Composition profiling of seized ecstasy tablets by Raman spectroscopySteven E. J. Bell , D. Thorburn Burns , Andrew C. Dennis , Lindsay J. Matchett and James S. SpeersAnalyst 125(10), 1811-1815 (2000)
(
http://dx.doi.org/10.1039/b005662f)
AbstractRaman spectroscopy with far-red excitation has been investigated as a simple and rapid technique for composition profiling of seized ecstasy (MDMA, N-methyl-3,4-methylenedioxyamphetamine) tablets. The spectra obtained are rich in vibrational bands and allow the active drug and excipient used to bulk the tablets to be identified. Relative band heights can be used to determine drug/excipient ratios and the degree of hydration of the drug while the fact that 50 tablets per hour can be analysed allows large numbers of spectra to be recorded. The ability of Raman spectroscopy to distinguish between ecstasy tablets on the basis of their chemical composition is illustrated here by a sample set of 400 tablets taken from a large seizure of >50000 tablets that were found in eight large bags. The tablets are all similar in appearance and carry the same logo. Conventional analysis by GC-MS showed they contained MDMA. Initial Raman studies of samples from each of the eight bags showed that despite some tablet-to-tablet variation within each bag the contents could be classified on the basis of the excipients used. The tablets in five of the bags were sorbitol-based, two were cellulose-based and one bag contained tablets with a glucose excipient. More extensive analysis of 50 tablets from each of a representative series of sample bags gave distribution profiles that showed the contents of each bag were approximately normally distributed about a mean value, rather than being mixtures of several discrete types. Two of the sorbitol-containing sample sets were indistinguishable while a third was similar but not identical to these, in that it contained the same excipient and MDMA with the same degree of hydration but had a slightly different MDMA/sorbitol ratio. The cellulose-based samples were badly manufactured and showed considerable tablet-to-tablet variation in their drug/excipient ratio while the glucose-based tablets had a tight distribution in their drug/excipient ratios. The degree of hydration in the MDMA feedstocks used to manufacture the cellulose-, glucose- and sorbitol-based tablets were all different from each other. This study, because it centres on a single seizure of physically similar tablets with the same active drug, highlights the fact that simple physical descriptions coupled with active drug content do not in themselves fully characterize the nature of the seized materials. There is considerable variation in the composition of the tablets within this single seizure and the fact that this variation can be detected from Raman spectra demonstrates that the potential benefits of obtaining highly detailed spectra can indeed translate into information that is not readily available from other methods but would be useful for tracing of drug distribution networks.
IntroductionRaman spectroscopy has been proposed as a useful method for screening of seized tablets and powders for illicit substances. 1–6 The two features of the technique which make it attractive for this purpose are the ability to record spectra with no sample preparation and the reasonably short (<1 min) collection times required. We have previously shown that Raman methods can be used to distinguish between ecstasy (i.e. MDMA, N-methyl-3,4-methylenedioxyamphetamine) and various other phenethylamine ecstasy analogues commonly seized in the UK. The compounds can be identified even when they are mixed with bulking agents (excipients) in seized tablets, as well as when they are presented as pure samples. 7 In itself, this rapid identification of the active drugs is useful but the fact that the spectra can also be used to identify excipient(s), the relative concentration of drug to excipient and even the degree of hydration of the active compounds means that the Raman spectrum of a seized sample has the potential to provide a more complete profile of the composition of seized tablets than other, more established, analytical methods. For example, gas chromatography-mass spectrometry (GC-MS) analysis, which is currently the method of choice for criminal prosecutions, 8 can be used to identify and quantify the active compound present but does not give any other information on tablet composition.
Seized ecstasy tablets are typically marked with an identifying logo (a form of product branding): these logos can be symbols of any type, such as animals or birds, but they are often copies of the logos of internationally recognized brand names, such as Rolex or McDonalds. Unfortunately, the logos are not a reliable guide to the source or composition of tablets, not least because as brands become popular the logos they carry are simply copied by other illicit manufacturers.
Since Raman spectra can be acquired rapidly it has been suggested that it should be possible to test hundreds of samples in a few hours, although to our knowledge this has not actually been done. In contrast, with conventional GC-MS analysis only small numbers of individual tablets are normally analysed because of the time taken for each measurement. The ability to test relatively large numbers of samples within reasonable times is obviously a useful characteristic of any method, since it increases the sample throughput and reduces cost, but in the case of ecstasy seizures it is particularly useful because it opens up the possibility of determining the degree of homogeneity within a single sample set composed of hundreds of apparently similar tablets. This information on the composition of many tablets might be used to determine if all the tablets in large seizures are simply a single batch which has been sub-divided for distribution or whether they have actually been manufactured in a different way and only later tableted in the same form. Similarly, information on sample homogeneity is needed if the composition data are to be used to determine whether samples that have been seized at a different time (and possibly even carry different identifying marks) in fact have the same chemical constituents and therefore have the same origin, which their manufacturers have tried to conceal by changing logos etc. In these comparisons of different seizures, the degree of homogeneity within the two sample sets to be compared must be known. Clearly with very inhomogeneous mixed batches of tablets sampling errors could easily lead to erroneous conclusions about the similarity of seized samples from different sources.
The purpose of this paper is to explore the extent to which possible potential advantages of Raman profiling of composition can be realised in practice. In particular, a large sample set consisting of 400 tablets taken from a single seizure has been studied. This seizure was composed of a series of bags of physically-similar tablets all bearing the same Mitsubishi logo, so it is almost ideal for testing the discriminating power of Raman methods.
ExperimentalRaman spectra were recorded using 810 nm excitation (Spectra-Physics (San Jose, CA, USA, Ti/sapphire laser pumped by a Spectra-Physics 2020 Ar+ laser, typically 70 mW at sample) using a 180° backscattering geometry. Typically, the laser was line focused (<100 m × 10 mm) with a cylindrical lens onto the sample, which was mounted on a rotating platform so that an averaged spectrum from the tablet was obtained. Scattered light was collected, passed through a Kaiser Optical Systems (Ann Arbor, MI, USA) holographic notch filter and then dispersed by a Jobin-Yvon (Longjumeau, France) HR640 single stage spectrograph (600 lines per mm grating) onto a Princeton Instruments (Trenton, NJ, USA) LN1152 liquid N2 cooled charge-coupled device (CCD) detector. Spectra were typically accumulated for 40 s and were exported to the Labcalc spectral manipulation package for processing and presentation. 9 Due to the nature of the samples all the spectra were superimposed on a smooth background of stray light which rose smoothly to the low cm1 end of the spectrum. This background was removed from all the spectra shown by digitally subtracting a similar stray light signal which was generated by placing a piece of normal blackboard chalk in the sample position. The spectrometer was calibrated using a standard 50/50 mixture of toluene and acetonitrile. 10 The positions of strong sharp bands are considered accurate to ± 2 pixels (ca. 3 cm1).
Seized ecstasy tablets were obtained from the Forensic Service Agency for Northern Ireland and were used as received.
ResultsAt the outset, the only information available for the samples was that they were in eight bags, each containing 50 off-white tablets with the Mitsubishi logo. Each of these sets of 50 tablets had been taken at random from one of the eight very large bags of tablets that comprised the original seizure of >50000 tablets. Conventional analysis of a random selection of the tablets had shown that they all contained MDMA. Due to the large number of tablets received, a preliminary data set consisting of the Raman spectra of 6 tablets from each bag was obtained, so that gross variations in MDMA and excipient content between the various sample bags could be determined. Fig. 1 shows this data as the averages of the spectra of each set of 6 tablets. All the spectra show clear sets of bands due to their MDMA content (marked on the figure as bands 716, 771 and 810 cm1) but they also have features due to the excipients present. Cursory examination of the spectra in Fig. 1 shows that the tablets in Item 1, bags 1, 2, 3 and 4, and Item 2, bag 3 are similar, as are Item 2, bags 1 and 2, leaving Item 2, bag 4 as a unique set. Since the samples clearly divide into three main types we have re-labelled the samples as 1(a)–(e), 2(a) and (b), and 3. All the samples 1(a)–(e) contain the same excipient, which was identified as sorbitol (band at 880 cm1) by comparison with the spectrum of pure sorbitol, samples 2(a) and (b) have cellulose as the excipient (band at 1124 cm1), while sample 3 is glucose-based (band at 896 cm1).
The spectra shown in Fig.1 were all collected from tablets which were mounted on a rotating platform in the spectrometer and with the laser line-focused (not point-focused) onto them. This sampling method was chosen so that the spectra were obtained over as large an area of the tablet face as possible, minimizing any problems due to inhomogeneity of the distribution of drug and excipient throughout the tablets. Illegally-produced tablets are not manufactured under conditions of strict quality control, so that sample inhomogeneity due to poor mixing of drug and excipient is likely to be encountered.
To determine the extent of any inhomogeneity, a single tablet from sample bag 1(e) was chosen at random for more detailed study. This tablet was mounted on a linear translation stage and the laser spot-focused rather than line-focused onto its surface. A series of spectra were then recorded as the sample was moved in 100 m increments across the diameter of the tablet. A plot of the absolute peak heights of the MDMA band at 810 cm1 and the excipient band at 880 cm1 across the tablet is shown in Fig. 2. It is clear from the figure that there is not a homogeneous distribution of MDMA or excipient throughout the tablet.
Even without further analysis, it is clear that the Raman data in Fig. 1 can be used to show that all the tablets studied do indeed contain MDMA (as opposed to MDA or MDEA, for example) and further, that they broadly classify into the three groups labelled 1, 2 and 3 which have different excipients. However, the spectra clearly contain more information than these two simple measures of composition and measurement of appropriate relative peak heights or areas can be used to extract quantitative data on, for example, the relative amounts of drug and excipient present.
In the initial data set, analysis of the variance of the relative peak heights of the strongest bands of the MDMA and excipients showed that the sets of 6 tablets taken from four of the bags with sorbitol-containing tablets (samples 1(a) and 1(c)–(e)) were indeed statistically similar, giving an F value of 0.58 compared with the critical value of 3.10 at the 5% confidence level. However, somewhat surprisingly, the tablets taken from the fifth sorbitol bag, (sample 1(b)) were found to be slightly different. The sets of tablets taken from the bags with cellulose excipient (sample 2(a) and (b)) were indistinguishable from each other (F = 0.062, Fcrit = 4.96 at the 5% level), while sample 3 was unique.
In the preliminary data we found a considerable degree of variation in drug/excipient ratios, even in tablets taken from the same bag. To provide more statistically-valid data and to confirm that the variation within each sample set was not due to the bags containing mixtures of just two or three distinct types of tablets, each with a different fixed composition, we recorded spectra of all 50 tablets from each of a representative series of sample bags. The samples chosen were: two of the bags with similar apparent drug/sorbitol ratios (1(a) and (e)), the sorbitol-containing sample (1(b)) which the preliminary data had suggested was different from the others, one cellulose-containing sample (2(a)) and the glucose-containing sample 3.
Fig. 3 shows the averages of the 50 spectra taken for each of the three sorbitol-containing samples. In all the 150 spectra used to make these averages there were variations in relative peak heights etc. but all the tablets were of the same general type i.e. there were no anomalous stray glucose- or cellulose-containing tablets in the sorbitol samples, similarly the bags of cellulose- or glucose-containing samples did not contain any sorbitol tablets.
In the spectra shown in Fig. 3 there is a clear difference in the relative intensities of the strongest MDMA (810 cm
-1) and sorbitol (880 cm
-1) bands between samples from bag 1(b) and the samples from the other 2 bags examined. However, comparison of averaged spectra can be misleading and to make a well-founded conclusion about whether samples are indeed statistically similar or distinct it is necessary to look at the distribution of values as well as their averages.