figure 4.12); however, the discrimination was not as clear as observed following total 91 PC2 (20.39%) 1.50E+00 -2.00E+00 2.00E+00 PC1 (57.01%) -1.50E+00 Figure 4.12: PCA scores plot based on selected compounds after logarithmic normalization for five MDMA exhibits extracted by LLE and analyzed by GC-MS using Temperature Program A. 92 area normalization (Figure 4.10). Hence, for all subsequent data analysis, selected compounds were used and the data were subjected to total area normalization prior to PCA 4.4.2 Comparison of Association and Discrimination of Exhibits Extracted by LLE and Analyzed using All Temperature Programs based on Selected Compounds Datasets were formed for each temperature program using selected compounds only. After normalization, each dataset was again subjected to PCA. The first two PCs accounted for 90.85% of the total variance using Temperature Program A (Figure 4.10), 90.21% using Temperature Program B (Figure 4.13), 95.29% using Temperature Program C (Figure 4.14), and 85.55% using Temperature Program D (Figure 4.15). Temperature Program D still displayed the greatest separation of exhibits and closest clustering of replicates, even though the first two PCs accounted for the least amount of variance compared to the other temperature programs. This can be explained by considering the instrumental variance observed by each temperature program. For example, using Temperature Program C, which accounted for the most variance on the first two PCs, the relative standard deviations (RSDs) for the selected compounds ranged from 7.47% to 27.82% and averaged 16.57%. These RSDs were much higher than those of Temperature Program D, which ranged from 1.25% to 28.73% and averaged 6.79%. This means much more instrumental variance was observed using Temperature Program C than D. Since the instrumental variance could be accounted on the first two PCs, more spread was visible in the scores plot among replicates using Temperature Program C. More PCs were required to describe the variance using Temperature Program D because the variance was mostly chemical, therefore the differences between samples were much greater. In addition, Temperature Program A still showed the least separation of exhibits and the most spread among replicates. However, there was improvement in the separation of exhibits 93 PC2 (25.27%) 6.00E+07 -1.20E+08 1.20E+08 PC1 (64.94%) -6.00E+07 Figure 4.13: PCA scores plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program B. 94 PC2 (11.52%) 6.00E+07 -1.50E+08 1.50E+08 PC1 (83.77%) -6.00E+07 Figure 4.14: PCA scores plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program C. 95 PC2 (27.19%) 8.00E+07 -1.50E+08 1.50E+08 PC1 (58.36%) -8.00E+07 Figure 4.15: PCA scores plot based on selected compounds for five MDMA exhibits extracted using LLE and analyzed by GC-MS using Temperature Program D. 96 using selected compounds over TICs for all four temperature programs. Besides Temperature Program D, Temperature Program B was the most efficient chromatographically, displaying better reproducibility of compounds than Temperature Programs A or C. This was reflected in the scores plots by closer clustering of replicates using Temperature Programs B and D (Figures 4.10 and 4.13-4.15). In the scores plot for each temperature program, Exhibits T-17 and MSU 900-01 were the closest associated exhibits. This was because these two exhibits were mainly composed of MDP2P-OH and MDMA, with caffeine as an adulterant. While the ratio of MDP2P-OH and MDMA was reversed in these two exhibits, both compounds were observed at similar abundances, and the two exhibits did not contain vastly different abundances of any compounds, especially the selected compounds, that would allow them to be more distinguished from one another. 4.5 Summary of Findings using LLE Using PCA to compare the four GC temperature programs, it was determined that Temperature Program D was most successful in associating replicates of each exhibit and discriminating the five exhibits. In the scores plots for the three other temperature programs, much of the spread in replicates and incorrect association of exhibits using TICs could be attributed to misalignments throughout the chromatograms, especially in the caffeine peak. Separation of exhibits in the scores plots markedly improved for each temperature program when selected compounds were used instead of TICs because misalignments were not an issue. However, Temperature Program D was the only program to fully associate all replicates with complete discrimination of the five MDMA exhibits in both instances. 97 REFERENCES 98 REFERENCES 1. Southern Association of Forensic Scientists. Forendex. March, 2012. Available at: http://forendex.southernforensic.org/index.php/home/index. (Accessed May, 2012) 2. Gimeno P, Besacier F, Chaudron-Thozet H, Girard J, Lamotte A. A contribution to the chemical profiling of 3,4-methylenedioxymethamphetamine (MDMA) tablets. Forensic Sci Int 2002;127:1-44. 3. Van Deursen M, Poortman-van der Meer A. Organic impurity profiling of 3,4methylenedioxymethamphetamine (MDMA) tablets seized in the Netherlands. Sci Justice 2006;46:135-152. 4. Bonadio F, Margot P, Delémont O, Esseiva P. Headspace solid-phase microextraction (HS-SPME) and liquid-liquid extraction (LLE): Comparison of the performance in classification of ecstasy tablets (Part 2). Forensic Sci Int 2008;182:52-56. 5. Bonadio F, Margot P, Delémont O, Esseiva P. Optimization of HS-SPME/GC-MS analysis and its use in the profiling of illicit ecstasy tablets (Part 1). Forensic Sci Int 2009;187:73-80. 6. Weyermann C, Marquis R, Delaporte C, Esseiva P, Lock E, Aalberg L, Bonzenko Jr. J, Dieckmann S, Dujourdy L, Zrcek F. Drug intelligence based on MDMA tablets data I. Organic impurities profiling. Forensic Sci Int 2008;177:11-16. 99 Chapter 5 Effect of Gas Chromatography Temperature Program on the Association and Discrimination of MDMA Exhibits Extracted using a Headspace Solid Phase Microextraction Procedure Five MDMA exhibits were extracted using a headspace solid phase microextraction (HS-SPME) procedure and analyzed by gas chromatography-mass spectrometry (GC-MS). For GC-MS analysis, two different temperature programs were used: Temperature Program A, which previously displayed the least separation of exhibits and greatest spread among replicates when extracted using LLE, and Temperature Program D, which displayed the most separation of exhibits and closest clustering of replicates using LLE. For each temperature program, selected compounds were subjected to principal components analysis (PCA) to investigate association of extracts from the same MDMA exhibit and discrimination among the five exhibits. This chapter firstly compares the association and discrimination achieved among the exhibits after HS-SPME extraction and analysis using each of these temperature programs. Only minor differences in chromatographic efficiency were observed between the two temperature programs, and the five MDMA exhibits were successfully discriminated using both programs. The association and discrimination achieved using HS-SPME compared to LLE was also investigated. When the two extraction procedures were compared, HS-SPME resulted in closer clustering in the scores plots than LLE, especially using Temperature Program A. Note: Throughout this chapter, samples from the same exhibit are referred to as "extracts" since, when using HS-SPME, each sample was from a separate extraction. This is in contrast to when the liquid-liquid extraction (LLE) procedure was used, and each sample from the same exhibit was taken from one extraction, and were therefore referred to as "replicates." 100 5.1 MDMA Exhibits Extracted by HS-SPME and Analyzed using Temperature Program A An exemplar TIC of each exhibit analyzed using Temperature Program A is shown in Figure 5.1. Provisional identifications of the major compounds based on mass spectral searches of the NIST library are labeled on the TICs of each exhibit. All of the major compounds identified using LLE were observed in the HS-SPME extracts, with the exception of benzaldehyde and isothiocyanic acid, which were present in some but not all exhibits using HS-SPME. This indicates that these compounds were present in the LLE solvents as well as in some of the MDMA exhibits. For example, benzaldehyde was only observed in Exhibit T-17 using HS-SPME but was observed in all five exhibits after LLE. This compound can be an 2 1 impurity in MDMA manufacturing (either as a precursor of ephedrine or methamphetamine ), as well as an impurity in toluene. Hence, while benzaldehyde is likely an impurity in Exhibit T-17, the compound’s presence in the remaining LLE extracts is more likely due to an impurity in the extraction solvent. In addition, diethyl phthalate, which is used as a binder in the tabletting 1 process, was observed only using HS-SPME. This is because diethyl phthalate is not soluble in toluene, and was therefore not extracted using the LLE procedure. Several impurities, such as piperonal, 3,4-methylenedioxy-N,N-dimethylbenzylamine (MD-DMB), and (3,4-methylenedioxyphenyl)-2-propanone-2-oxime, were barely visible above the baseline using LLE, but were observed at significant levels using HS-SPME. This is because of the ability to pre-concentrate compounds on the fiber using HS-SPME, resulting in a higher concentration of compound being carried onto the GC column. In addition, two minor impurities from MDMA production were observed only using HS-SPME due to the pre-concentration of the fiber: phenyl-2-propanone (P2P) and methylenedioxyphenyl-2-propanone (MDP2P). P2P is a 1 common precursor of methamphetamine, which was present in all five exhibits, and MDP2P is 101 MDP2P-OH 6.00E+07 methamphetamine MDEA/MDDMA N-formyl methamphetamine caffeine methylenedioxyphenyl2-propanone (MDP2P) diethyl phthalate 3,4-methylenedioxyN,Ndimethylbenzylamine isothiocyanic nicotine acid piperonal phenol phenyl-2propanone MDMA lidocaine 0.00E+00 5 10 15 20 Retention Time (min) Figure 5.1: (a) TIC of Exhibit T-17 after HS-SPME and GC-MS analysis using Temperature Program A. 102 25 N-formyl MDMA Figure 5.1 Cont'd ketamine 5.00E+07 MDMA methamphetamine MDP2P-OH N-formyl methamphetamine methylenedioxyphenyl2-propanone (MDP2P) phenol phenyl-2propanone (P2P) diethyl phthalate caffeine diphenhydramine MDEA/ acetaminophen MDDMA piperonal nicotine 0.00E+00 5 10 15 20 Retention Time (min) Figure 5.1: (b) TIC of Exhibit T-27 after HS-SPME and GC-MS analysis using Temperature Program A. 103 25 N-formyl MDMA Figure 5.1 Cont'd 6.00E+07 MDP2P-OH MDMA N-formyl methamphetamine methamphetamine methylenedioxyphenyl2-propanone (MDP2P) MDEA/ MDDMA 3,4-methylenedioxyN,Ndimethylbenzylamine phenol phenyl-2propanone (P2P) N-formyl MDMA diethyl phthalate nicotine diphenhydramine piperonal 0.00E+00 5 10 15 20 Retention Time (min) Figure 5.1: (c) TIC of Exhibit T-29 after HS-SPME and GC-MS analysis using Temperature Program A. 104 25 Figure 5.1 Cont'd MDP2P-OH 6.50E+07 MDMA N-formyl methamphetamine methamphetamine N-formyl MDMA diethyl phthalate caffeine MDEA/ MDDMA methylenedioxyphenyl2-propanone (MDP2P) phenol phenyl-2propanone (P2P) nicotine piperonal lidocaine diphenhydramine (3,4-methylenedioxyphenyl)2-propanone-2-oxime 0.00E+00 5 10 15 20 Retention Time (min) Figure 5.1: (d) TIC of Exhibit T-30 after HS-SPME and GC-MS analysis using Temperature Program A. 105 25 Figure 5.1 Cont'd MDP2P-OH 6.50E+07 MDMA methamphetamine N-formyl MDMA N-formyl methamphetamine caffeine diethyl phthalate MDEA/ MDDMA phenol phenyl-2propanone (P2P) methylenedioxyphenyl2-propanone (MDP2P) lidocaine (3,4-methylenedioxyphenyl)2-propanone-2-oxime piperonal 0.00E+00 5 10 15 20 25 Retention Time (min) Figure 5.1: (e) TIC of Exhibit MSU 900-01 after HS-SPME and GC-MS analysis using Temperature Program A. 106 2 an intermediate produced during MDMA production by either isosafrole or piperonal. Trace amounts of nicotine were also observed in Exhibits T-17, T-27, T-29, and T-30 using HS-SPME only, because of the pre-concentration of the fiber. A higher background and several additional peaks were also present in the HS-SPME TICs because of the fiber itself. The polydimethylsiloxane/divinylbenzene (PDMS/DVB) coating on the SPME fiber slowly degraded in the heated GC inlet and released compounds onto the column, which were observed in the TICs. These compounds are not labeled in Figure 5.1 as they did not contribute to the MDMA impurity profiles. Along with observation of additional compounds, pre-concentrating the fiber leads to peak broadening throughout the chromatograms. Because a greater mass of each compound was carried onto the GC column using HS-SPME, compounds that were already at a high abundance using LLE overloaded the column when extracted using the HS-SPME procedure. This is illustrated in Figure 5.2 which shows chromatograms of Exhibit T-17 extracted using HS-SPME (top) and LLE (bottom). Baseline separation was not observed for all peaks using HS-SPME, and many of the major compounds observed using LLE displayed extensive fronting using HS-SPME due to overloading the GC. 107 6.00E+07 0.00E+00 1.00E+06 0.00E+00 5 10 15 20 25 Retention Time (min) Figure 5.2: TICs of Exhibit T-17 after HS-SPME (top) and LLE (bottom) and GC-MS analysis using Temperature Program A. 108 5.2 MDMA Exhibits Extracted by HS-SPME and Analyzed using Temperature Program D Each of the compounds identified in the five MDMA exhibits when analyzed using Temperature Program A was also observed in all extracts using Temperature Program D. However, P2P formed a more abundant peak with Temperature Program D than with Temperature Program A because the compound eluted during a hold in the program, which allowed sufficient time for the compound to more fully elute from the column. This was in contrast to Temperature Program A, in which the same compound eluted during a temperature ramp. As an example, Figure 5.3 shows Exhibit T-17 analyzed by Temperature Program D with the major compounds identified. 5.3 Association and Discrimination of Exhibits Extracted by HS-SPME based on PCA using Selected Compounds Due to the peak broadening that was observed using the HS-SPME procedure, retention time misalignments of compounds were observed throughout the TICs. Since misalignments are a source of non-chemical variance that is introduced into the data, the full TICs were not used for data analysis. Instead, PCA was performed based on selected compounds, as was done for the LLE data. More compounds were consistently observed using HS-SPME than LLE, therefore PCA was performed based on more compounds for the HS-SPME data than for the LLE data (Table 5.1 and Table 4.1). Since P2P could not be integrated at a threshold above baseline in all exhibits, this compound was not used in data analysis. Diethyl phthalate and MDEA/MDDMA could not be resolved from one another in one extract of Exhibit T-17 using Temperature Program D, and, again, were not used in data analysis. Extracts of each exhibit were normalized by total area in both datasets (one dataset per temperature program) prior to PCA based on the selected compounds (Table 5.1). 109 5.50E+07 MDP2P-OH methamphetamine MDEA/MDDMA caffeine MDP2P MDMA diethyl phthalate MD-DMB nicotine isothiocyanic acid N-formyl methamphetamine N-formyl MDMA lidocaine P2P phenol piperonal 0.00E+00 5 10 15 20 25 30 Retention Time (min) Figure 5.3: TIC of Exhibit T-17 after HS-SPME and GC-MS analysis using Temperature Program D. 110 35 Table 5.1 Selected compounds for data analysis of MDMA exhibits extracted by HS-SPME using Temperature Programs A and D. Temperature Program Compound A D phenol   P2P  methamphetamine   piperonal   MDP2P   MDP2P-OH   MDMA   N-formyl MA   diethyl phthalate  MDEA/MDDMA  N-formyl MDMA   Note: Check indicates compound present and used in data analysis. 5.3.1 Association and Discrimination of Exhibits Extracted by HS-SPME and Analyzed using Temperature Program D based on PCA using Selected Compounds The PCA scores plot based on nine selected compounds for the five exhibits extracted using HS-SPME and analyzed using Temperature Program D is shown in Figure 5.4, in which the first two PCs accounted for 99.17% of the total variance. Extracts of each exhibit were closely clustered, with some spread observed along PC2 in extracts of Exhibit MSU 900-01. Exhibit T-30 was the only exhibit positioned positively on PC1, and Exhibits T-29 and MSU 900-01 were the only exhibits positioned positively on PC2. Positioning of the exhibits could be explained with reference to the corresponding loadings plot, which is shown in Figure 5.5, with PC1 on the x-axis and PC2 on the y-axis. From the loadings plot, MDMA and MDP2P-OH contributed most to the variance described: MDMA had the highest positive weighting on PC1 while MDP2P-OH had the highest positive weighting on PC2. Methamphetamine, N-formyl methamphetamine, and N-formyl MDMA were each 111 PC2 (20.31%) 8.00E+09 -2.00E+10 2.00E+10 -8.00E+09 PC1 (78.86%) Figure 5.4: PCA scores plot for five MDMA exhibits after HS-SPME and GC-MS analysis using Temperature Program D. 112 1.00E+00 PC2 phenol P2P methamphetamine piperonal MDP2P MDP2P-OH MDMA N-formyl MA N-formyl MDMA -1.00E-01 -3.00E-01 1.00E+00 PC1 Figure 5.5: PCA loadings plot for five MDMA exhibits after HS-SPME and GC-MS analysis using Temperature Program D. 113 positioned slightly positively on both PCs. The remaining compounds were positioned at, or close to, zero on both PCs and therefore had little contribution to the positioning of the exhibits in the scores plot. In the scores plot (Figure 5.4), Exhibit T-17 was positioned negatively on both PC1 and PC2. Of the selected compounds used for data analysis, this exhibit was composed mainly of MDP2P-OH, methamphetamine, and N-formyl MDMA. Since this exhibit contained very little MDMA (the lowest abundance of all exhibits), this compound contributed negatively in the mean-centered data for this exhibit. When this negative value was multiplied by the positive eigenvector for MDMA, which had the greatest effect on positioning on PC1 (Figure 5.5), the result was a negative number. Therefore, this exhibit was positioned most negatively on PC1. Similarly, the abundance of MDP2P-OH in Exhibit T-17 was slightly below average. Thus, when the mean-centered data were considered, MDP2P-OH had a negative contribution that, when multiplied by the positive eigenvector of this compound on PC2, positioned the exhibit negatively on PC2. Exhibit T-27 was positioned negatively on both PCs in the scores plot (Figure 5.4), and was the most negatively positioned exhibit on PC2. This exhibit was dominated by ketamine, which was not among the selected compounds considered in data analysis. Similar to Exhibit T-17, positioning of this exhibit was influenced by the lower abundance of MDMA and MDP2P-OH. Although MDMA was the second most prominent peak in this exhibit, when the mean-centered data were considered this compound still had a negative contribution. When multiplied by the positive eigenvector for MDMA on PC1 (Figure 5.5), the resulting positioning for this exhibit was negative. On PC2, the positioning was influenced by the low abundance of MDP2P-OH in this exhibit. When the mean-centered data were again considered, a negative 114 value for this compound was multiplied by the positive eigenvector on PC2, resulting in an overall negative contribution to the positioning on this PC. Exhibit T-29, which was dominated by MDP2P-OH, was positioned negatively on PC1 and positively on PC2 (Figure 5.4). Positioning on PC1 was due to low abundances of methamphetamine and MDMA, which contributed negatively when the mean-centered data were considered. Both of these compounds were weighted positively on PC1, therefore, when multiplied by the negative mean-centered values for this exhibit, Exhibit T-27 was positioned negatively on PC1. This exhibit was positioned positively on PC2 because of the high abundance of MDP2P-OH, which was weighted positively on this PC (Figure 5.5). Exhibit T-30 was positioned positively on PC1 and negatively on PC2 (Figure 5.4). This exhibit contained a large abundance of MDMA, MDP2P-OH and methamphetamine. Compared to other exhibits, the abundance of MDMA in Exhibit T-30 was so large that it was the only exhibit with a positive contribution in the mean-centered data. Therefore, Exhibit T-30 was the only exhibit to load positively on PC1. Although methamphetamine and MDP2P-OH had a positive effect on the positioning on PC2 (Figure 5.5), the negative contribution from MDMA outweighed the positive contributions, thus positioning this exhibit negatively, but close to zero, on PC2. Exhibit MSU 900-01 contained the largest abundance of MDP2P-OH in the five exhibits, and was positioned negatively on PC1 and most positively on PC2, with some spread among extracts on PC2 (Figure 5.4). On PC1, a low abundance of MDMA influenced the positioning of this exhibit negatively. However, the overall positioning was close to zero due to the positive contributions of MDP2P-OH and methamphetamine in this exhibit. On PC2, Exhibit MSU 900-01 was positioned most positively because of the high abundance of MDP2P-OH, 115 which had the greatest effect on positioning on this PC (Figure 5.5). Slight differences in the total area of MDP2P-OH were observed among extracts in this exhibit, which was observed on PC2 as spread among the extracts. 5.3.2 Association and Discrimination of Exhibits Extracted by HS-SPME and Analyzed using Temperature Program A based on PCA using Selected Compounds The PCA scores plot based on ten selected compounds for the five exhibits extracted using HS-SPME and analyzed using Temperature Program A is shown in Figure 5.6, in which the first two PCs accounted for 97.87% of the total variance among the exhibits. This was slightly less than Temperature Program D (99.17%), but still higher than all temperature programs using LLE. This means that nearly all of the variance, both instrumental and chemical, among the selected compounds could be accounted for using just two PCs when the exhibits were extracted by HS-SPME. In the scores plot for HS-SPME using Temperature Program A, the five exhibits were clearly distinguished, with minor spread among extracts of Exhibits T-17, T-29, T-30 and MSU 900-01, mostly along PC2. Exhibits T-30 and MSU 900-01 were positioned positively on PC1, with Exhibit MSU 900-01 close to zero, and the remaining three exhibits positioned negatively on this PC. On PC2, Exhibits MSU 900-01, T-17, and one extract of T-29 were positioned positively, while Exhibits T-27, T-30, and the remaining two extracts of T-29 were positioned negatively. The positioning of each exhibit in the scores plot could be explained with reference to the loadings plot, which is shown in Figure 5.7. Since the same compounds were observed in the TICs using both temperature programs, and there were few differences in the abundances of the compounds, the compounds contributing most to the variance were the same for both Temperature Programs A and D. Therefore, the 116 PC2 (22.27%) 8.00E+09 -2.00E+10 2.00E+10 -8.00E+09 PC1 (75.60%) Figure 5.6: PCA scores plot for five MDMA exhibits after HS-SPME and GC-MS analysis using Temperature Program A. 117 1.00E+00 PC2 phenol methamphetamine piperonal MDP2P MDP2P-OH MDMA N-formyl MA diethyl phthalate MDEA/MDDMA N-formyl MDMA -1.00E-01 1.00E+00 PC1 -4.00E-01 Figure 5.7: PCA loadings plot for five MDMA exhibits after HS-SPME and GC-MS analysis using Temperature Program A. 118 loadings plots for these two datasets were very similar. Like the loadings plot for Temperature Program D, MDMA had the greatest positive effect on positioning on PC1, and MDP2P-OH had the greatest positive effect on positioning on PC2 using Temperature Program A (Figure 5.7). Methamphetamine, N-formyl methamphetamine, and N-formyl MDMA were each weighted positively on both PCs, while MDEA/MDDMA was weighted negatively on PC1 and slightly positively on PC2. The remaining four compounds were positioned close to zero on both PCs and therefore had little effect on the positioning of exhibits in the scores plot. Since the TICs (Figures 5.1 and 5.3) and PCA loadings plots (Figures 5.5 and 5.7) were similar for the two temperature programs, positioning of exhibits in the scores plot for Temperature Program A (Figure 5.6) was very similar to the positioning observed for Temperature Program D (Figure 5.4), for the same reasons as described previously (section 5.4.1). Despite this similarity, some differences were observed in the scores plot: using Temperature Program A, Exhibit T-17 was positioned positively on PC2, Exhibit T-29 was positioned more negatively, around zero, on PC2, and Exhibit MSU 900-01 was positioned positively on PC1. These differences were due mostly to the inclusion of MDEA/MDDMA as one of the selected compounds for Temperature Program A. Exhibit T-17 contained the greatest abundance of MDEA/MDDMA of all exhibits, resulting in a positive contribution on PC2 by this compound (Figures 5.6 and 5.7). In addition, the low abundance of MDMA in Exhibit T-17 using Temperature Program A contributed negatively to the mean-centered data. When multiplied by the negative loading for MDMA on PC2, the result was a positive loading for MDMA on PC2. The combination of these factors led to an overall positive positioning of Exhibit T-17 on PC2, using Temperature Program A. 119 Exhibit T-29 was positioned more negatively, around zero, on PC2 using Temperature Program A (Figure 5.6) compared to Temperature Program D. This exhibit contained a low abundance of MDEA/MDDMA. Therefore, this compound contributed negatively to the positioning in the scores plot when the mean-centered data were considered. In addition, the abundance of MDP2P-OH was below average in this exhibit using Temperature Program A but above average using Temperature Program D. Since MDP2P-OH was weighted positively on PC2 (Figure 5.7 and Figure 5.5) for both temperature programs, this compound had a negative loading when the mean-centered data were considered for Temperature Program A, but a positive loading in Temperature Program D. In addition, variability in the abundance of both MDMA and MDP2P-OH observed among extracts of this exhibit led to the spread observed along PC2 in the scores plot. Exhibit MSU 900-01 was positioned positively on PC1 using Temperature Program A (Figure 5.6). This exhibit contained a low abundance of MDEA/MDDMA, resulting in a negative contribution in the mean-centered data. When multiplied by the negative eigenvector for PC1, there was a positive contribution from this compound. This, in addition to the positive contribution of MDP2P-OH in this exhibit, positioned Exhibit MSU 900-01 positively on PC1. However, the overall positioning of the exhibit remained close to zero due to the negative contribution of MDMA in this exhibit, as a result of the low abundance of this compound. 5.4 Comparison of Association and Discrimination of Exhibits Extracted by LLE and HS-SPME based on PCA using Selected Compounds PCA scores plots for the five MDMA exhibits extracted using HS-SPME and analyzed using Temperature Programs A and D were visually compared to the scores plots generated after extraction using LLE. In general, greater discrimination of exhibits was observed using 120 HS-SPME over LLE, which is evidenced by the scales for the scores plots (Figures 5.8 and 5.9). In both figures, the scales for the HS-SPME scores plot are two orders of magnitude greater than those for the LLE scores plots. This means that, although the separation of exhibits appears to be the same or only slightly better using HS-SPME, it was actually much greater. However, this also means that spread among extracts was greater using HS-SPME. Detailed comparisons of the two temperature programs are given in the subsequent sections. 5.4.1 Comparison of Association and Discrimination of Exhibits Analyzed using Temperature Program D Using Temperature Program D, each exhibit formed a distinct cluster using both extraction procedures. As mentioned above, the discrimination of exhibits was greater using HS-SPME than LLE. This was likely because of the additional compounds used in data analysis. Since more compounds were consistently extracted using HS-SPME than LLE, nine compounds were used in the data analysis using HS-SPME, compared to only six compounds using LLE. These additional compounds provided more discrimination, which allowed for better separation of exhibits. Despite greater discrimination, more spread was observed among extracts of each exhibit. This is due to HS-SPME being a less reproducible extraction procedure than LLE because of the variability in the amount of sample extracted from the headspace onto the fiber. In the TICs for the HS-SPME extracts, peak areas ranged in relative standard deviations (RSDs) among extracts from 1.55% to 32.19%, and averaged 12.53%. These were higher than the RSDs for LLE replicates, which ranged from 1.25% to 28.73% and averaged 6.79%. The increased variability using HS-SPME was evidenced by the increased spread among extracts in the scores plot. 121 8.00E+09 PC2 (27.19%) PC2 (20.31%) 8.00E+07 -1.50E+08 1.50E+08 -8.00E+07 -2.00E+10 PC1 (58.36%) 2.00E+10 -8.00E+09 PC1 (78.86%) Figure 5.8: PCA scores plots based on selected compounds for five MDMA exhibits extracted using LLE (left) and HS-SPME (right) and analyzed by GC-MS using Temperature Program D. 122 8.00E+09 PC2 (12.19%) PC2 (22.27%) 6.00E+07 -1.50E+08 1.50E+08 -6.00E+07 -2.00E+10 PC1 (78.66%) 2.00E+10 -8.00E+09 PC1 (75.60%) Figure 5.9: PCA scores plots based on selected compounds for five MDMA exhibits extracted using LLE (left) and HS-SPME (right) and analyzed by GC-MS using Temperature Program A. 123 5.4.2 Comparison of Association and Discrimination of Exhibits Analyzed using Temperature Program A Using Temperature Program A, the five exhibits formed five distinct clusters in the scores plot after extraction using HS-SPME but not after using LLE. As mentioned previously, the additional discrimination observed using HS-SPME was likely a result of the additional compounds used for data analysis: ten compounds were used for HS-SPME analysis using this temperature program compared to only seven compounds for the LLE extracts. Although there appeared to be more spread among replicates using LLE, when these two scores plots were evaluated using the same scale, more spread was actually observed among extracts using HS-SPME. RSDs were also higher for HS-SPME than LLE, ranging from 11.37% to 57.73% and averaging 22.78% using HS-SPME, compared to a range of 8.59% to 29.84% and average of 18.82% using LLE. This, again, was due to the poor reproducibility of this extraction procedure, which led to greater variability in peak areas. 5.5 Summary of Findings using HS-SPME Using PCA to analyze the HS-SPME data, it was determined that both Temperature Programs A and D were successful in discriminating the five MDMA exhibits. Compared to LLE, more spread was observed among extracts using HS-SPME due to the irreproducibility of this extraction procedure. However, greater discrimination of exhibits was observed using HS-SPME. This was especially true for Temperature Program A, in which five separate clusters were visible in the HS-SPME scores plot, but were not clearly distinguished using LLE. In addition, several compounds indicative of MDMA synthesis were observed using HS-SPME that were either present at very low abundances or were not observed at all using LLE. Despite the 124 irreproducibility of the HS-SPME procedure, the additional compounds observed provide beneficial information for determining the MDMA synthesis route that may make this extraction procedure favorable for profiling purposes. 125 REFERENCES 126 REFERENCES 1. Bonadio F, Margot P, Delémont O, Esseiva P. Optimization of HS-SPME/GC-MS analysis and its use in the profiling of illicit ecstasy tablets (Part 1). Forensic Sci Int 2009;187:73-80. 2. Gimeno P, Besacier F, Chaudron-Thozet H, Girard J, Lamotte A. A contribution to the chemical profiling of 3,4-methylenedioxymethamphetamine (MDMA) tablets. Forensic Sci Int 2002;127:1-44. 127 Chapter 6 Conclusions and Future Work 6.1 Conclusions The aim of impurity profiling is to determine the synthesis method used to produce illicit substances, and to compare the additives and impurities in these substances. The information obtained from impurity profiles can be used to link exhibits to each other or to a clandestine laboratory based on similarities observed between separate exhibits. However, MDMA profiles cannot currently be compared between forensic laboratories due to the different extraction and analysis procedures used. This research was intended to compare current MDMA impurity profiling methods and determine if any were more beneficial than others. Establishing a standard extraction procedure and GC temperature program would allow for comparisons of MDMA profiles between forensic laboratories and would greatly benefit law enforcement in their ability to link exhibits to one another and to clandestine laboratories. The objectives of this research were to use statistical procedures to investigate the effects of extraction procedure and GC temperature program on association and discrimination of MDMA exhibits based on chemical profiles. Five MDMA exhibits were extracted using a previously optimized liquid-liquid extraction (LLE) procedure and then analyzed by four temperature programs published in the literature. The resulting chromatograms were evaluated using principal components analysis (PCA), and comparisons of the four temperature programs were made. The same five MDMA exhibits were then extracted using a previously optimized headspace solid phase microextraction (HS-SPME) procedure, also published in the literature, and then analyzed by two of the temperature programs: that which resulted in the most discrimination of exhibits using LLE, and that which resulted in poor discrimination of exhibits and the most spread among replicates using LLE. 128 6.1.1 Effect of Temperature Program on MDMA Impurity Profiles Five MDMA exhibits were analyzed in replicate by each of the four temperature programs after extraction using LLE. The chromatograms from all replicates of all exhibits formed a dataset for each temperature program, and the four datasets were each subjected to alignment and normalization pretreatment procedures. Following data pretreatment, the datasets were analyzed using PCA. When the total ion chromatograms (TICs) were analyzed, only Temperature Program D was successful at discriminating the five MDMA exhibits. Retention time misalignments were observed throughout the chromatograms and could not be effectively corrected using a correlation optimized warping (COW) alignment algorithm. Therefore, impurity profiles were assessed based on peak area of selected compounds instead of the TICs. This eliminated the issue of misalignments because retention time was no longer considered. When each of the temperature programs was analyzed using selected compounds, improvement in discrimination of exhibits was observed. However, Temperature Program D was still most efficient at discriminating exhibits and closely associating replicates, and Temperature Program A was still the least efficient. Temperature Program C was also successful in discriminating the five exhibits, but displayed spread among replicates. Temperature Program B displayed closer clustering of replicates, but exhibits T-17 and MSU 900-01 could not be distinguished on the scores plot. In general, Exhibits T-27 and T-29 were most distinguishable on the scores plot. Exhibit T-27 contained a high abundance of MDMA, while Exhibit T-29 contained high abundances of MDMA and MDP2P-OH. These two compounds consistently contributed most to the variance, therefore these two exhibits were positioned apart from the rest on the scores plots. Exhibits T-17 129 and MSU 900-01 were located closest to each other on the scores plot using each temperature program, and were sometimes indistinguishable. This was because these two exhibits contained similar concentrations of most selected compounds, including low abundances of MDP2P-OH, MDMA and N-formyl MDMA. Exhibit T-30, which also contained a low abundance of MDMA, was positioned near Exhibits T-17 and MSU 900-01 as well, but was typically distinguishable because it contained a higher abundance of MDP2P-OH and the greatest abundance of N-formyl MDMA of all exhibits. The spread among replicates using Temperature Program A was attributed to variability in the MDP2P-OH peak, which eluted after a hold and at the start of a slow ramp using this temperature program. This led to more fluctuation in the instrument, which was evidenced by a high relative standard deviation (RSD) for this compound using Temperature Program A. In contrast, MDP2P-OH eluted during a hold using Temperature Program D. The instrument, therefore, was more stable, resulting in a low RSD for this compound. Although Temperature Program D was the most successful at clustering replicates and discriminating the five exhibits, it was also the longest temperature program. At a total length of 53 minutes, this program is not ideal for implementation in a forensic laboratory because it greatly reduces the number of samples that can be analyzed daily. While the 12-minute hold in the middle of this temperature program provided better precision in peak abundance, it may be longer than is necessary. Minor adaptations to the temperature program, such as shortening this hold time, may make it more useful in a forensic laboratory, while still keeping the features that made it the most successful at associating replicates and discriminating MDMA exhibits. 130 6.1.2 Effect of Extraction Procedure on MDMA Impurity Profiles Differences were observed between chromatograms obtained after the exhibits were extracted using the LLE procedure and those obtained following extraction using the HS-SPME procedure. In general, chromatograms displayed peaks that were more Gaussian-shaped using LLE than HS-SPME, and the peak areas among replicates were more reproducible. However, more trace compounds were extracted using HS-SPME, providing more beneficial information to determine synthesis routes. While both extraction procedures were successful at discriminating exhibits using Temperature Program D, improvements were observed using Temperature Program A when HS-SPME was the extraction procedure compared to LLE. The differences observed in the chromatograms were reflected in the PCA scores plots for the two extraction procedures. On the scores plots, there was greater discrimination of exhibits but also greater spread among replicates, when HS-SPME was used as the extraction procedure. Greater discrimination of exhibits was due to the greater number of selected compounds that could be used for analysis, which provided more discriminatory power. However, due to HS-SPME being a less reproducible extraction procedure, greater spread was observed among replicates using this extraction procedure for both temperature programs. The temperature program did not have as great of an effect on HS-SPME data; the five MDMA exhibits were successfully discriminated using both Temperature Programs A and D when HS-SPME was used as the extraction procedure. Minor fluctuations in the instrument due to the temperature program were visible as spread among replicates when LLE was used as the extraction procedure. However, due to the irreproducibility of the HS-SPME extraction procedure, spread among replicates existed due to the extraction procedure itself, therefore 131 masking the minor effects of the temperature program. Thus, both temperature programs displayed five distinct clusters, but with some spread among replicates. 6.2 Future Work After analyzing the chromatograms collected using different GC temperature programs, it is apparent that some are more successful than others. However, the most successful temperature program was also the longest, and is not practical for use in a forensic laboratory. Future investigations should focus on determining the characteristics that made temperature programs successful (e.g. hold at 150 °C) and then optimize these characteristics (e.g. shorter hold period). Other parameters, such as final temperature and hold time, could be decreased since there were typically no compounds eluting at the highest temperature or during the final few minutes of the program. Although Temperature Programs B and C were not analyzed by HS-SPME due to instrument time limitations, it may be beneficial to analyze HS-SPME samples using at least Temperature Program C, which was the next most successful at discriminating exhibits using LLE. This temperature program, similar to Temperature Program D, contained a hold at 150 °C, but the hold was 5.5 minutes compared to 12 minutes using Temperature Program D. The overall analysis time of this temperature program was 39.5 minutes, which is slightly more practical in a forensic laboratory setting than the 53 minutes for Temperature Program D. More MDMA exhibits should be used, if possible, to increase the dataset available for statistical analysis. Since MDMA tablets contain such an array of impurities and additives, it is necessary to compare temperature programs and extraction procedures using a larger sample population, to ensure the optimal procedures are truly optimal for the wide range of samples that 132 could potentially be analyzed. While the five MDMA exhibits analyzed in this research were all different, each contained similar starting materials (e.g. MDP2P) and adulterants (e.g. caffeine), and did not sufficiently cover the diversity of MDMA impurities. In addition, or if further exhibits could not be obtained, simulated exhibits could be used. Mixtures simulating MDMA tablets could be created using reference standards for compounds commonly found in the exhibits (e.g., methamphetamine, piperonal, MDP2P, MDMA, MDEA, N-formyl MA, caffeine, N-formyl MDMA, etc.). These mixtures could have varying abundances of each of the compounds, which could then be extracted by LLE and HS-SPME and analyzed using the four temperature programs. This would allow for more control over the individual compound concentrations, allowing optimization of the procedures for a range of compound concentrations. In this research, data analysis was performed using PCA. Future work could include additional data analysis procedures, such as Pearson product moment correlation (PPMC) coefficients, soft independent modeling classification analogy (SIMCA), or degree of class separation. These data analysis procedures would provide numerical values to describe the variability in the datasets, and may allow for further distinction of MDMA exhibits, beyond the visual discrimination of PCA scores plots. In conclusion, further investigation of GC temperature programs needs to be performed since some programs were more efficient at associating replicates and discriminating MDMA exhibits than others. New temperature programs could be created using the beneficial aspects of the temperature programs investigated in this research, while keeping the overall analysis time short for practical implementation in a forensic laboratory. While peak areas using HS-SPME were less precise than LLE and, in general, the chromatograms were less reproducible, several 133 additional compounds were extracted using this procedure over LLE. Because the information gained using HS-SPME is highly beneficial in determining MDMA synthesis route, this extraction procedure shows great potential and should not be discounted simply because it is a less reproducible method than LLE. Combining information obtained using both extraction procedures may be the most effective method for associating replicates and discriminating MDMA exhibits using PCA. 134