Nitrogen-14 Nuclear Quadrupole Resonance Spectroscopy: A

Feb 5, 2013 - ABSTRACT: We report the detection and analysis of a suspected counterfeit sample of the antimalarial medicine Metakelfin through develop...
4 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Nitrogen-14 Nuclear Quadrupole Resonance Spectroscopy: A Promising Analytical Methodology for Medicines Authentication and Counterfeit Antimalarial Analysis Jamie Barras,*,† Darragh Murnane,‡ Kaspar Althoefer,† Sulaf Assi,‡ Michael D. Rowe,† Iain J. F. Poplett,† Georgia Kyriakidou,† and John A. S. Smith† †

Department of Informatics, King’s College London, Strand, London WC2R 2LS, United Kingdom Department of Pharmacy, University of Hertfordshire, College Lane, Hatfield AL10 9AB, United Kingdom

‡

S Supporting Information *

ABSTRACT: We report the detection and analysis of a suspected counterfeit sample of the antimalarial medicine Metakelfin through developing nitrogen-14 nuclear quadrupole resonance (14N NQR) spectroscopy at a quantitative level. The sensitivity of quadrupolar parameters to the solid-state chemical environment of the molecule enables development of a technique capable of discrimination between the same pharmaceutical preparations made by different manufacturers. The 14N NQR signal returned by a tablet (or tablets) from a Metakelfin batch suspected to be counterfeit was compared with that acquired from a tablet(s) from a known-to-begenuine batch from the same named manufacturer. Metakelfin contains two active pharmaceutical ingredients, sulfalene and pyrimethamine, and NQR analysis revealed spectral differences for the sulfalene component indicative of differences in the processing history of the two batches. Furthermore, the NQR analysis provided quantitative information that the suspected counterfeit tablets contained only 43 ± 3%, as much sulfalene as the genuine Metakelfin tablets. Conversely, conventional nondestructive analysis by Fourier transform (FT)-Raman and FT-near infrared (NIR) spectroscopies only achieved differentiation between batches but no ascription. High performance liquid chromatography (HPLC)-UV analysis of the suspect tablets revealed a sulfalene content of 42 ± 2% of the labeled claim. The degree of agreement shows the promise of NQR as a means of the nondestructive identification and content-indicating first-stage analysis of counterfeit pharmaceuticals.

C

Reports of newly developed techniques for detection of counterfeiting of medicines have typically focused on high-value medicines used in developed economies such as sildenafil citrate for erectile dysfunction or the cholesterol-lowering atorvastatin calcium. However, globally, antimalarials appear to be one of the most targeted classes of medicines by counterfeiters.5 Metakelfin is a combination therapy consisting of sulfalene (also known as sulfametapyrazine or sulfamethoxypyrazine) and pyrimethamine in the same tablet (500 mg of sulfalene, 25 mg of pyrimethamine). Due to a rise in resistance and a spate of counterfeiting, it is no longer recommended as a treatment by the WHO. Indeed, its importation has been banned by Uganda and Tanzania, although genuine and counterfeit examples are still widely available, due to the continued popularity of this relatively low-cost treatment with consumers across Sub-Saharan Africa.6 Clearly, substitution of antimalarial API poses severe risks to patients, but substandard products also pose a risk of parasite tolerance and emergence of resistance when insufficient dose content or plasma concen-

ounterfeit medicines are a serious threat to human health, and the problem is a global one.1 A counterfeit drug is one which is deliberately and fraudulently mislabeled with respect to its identity and/or source. Counterfeit products may contain the correct ingredients, the wrong ingredients, no active ingredients, or insufficient active ingredient. Although the focus of the above-mentioned activity is on deliberately counterfeited medicines, there is an additional health risk from substandard medicines, medicines that do not contain the correct amount of active ingredient due to poor quality assurance at point of manufacture2,3 or degradation due to inappropriate storage in parallel supply chains rather than via any deliberate act. It is possible also for substandard medicines to possess altered bioavailability of the API (active pharmaceutical ingredient(s)) due to altered API solid state or excipient profiles in alternative production routes. Many techniques have been developed for detection of counterfeit medicines including visual, covert tagging, and chemical fingerprinting, which can be destructive (e.g., high performance liquid chromatography (HPLC) or nondestructive (e.g., Raman).4 There is an essential requirement for midsupply chain as well as point-of-dispensing analytical techniques to safeguard patients against substandard and counterfeit medicinal products. © 2013 American Chemical Society

Received: November 9, 2012 Accepted: February 5, 2013 Published: February 5, 2013 2746

dx.doi.org/10.1021/ac303267v | Anal. Chem. 2013, 85, 2746−2753

Analytical Chemistry

Article

known to be counterfeit; Pfizer was also able to supply pure samples of the active ingredients to support further analysis of the tablets.

trations of API are achieved. Therefore, analytical techniques which can reveal counterfeit substitution and possess the analytical power to indicate the standard of medicines are required for antimalarials. Several of the current authors have previously7 proposed that nuclear quadrupole resonance (NQR) spectroscopy may be applied to the problem of determining if a medicine is genuine and of the appropriate quality, so-called “medicines authentication”. More recently, we have described the first stages of a project to develop such a system.8 NQR groups in several countries working with polycrystalline samples of pharmaceutical compounds have been able to demonstrate that the method is quantitative, that different polymorphs of the same pharmaceutical compound can be distinguished, as can anhydrous materials from hydrates, and that there is no interference from the excipient or coating material.7−13 Thus, NQR, which produces spectral information and parameters which are comparatively easy to interpret, might offer improvements over alternative techniques such as Fourier transform-near infrared (FT-NIR) or FT-Raman spectroscopies, which can be sensitive to coatings, packaging, and fluorescence, and may require advanced high cost instrumentation14 and chemometric techniques.15 Approximately 50% of atoms in the periodic table have quadrupolar nuclei, so signals could be acquired from the vast majority of medicines, the only caveat being that, when the quadrupolar isotope is in low abundance, such as 33S and 17O, inspection times will be longer. The quadrupolar parameters are so sensitive an indicator of the chemical constitution of the solid-state environment of the molecule that the technique can discriminate between the same pharmaceutical preparations made by different manufacturers.12,13 The problems posed by spurious signals, such as those observed from piezoelectric materials included as excipient in the tablets or RFI, are largely eliminated by cycling the phases of the radiofrequency (RF) pulses and then using the novel parametric data processing and classification techniques that have been developed by KCL in association with Lund University, Sweden, to deal with these problems.16 The challenge for NQR in this application is in the low filling factors (ratio of sample volume to coil volume) that can be expected with blister packs, the standard configuration of packaging in most parts of the world; however, using RF methods, the contents of medicine bottles and blister packs can be examined through multiple layers of packaging without the need to remove the contents from cartons, plastic bags, drums, or other additional packaging of various sizes and shapes. We report here the comparative benefits of NQR for detection and analysis of a counterfeit antimalarial medicine labeled as manufactured by Pfizer in Italy (“suspect Metakelfin”). As an initial attempt to surmount the problems of low filling factors, a technique for studying loose tablets was developed in an approach that compared the NQR signal returned by a tablet(s) from a medicine pack suspected to be counterfeit with the signal acquired from an equivalent number of tablets from a known-to-be-genuine pack of the same medicine from the same named manufacturer. By sourcing “genuine Metakelfin” marketed by Pfizer in Italy for confirmation, false alarms from NQR signal differences caused by different production methods used in alternative legitimate supply chains were avoided. It should be noted here that the detection and analysis were carried out prior to contact with the named manufacturer, Pfizer. Once the analysis had been done and contact made, Pfizer was able to confirm that the tablets came from a batch

■

EXPERIMENTAL SECTION Test Materials. Metakelfin (500 mg of sulfalene/25 mg of pyrimethamine) tablets were sourced from a community pharmacy in Rome, Italy, in accordance with United States Pharmacopeia Drug Quality and Information Programme Guidelines. Suspect counterfeit samples were provided by Rockhopper TV following collection in Tanzania. Reference material of sulfalene (Lot# 06013-QCS) and pyrimethamine (Lot#08022-QCS) were generous gifts from Pfizer Ltd. (Sandwich, United Kingdom). Acetonitrile was HPLC-grade and was sourced from Fisher Scientific (Loughborough, UK); reverse osmosis water was produced in-house using an ElgaLabWater system (Marlow, United Kingdom). Sodium dihydrogenphosphate dihydrate (Rectapur grade) and sodium perchlorate (Analar grade) were from BDH Prolabo (VWR International Ltd., Lutterworth, United Kingdom). NQR Spectrometer. The NQR spectrometer setup used for this work is a Tecmag Libra 0.1−12.5 MHz Pulsed RF spectrometer with one transmit and one receive channel. The spectrometer is controlled via a MacIntosh PC. This spectrometer is used in conjunction with a Heatherlite 0.5− 5.5 MHz pulsed RF amplifier. The two RF resonators used for this work are described below. The solenoid NQR coil is built to take small pyrex test tubes of ca. 10 mm diameter (VWR Limited). A 40 mm long coil with 110 turns of 0.315 enamelled copper wire (RS Limited) was wound on a wax filled paper tube of slightly larger diameter (∼10.5 mm) than the sample tubes. The outside of the coil was embedded in epoxy resin to hold the wire in place, and the wax core and paper were removed. Thus, the coil has no former tube, and the inside surface of the coil is just the enamel coating of the copper wire, maximizing the coil filling factor. The inductance of this coil was measured as ∼26 μH and can be tuned and matched to 50 Ω over the frequency range of 2−4.5 MHz using the 5−120 pF variable capacitors that are standard within the NQR group. The Q of the NQR probe at 3.2 MHz was about 45, but this was damped to around 18 with a 10 Ω resistor for most of the work described here, as this increases the detection bandwidth and reduces spurious ring-down signals. The coil was supported concentrically in a length of “Perspex” tube (Farnell Limited), with the top and bottom glued in place, and it was mounted in a small aluminum diecast box which has a hole above the coil for inserting the sample. The unilateral NQR sensor is based around a 2.5 cm flat spiral RF coil. Since small coils have a low inductance (L), tuning to low frequencies requires a large capacitance; while this can be made up of fixed and variable capacitance in parallel to provide tuning adjustment, the range is very limited in practice. For this reason, any increase in inductance is useful; hence, a two layer spiral was used since the mutual inductance of the layers increases the total L.17 The RF coil was wound with 12 turns of 300 × 0.040 mm copper Litz wire (Scientific Wire Company) for each spiral layer, and the inductance was about 3 μH. The spiral coil was fixed inside a 5 mm slice of 32 mm OD acrylic tube. In order to shorten the dead-time following the end of the pulse, a Q-switch was employed. The Q-Switch is of a KCL design similar in basic concept to that detailed in ref 18. A 9 turn (about 4 mm deep) solenoid Q damping coil (L ≈ 5 μH) was wound using 0.4 mm diameter 2747

dx.doi.org/10.1021/ac303267v | Anal. Chem. 2013, 85, 2746−2753

Analytical Chemistry

Article

Figure 1. Standard normal variate-second derivative (SNV-D2) NIR absorption spectra of genuine (a) and counterfeit (b) Metakelfin tablets and the principal component analysis scores plot (c) of the SNV-D2 NIR spectra of the authentic (blue) and counterfeit (magenta and black) Metakelfin tablets. The 95% equal frequency ellipses are drawn around the authentic tablets scores.

enamelled copper wire (RS Limited) and fixed around the acrylic tube slice. FT-Raman and FT-NIR Analysis. The NIR spectra were recorded using an Identicheck FT-NIR spectrometer (PerkinElmer, Seer Green, United Kingdom) over the range of 4000−7500 cm−1. Twenty spectra were taken from each tablet from both sides. Each spectrum was the sum of 16 scans. The spectra were exported to Unscrambler 9.2, and standard normal variate-second derivative (SNV-D2) was applied as the pretreatment method. Principal component analysis (PCA) was applied by taking as many as 20 spectra from each batch of tablets. The first two PC scores (accounting for the highest variances) were plotted, and the 95% equal frequency ellipse was drawn around the authentic tablets scores. Raman spectra were recorded on the Renishaw inVia Raman microscope with 5× objective lens, equipped with 785 nm NIR laser, 300 mW laser power, and charge coupled device (CCD) detector. Twenty spectra were taken from both sides from each tablet over the range of 250−2000 cm−1. Each spectrum was the sum of four scans, and the raw spectra were exported to Unscrambler 9.2 where CWS and PCA were applied as above for FT-NIR analysis. Spectroscopic Sample Presentation. Loose tablets removed from the primary packaging were used for all spectroscopic measurements. Intact tablets were analyzed in the case of FT-Raman and FT-NIR measurements. For the unilateral coil NQR measurements, single tablets were placed directly above the center of the coil. For the solenoid, a combination of tablets broken in half and powdered tablets were packed into a Perspex test tube in order to maximize the filling factor of the coil. It was found that the small pyrex test tubes could hold 3.5 tablets in this way. An identical test tube was filled with an equivalent quantity of pure sulfalene powder for quantitative comparison measurements in the solenoid. Destructive Chemical Assay of Drug Content Using High Performance Liquid Chromatography. A gradient

HPLC method was employed on an Agilent 1100 instrument (Agilent Technologies UK Ltd., Wokingham, United Kingdom) to resolve sulfalene from pyrimethamine using a reverse phase Inertsil-ODS2 (250 mm × 4.6 mm, 5 μm particle size) from Capital HPLC (Broxburn, United Kingdom) at 25 °C. The mobile phase consisted of acetonitrile and an aqueous mixture, 40 mM sodium dihydrogenphosphate buffer (pH 3.5) with 10 mM sodium perchlorate (Buffer A). The mobile phase gradient at a flow rate of 1 mL min−1 was as follows: 20% acetonitrile increased to 35% over 10 min, held at 35% acetonitrile for 5 min, and reduced to 20% acetonitrile over 2 min to regenerate the column. Detection was at 270 nm, and quantification was against an external standard calibration curve in the range of 2.46−123 μg mL−1 (pyrimethamine) and 2.92−146 μg mL−1 (sulfalene) using the Pfizer-supplied quality standards. Tablet Analysis. Two “genuine” tablets were weighed and ground separately to a fine flowing powder using a glass mortar and porcelain pestle for 2 min. Approximately 0.07 g of powdered tablets was accurately weighed (in triplicate) into a 50 mL volumetric flask. The material was dissolved with the aid of sonication for 15 min in a mixture of 50/50 (% v/v) acetonitrile/buffer A. Samples (10 mL) were withdrawn from each solution and filtered using a 0.22 μm membrane filter (Millex GP PES membrane). The filtrate was transferred to 2 mL HPLC vials and sealed (for determination of pyrimethamine). One mL of each filtrate was diluted to 25 mL with dissolution solvent and transferred to a separate 2 mL HPLC vial (for determination of sulfalene). Two separate intact “suspect” tablets were treated as above. Four further samples were analyzed from the 3.5 tablets previously tested in the solenoid NQR experiments, the halved tablets being powdered, dissolved, and prepared in an identical fashion to the intact “genuine” tablets. 2748

dx.doi.org/10.1021/ac303267v | Anal. Chem. 2013, 85, 2746−2753

Analytical Chemistry

Article

Figure 2. Raman spectra of genuine (a) and counterfeit (b) Metakelfin tablets and the principal component analysis scores plot (c) of the Raman spectra of reference authentic Metakelfin tablets (red) and test authentic (green) and counterfeit (blue and magenta) Metakelfin tablets. The 95% equal frequency ellipses are plotted around the authentic tablets.

■

RESULTS AND DISCUSSION Organoleptic and Visual Analysis. The “suspect” Metakelfin tablets were labeled as being manufactured by Pfizer, Italy, in common with the “genuine” sourced batch. The “suspect” tablets presented with a mottled appearance and were more friable than the genuine material, suggesting poorer manufacture or altered formulation composition. Spectroscopic Analysis. FT-NIR. The average SNV-D2 spectra of a “genuine” and “suspect” Metakelfin tablet are presented in Figure 1 and were analyzed in accordance with Assi et al.19 Correlation in wavelength space (CWS) chemometric analysis of the SNV-D2 spectra (calculation of the correlation coefficient between the mean “genuine” tablet spectrum and a mean sample spectrum) returned a misclassification of the two “suspect” tablets as being genuine with spectral correlation values being above the 0.95 cutoff (“genuine” tablet 2: r2 = 0.9999; “suspect” tablet 1: r2 = 0.9673; “suspect” tablet 2: r2 = 0.9678). Previous work has shown that misclassification of products can be avoided by a strategy of principal component analysis (PCA) verification of the CWS results. The PCA with 95% equal frequency ellipse discriminated the “suspect” material clearly from the authentic tablets. However, it was not possible to determine the source of this difference, and it is important to bear in mind that a nonauthenticated sample containing the correct dose of API may pose less risk to a patient at the point of dispensing (e.g., counterfeiting by repackaging a generic substitute for a branded product). FT-Raman. The average spectra of a “genuine” and “suspect” Metakelfin tablet are presented in Figure 2, revealing the presence of active ingredient in both tablet batches. Raman spectroscopy offers benefits compared to NIR analysis, which include higher sensitivity to chemical identity and lack of interference from water content,19 which may be a major component of spectral variation from NIR analysis due to

different tablet storage prior to authentication. CWS analysis failed to classify the tablets, however, with correlation coefficients >0.98 for both “suspect” tablets and the “genuine” test tablet against the reference authentic tablet. PCA on the Raman spectra for the tablets (Figure 2) revealed clear differences between the tablets; however, the “genuine” tablets displayed Type I error. Thus FT-Raman analysis failed to offer conclusive authentication and did not adequately resolve the sources of spectral variation, which can arise due to, e.g., excipient profile, solid state (polymorph) employed, or relative concentration of the active ingredient(s).14,15,20 14 N NQR. As the sulfalene is present in a far larger amount than the pyrimethamine, it is the target of choice for the 14N NQR analysis. Sulfalene, C11H12N4O3S, is a sulphonamide antibacterial; crystals of sulfalene are orthorhombic in form, Pbca spacegroup, with z = 8.21 Three of the four nitrogens are involved in hydrogen bonding (Figure 3).

Figure 3. (Left) Sulfalene/sulfametopyrazine/sulfamethoxypyrazine; (right) pyrimethamine.

The full procedure by which an NQR “fingerprint” is generated for a previously unstudied active pharmaceutical ingredient (API) in a particular pharmaceutical preparation is described elsewhere.8 Here, attention was focused on the 2500−3700 kHz region of the 14N NQR spectrum, as work in Slovenia9 and in the UK7 has demonstrated that 14N lines in sulfa drugs are often to be found in this region. Using several of the genuine Metakelfin tablets and the solenoid coil described in the Experimental Section above, two 14N lines were quickly 2749

dx.doi.org/10.1021/ac303267v | Anal. Chem. 2013, 85, 2746−2753

Analytical Chemistry

Article

located close together at 3078 and 3062 kHz. Further lines were subsequently discovered at other frequencies. A parallel study of the 14N NQR lines in pure pyrimethamine (purchased from Sigma Ltd.) showed that none of these lines corresponded to lines in pyrimethamine and so were assigned to sulfalene. NQR signals were detected with a multiple-pulse pulse sequence of the so-called “Pulse-Spin-Locking” (PSL) type: P1±x − τ − (τ − P2y − τ)n where both the P1 and P2 are 90°equivalent-flip-angle RF pulses (119° for powder samples), with the phases indicated by x and y subscripts; τ was 700 μs and n = 64. Such RF sequences produce trains of refocused echoes (the overall envelope of which decays exponentially with the time constant T2e) which can be summed for signal enhancement. Site assignments were made through the use of “PolarisationEnhanced NQR” (PENQR).22 PENQR can in favorable cases produce large enhancements of 14N signals. The method works by transferring proton NMR polarization generated in a high static magnetic field (which need not be homogeneous) to the quadrupolar nuclei by the “level crossing” with the much lower NQR frequencies which occurs when the magnetic field is cycled to zero in a finite time. Critically, the degree of signal enhancement is dependent on the proximity of protons to the nitrogen and is generally greatest when there are one or more protons directly bonded to the nitrogen and the least, or negligible, when there are no protons direction bonded to the nitrogen. It is therefore ideally suited to assigning lines in materials where two or more of these conditions are met, as in sulfalene, where nitrogen is present as −NH2, −NH−, and  N−. PENQR of the lines identified in sulfalene showed that the 3079 kHz line belonged to one of the −NH2 groups, while the 3062 kHz line was a ring nitrogen. Having found suitable lines for the analysis, an equivalent number of the “suspect” Metakelfin tablets were run under identical conditions in a search for the same two 3078 and 3062 kHz lines. This search revealed that the “suspect” batch did have lines at these two frequencies, and thus, we would conclude that they do contain sulfalene. However, as Figure 4 shows, the signal obtained from the “suspect” tablets were markedly lower in intensity than those obtained from the “genuine” tablets. The weak feature near 3108 kHz is a quad image artifact of the 3062 kHz line. Note also the shoulder on the left of the lower frequency line of the genuine tablets; this is discussed later. There are a number of characteristic time constants associated with an NQR response, the relative values of which must be considered when trying to gauge why two samples return different responses, particularly when those two samples are meant to be the same [material]. These time constants are summarized in Table 1. In the vector model, the pulsed RF B1 field in the xy plane rotates the bulk system magnetization from its equilibrium position in the z direction into the xy plane. It is this magnetization that usually produces the detected signals. In NQR, strictly speaking, there is no bulk equilibrium magnetization since there are 2 opposing systems of magnetization along z and −z which cancel; however, following an RF pulse, they can both come into phase in the xy plane. It can also be seen from Figure 4 that the lines from the “suspect” Metakelfin tablets are noticeably narrower than those from the genuine tablets. This is suggestive of differences between the genuine and suspect tablets in the number of defects in the sulfalene crystals, the processing involved during tabletting, and/or the presence of paramagnetic impurities. As

Figure 4. 14N NQR spectra of genuine Metakelfin (red/larger peaks) and suspect Metakelfin (green/smaller peaks) in the region of 3035− 3135 kHz obtained under the same experimental conditions showing sulfalene peaks at 3062 and 3078 kHz. Excitation frequency, 385 kHz. 3.5 tablets of each inside the solenoid NQR coil. Note the shoulder on the left of the lower frequency line of the genuine tablets.

the latter would have an effect of the signal returned by the pulse sequence, complicating any attempt at a quantitative analysis, the spin−lattice, T1, and echo-train decay, T2e, relaxation times for the lines in both samples were measured. These are presented for comparison in Table 2. The spin− lattice relaxation times were estimated using a saturationrecovery approach and a single-exponential fit, the echo-train decay time by fitting a single exponential to overall envelope of the decaying echo train. Note, however, that measurements were made off-resonance and for comparison between the two samples only; they are not definitive. Within the margins of error, there were no appreciable differences between the estimations of these time constants between the genuine and suspect Metakelfin tablets. This excludes the possibility of any influence of signal intensity from paramagnetic impurities in one sample of the other and strongly suggests that the differences in intensity observed between the genuine and suspect Metakelfin is due to a difference in the amount of sulfalene present in the two samples. The unilateral sensor was used to examine single tablets from the genuine and suspect packs. Due to the over 3fold reduction in sample size, the returned signals were noisier; however, as Figure 5 shows, the results were broadly the same as those found with the multiple tablets inside the solenoid. Once again, the sequence employed was a multiple-pulse pulse sequence of the so-called “Pulse-Spin-Locking” (PSL) type. However, here, the excitation frequency was 3070 kHz. Table 3 shows the results of the comparison in returned signals. Two methods were used for extracting the intensity of the response, and the results were compared. In the frequency domain (obtained from the Fourier transform of the acquired echo signal), the integrated spectral line intensities were measured. In the time domain, the matrix pencil (MP) technique23 was used to extract the amplitudes of the individual components of the echo signal arising from each spectral line. The MP technique is a method for extracting the signal parameters of amplitude, decay time, frequency, and phase from 2750

dx.doi.org/10.1021/ac303267v | Anal. Chem. 2013, 85, 2746−2753

Analytical Chemistry

Article

Table 1. Time Constants of the NQR Response time constant T1 T2

description The longitudinal or spin−lattice relaxation time. The time taken for the system to return to equilibrium after an RF pulse. The actual material transverse or spin−spin relaxation time caused by homogeneous interactions in the material. In the absence of any defects or impurities etc, in the solid material, this is the time taken for the transverse or xy magnetization and hence the signal to decay after the RF pulse. In practice, this can only be measured from echo maxima in certain multipulse sequences. The observed transverse relaxation time when inhomogeneous line broadening or dephasing of the transverse magnetization occurs in solid materials due to defects and impurities etc. This is the exponential decay time constant for the FID signal following the RF pulse. An extended transverse relaxation time produced by certain multiple pulse sequences such as the PSL sequence which modify the relaxation process in the material. T2e is the exponential decay time of the train of echo signals produced by such sequences.

T2* T2e

Table 2. Comparison of Spin−Lattice (T1) and Echo-Train Decay (T2e) Relaxation Times for the 3062 and 3078 kHz 14 N NQR Lines of the Genuine and Suspect Metakelfin Tabletsa 3062 kHz line

a

Table 3. Comparison of Returned Signals Using Integrated Line Intensity and Echo Amplitude Estimated Using a Matrix Pencil (MP) Methoda sample

3078 kHz line

genuine Metakelfin 1 tablet

time constants

genuine

suspect

genuine

suspect

T1/s T2e/s

1.70 0.65

1.65 0.71

1.70 0.16

1.72 0.16

3078

3062 3078 3062 Integrated Line Intensity [×108] 1.64 1.95 0.69 0.86 3.59 (both lines integrated 1.55 (both lines integrated together) together) MP Echo Signal Amplitude [×105] 6.60 8.23 2.90 3.50

Note: measurements made off-resonance and for comparison only.

a

an exponentially decaying signal, which when multiple spectral lines are present is the sum of exponentially decaying signals with different parameters. An exponentially decaying complex signal is modeled as a sum over the various spectral components k as:

∑ ak e(β + i2πf )t k

k

One tablet of each placed on top of the NQR unilateral sensor.

Comparisons were made using data both from all lines integrated separately and the lines integrated together. The comparison of the single tablet produced an estimation of the ratio of sulfalene in the genuine Metakelfin tablets compared to the suspect batch of 2.32 ± 0.06 to 1, discounting measurements errors. This result suggests strongly that each suspect Metakelfin tablet is deficient in sulfalene to the same degree and to an order of 2.1 times lower than a genuine Metakelfin tablet. Thus, the conclusion of the quantitative NQR analysis is that the suspect tablets contain only 43 ± 3% as much sulfalene as the genuine tablets. As a check on the reliability of the quantitative data, a mass of pure sulfalene powder equialvent to 1.1× the mass of sulfalene in the genuine tablets was subjected to the same NQR measurements. Table 4 summarizes the time constants measured for all three samples, and Figure 6 shows the returned spectrum from the sulfalene powder. As T2e is a function of the relationship between T2* and 2τ (the time between the pulses in the echo train), data were acquired at several 2τ values to ensure that the results at any one given 2τ value were not influenced by the difference in T2* between the two lines. As can be seen from Figure 6 and as shown in Table 4, the lines in the powder are much narrower than in either of the samples in tablet form, which is due to compression effects or crystal damage during the manufacturing process. The spectra of the pure sulfalene material revealed that the shoulder on the 3062 kHz line of the genuine tablets (Figures 4 and 5) is, in fact, another line at 3067 kHz. The origin of this line is not known, but its presence in the raw material (a) explains its presence in the tablets and (b) discounts any effect of the tablet manufacturing process as the origin of this line. At the same time, the fact that the ratio of the 3067:3062 kHz lines remains constant through the genuine samples suggests that this is not a degradation product. The data was too noisy to tell if the shoulder was also present in the spectra of the suspect tablets;

Figure 5. 14N NQR spectra of genuine Metakelfin (red/larger peaks) and suspect Metakelfin (green/smaller peaks) in the region of 3010− 3120 kHz obtained under the same experimental conditions showing sulfalene peaks at 3062 and 3078 kHz. Excitation frequency, 3070 kHz. One pill of each placed on top of the center of the unilateral NQR sensor. Note the shoulder on the left of the lower frequency line.

S=

“suspect” Metakelfin 1 tablet Line (kHz)

k

(1)

where ak = |ak|e is the complex amplitude with θk being the phase; βk is the decay rate or damping term which is equal to 1/ T2k* where T2* is the decay time and f k is the frequency. The spectral line full width at half height (fwhh) of each component is given by 1/πT2k* = βk/π. (iθk)

2751

dx.doi.org/10.1021/ac303267v | Anal. Chem. 2013, 85, 2746−2753

Analytical Chemistry

Article

Table 4. Measured 14N NQR Parameters from All Three Samples Pfizer Sulfalene

genuine Metakelfin “suspect” Metakelfin

line [kHz]

T2* [μs]

T1 [s]

T2e [ms] 2τ = 0.42 ms

T2e [ms] 2τ = 1.4 ms

T2e [ms] 2τ = 2.68 ms

3078 3067 3062 3078 3062 3078 3062

140 330 267 68 90 93 136

1.2 1.1(5) 1.2 1.2 1.1 1.2 1.2

183

277

283

956 160 652 158 705

790 274 774 241 695

714 278 699 211 708

Genuine Tablets. The determined dose contents were 21.9 and 20.8 mg of pyrimethamine (or 86.4% and 81.9%) for Tablet 1 and Tablet 2, respectively. The corresponding doses of sulfalene were 491.3 and 488.7 mg (or 96.2% and 97.4%), respectively. Notwithstanding the lack of statistical significant sampling, the dose content was not outside the European Pharmacopoeial limits of not greater than two tablets lying outside the range of dose ± 15% and not more than one tablet outside the range of dose ± 25%. Suspect Tablets. Due to the greater quantity of the fake Metakelfin supplied, it was possible to perform a more statistically relevant analysis of six separate dosage units. The content of pyrimethamine was 25.6 ± 1.4 mg (i.e., % CV, 5.45%), and the content of sulfalene was 212.2 ± 8.9 mg (i.e., % CV, 4.23%). Although all samples were acceptable with respect to their content uniformity (i.e., within the ±5% limit), only the dose content of pyrimethamine was within tolerance for the stated content. Sulfalene, conversely, was present at 42.2 ± 1.8% of the stated content in the suspect Metakelfin batch.

Figure 6. 14N NQR spectrum of pure Pfizer sulfalene powder. PSL sequence 2τ = 2.68 ms. The sum of two measurements with all 256 echoes summed. Excitation at 3085 kHz; 6 s repeat time. Lines at 3078.8, 3066.9, and 3062.4 kHz. The weak feature near 3108 kHz is a quad image artifact of the 3062 kHz line. The shoulder seen on the spectra from the tablets is here resolved as a separate line at 3067 kHz.

■

CONCLUSIONS This study has been necessarily limited given that only one sample of the counterfeit batch of Metakelfin was available and limited funds precluded the purchase of large numbers of samples of the genuine material also. However, it can be said that a body of evidence confirmed differences between the genuine Metakelfin tablets and a batch suspected to be substandard or counterfeit in nature. FT-NIR and FT-Raman analysis, now routine detection tools in the fight against global counterfeiting, misclassified the tablets as being authentic in correlation of wavelength space. With a PCA-chemometric approach, spectral differences were observed, but neither Raman nor NIR-analyses were able to identify accurately the sources or causes of variation between formulations. 14N NQR analysis of the genuine and suspect Metakelfin tablets and the pure sulfalene powder lead us to conclude that the suspect tablets contained only 43 ± 3% as much sulfalene as the genuine tablets. The latter conclusion was verified by destructive HPLC determination, which also revealed the presence of the stated content of pyrimethamine in the suspect batch. Unlike other authentication techniques, we have shown that 14 N NQR spectroscopy supplied information confirmatory of substandard tablets (i.e., drug content) without the need for complex chemometric and external calibration approaches. There were additional differences between the returned signals from the genuine and suspect tablets in the width of the NQR lines that point to production differences between the tablets. This work has shown the strength of an NQR approach, which was suitable for rapid, single tablet authentication of suspected counterfeit Metakelfin tablets. Furthermore, the NQR approach provided information indicative of drug content and may be an

there appears to be a shoulder in the single tablet results (Figure 5), but this is less obvious in the 3.5 tablet results (Figure 4). The lack of resolution of the lines in the tablets is consistent with solid state and processing damage which is known to occur in common tablet manufacturing operations. A comparison of the line intensities between the powder and the genuine Metakelfin tablets returned an average ratio of four repeated measurements of 1.14 ± 0.06, consistent with the ratio of the masses (1.1:1), confirming the validity of the quantitative analysis of the different samples. Estimations of intensity were made by both treating the 3062 and 3067 kHz lines separately and taking them together. Destructive Chromatographic Analysis of Metakelfin Tablets. The chromatrogram of the mixed calibration standards (sulfalene and pyrimethamine) are presented in Figure S-1, Supporting Information, and demonstrated good resolution of the components at all concentration levels. The peak shapes were good with low levels of tailing with elution times (n = 19 standard injections) of 9.69 ± 0.02 min (sulfalene) and 13.11 ± 0.04 min (pyrimethamine). Calibration demonstrated excellent linearity for both drugs (r2 > 0.9999) with good precision (