Anal. Chem. 2009, 81, 5574–5576
Quantitative 35Cl Nuclear Quadrupole Resonance in Tablets of the Antidiabetic Medicine Diabinese Elizabeth Tate, Kaspar Althoefer, Jamie Barras, Michael D. Rowe, and John A. S. Smith* King’s College London, Department of Mechanical Engineering, Strand, London, WC2R 2LS, United Kingdom Gareth E. S. Pearce Merck, Sharp and Dohme, Hertford Road, Hoddesdon, Hertfordshire, EN11 9BU, United Kingdom Stephen A. C. Wren AstraZeneca, Silk Road Business Park, Charter Way, Macclesfield, Cheshire, SK10 2NA, United Kingdom Pulsed 35Cl nuclear quadrupole resonance (NQR) experiments have been performed on 250-mg tablets of the antidiabetic medicine Diabinese to establish the conditions needed for noninvasive quantitative analysis of the medicine in standard bottles. One important condition is the generation of a uniform radio-frequency (RF) field over the sample, which has been achieved by two designs of sample coil: one of variable pitch, and the other a resonator that has been fabricated from a single turn of copper sheet with a longitudinal gap bridged by tuning capacitors. The results from blind tests show that the number of tablets in a bottle could be predicted to within (3%. In recent years, several papers have discussed the potential applications of pulsed nuclear quadrupole resonance (NQR) to pharmaceutical analysis1,2 and drug development.3 The technique has several advantages: no static magnetic field is required, so the radio-frequency (RF) probe can be designed to accommodate the objects to be examined; and there is little limit to their volume, in that values as large as 8000 L have been used.4 Although NQR signals are only seen in solids, both crystalline and amorphous or glassy phases can be studied5 and the technique is noninvasive, so the container does not have to be opened. Furthermore, diluents and excipient do not interfere because of the high specificity of NQR spectra. Polymorphs are readily distinguished and their proportions estimated by means of parametric data processing of the time domain signals.6 At low RF values, * To whom correspondence should be addressed. E-mail: John.Smith@ kcl.ac.uk. (1) Balchin, E.; Malcolme-Lawes, D.; Poplett, I. J. F.; Smith, J. A. S.; Pearce, G. E. S.; Wren, S. A. C. Anal. Chem. 2005, 77, 3925–3930. (2) Perez, S. C.; Cerioni, L.; Wolfenson, A. E.; Faudone, S; Cufini, S. L. Int. J. Pharm. 2005, 298, 143–152. (3) Latosinska, J. N. Expert Opin. Drug Discovery 2007, 2 (2), 225–248. (4) Barras, J.; Gaskell, M. J.; Hunt, N.; Jenkinson, R. I.; Mann, K.; Peddar, D.; Shilstone, G. N.; Smith, J. A. S. Appl. Magn. Reson. 2004, 25, 411–437. (5) Taylor, P. C. Z. Naturforsch. 1996, A51, 603–611. (6) Butt, N. R.; Somasundaram, S. D.; Jakobsson, A.; Smith, J. A. S. Signal Process. 2008, 88, 834–843.
5574
Analytical Chemistry, Vol. 81, No. 13, July 1, 2009
developments in spectrometers using SQUIDS7 or atomic magnetometers8 offer the potential to provide higher sensitivity. The present technical note extends our earlier 35Cl NQR studies of the diuretic Furosemide1 by discussing the important features of an instrument based on commercially available pulsed RF spectrometers, which can be used in quantitative analysis, and presents new experimental results for 35Cl NQR in the antidiabetic drug Diabinese at 34.33 MHz. EXPERIMENTAL SECTION Two commercial pulsed magnetic resonance spectrometers were used in this worksan Apollo spectrometer, from Tecmag,9 and a Spin-NMR spectrometer, from Spincore;10 both use a Tomco Model BT00800, 800 W, 20-300 MHz RF power amplifier and two Miteq Model AU-2A-0150-1103 signal preamplifiers. For the Apollo spectrometer, we used the Tecmag NTNMR software to generate the multiple-pulse sequences, as well as to display and process the acquired NQR signals. For the Spincore spectrometer, the signal data were exported to our own software, which was developed using Matlab [The Mathworks, Inc., Natick, MA.]. This software enables standard data processing procedures to be performed via a graphic user interface (GUI), the features of which include baseline correction, fast Fourier transformation, shifting of data points, signal-to-noise ratio (SNR) calculations, spectral line-shape fitting, and exponential decay fits to time domain data. Sequences on the Spincore system are written in C++, and, to generate a multiple pulse sequence such as pulsed spin-locking (PSL), a one-dimensional batch file was used to start the sequence, as well as to set the frequency, pulse widths and separations, number of points, dwell time, time between scans, and their number. Two-dimensional batch files were run by inserting a series of commands into a one-dimensional batch file and saving the data at the appropriate point, to generate a series of two(7) Augustine, M. P.; Ton That, D. M.; Clarke, J. Solid State Magn. Reson. 1998, 11, 139–156. (8) Shaf, V.; Knappe, S.; Schwind, D. D. D.; Alem, O.; Romalis, M. V. Appl. Phys. Lett. 2006, 89, 214106. (9) Tecmag is located in Houston, TX. (10) Spincore is located in Gainsville, FL. 10.1021/ac900656e CCC: $40.75 2009 American Chemical Society Published on Web 06/03/2009
Figure 2. Variation in induction-loop voltage for a single-turn solenoid (STS) with a round cross section. Figure 1. Variation in induction-loop voltage along the length of a variable-pitch solenoid.
dimensional data files that could be processed and viewed using our software in the Matlab environment. In both spectrometers, the 90eff° pulse width was established from the Bessel function plot of signal intensity versus pulse length at constant RF power. The RF probe was designed to accommodate standard-sized medicine bottles and consisted of a coil parallel tuned, series matched using two Jennings variable vacuum capacitors contained within a die-cast aluminum box for screening. Signal preamplifier protection was provided by quarter-wave units optimized for frequencies in the range of 34-36 MHz. For quantitative work, it is important to generate a uniform RF B1 across the entire sample; for this purpose, two different coil designs were tested, both of which had similar Q-factors of ∼120. Coil A was constructed from 11 turns of 1.5-mmdiameter copper wire wound with variable pitch11 on a plastic former 50 mm in diameter. The field homogeneity was measured by means of a small induction loop (with an area that was approximately one-tenth that of the NQR coil), which was moved along the central axis of the solenoid over a distance of 20 mm from either side of its center. Figure 1 shows the observed variation in the detected voltage; the coil center is located at a distance of 2.25 cm, and the variation across 20 mm on either side of this was close to ±9%, largely because of errors in the winding. The sharp dropoff at distances greater than 20 mm from the center emphasizes the importance of locating the entire sample in the homogeneous region of the RF field. Coil B, which is a more novel design, was a single-turn resonator that was fabricated from a 0.7-mm-thick copper sheet, with a gap left between the ends of the turns, to give a longitudinal slot. Both square and round cross-sectional resonators were made, and their performance tested. The square-cross-section resonator had a length of 102 mm and a width of 46 mm (comprising a volume of ∼216 cm3), with a 7-mm longitudinal slot. The roundcross-section resonator was the same length and had a diameter of 57 mm (comprising a volume of ∼260 cm3). For resonators of this type, the inductance L is approximated by
L)
µ0A l
(1)
where A is the cross-sectional area and l is the length. Tuning capacitance was provided by bridging the 7-mm gap with 40 uniformly spaced 4.7 pF ceramic capacitors, and adjustment was provided by a variable vacuum capacitor. The square-cross-section single-turn resonator gave a variation of ±3% in the peak-to-peak voltage from the induction loop over a distance of ±20 mm from the center. The round-cross-section single-turn solenoid (STS) gave a better performance (see Figure 2), because the voltage variation over a distance of ±20 mm from the center was symmetrical and less than ±1%. This is consistent with the general rule that the field strength only decreases significantly beyond one radius from the center of the solenoid. The Q-factors of both round and square cross-section designs were comparable: both were ∼120. RESULTS AND DISCUSSION Several performance checks were undertaken with both spectrometers and coil designs. Operating Conditions. 35Cl NQR measurements were conducted at 34.33 MHz at room temperature on 18 250-mg tablets of the antidiabetic medicine Diabinese (chlorpropamide, phase A)2 that were contained in a standard medicine bottle with a diameter of 40 mm and volume of 75 cm3. A PSL sequence, 90eff°-τ - (90eff° - 2τ - )n was used in these experiments, with a 90eff° pulse width of 50 µs and τ ) 350 µs. The first four echo signals (n ) 4) of each PSL sequence were summed to enhance the signal intensity, and the sequence was repeated 5000 times for signal averaging with a total experimental time of 20 min. A typical echo signal is shown in Figure 3. Reproducibility. A set of 51 such echo signals was obtained for the Diabinese sample over a period of 17 h, and several different signal parameters were measured so that they could be examined for reproducibility. The measured signal parameters included the integrated signal intensity of the spectral line in the frequency domain following echo Fourier transformation of the time domain echo, the peak intensity, and the signal-to-noise ratio (SNR) of the spectral line. The signal parameters were measured (11) Leifer, M. C. J. Magn. Reson. 1993, A105, 1–6. (12) Adams, M. J. Chemometrics in Analytical Spectroscopy; RSC Analytical Spectroscopy Monographs; RSC: London, 1995; Chapter 6.
Analytical Chemistry, Vol. 81, No. 13, July 1, 2009
5575
Figure 3. Typical NQR echo signal (real component) from 18 250mg tablets of Diabinese at room temperature.
with and without baseline correction of the raw data. Small variations in the ambient temperature of ±5 °C, over the 17-h period of these measurements, had no significant effect on the results. The best reproducibility (±4% variation) was obtained for the integrated signal intensity of the spectral line. The corresponding measurements on the peak spectral line intensity gave a variation of ±5%. The SNR measurements gave a worse performance. Based on this study, the integrated signal intensity of the spectral line, after Fourier transformation of the echo signal, was used for subsequent quantitative work. Quantitative Experiments. These were designed to extend our previous measurements on the diuretic Furosemide;1 here, the spectrometer gain was set to give an acceptable signal for 4 tablets of Diabinese and not changed thereafter. The number of 250-mg tablets was varied over a range of 4-27 tablets, and the integrated spectral line signal intensity (I) is plotted against the number of tablets n in Figure 4. These measurements were repeated and a least-squares fit to a plot of the combined results gave the equation I ) 0.2553n + 0.0064
(2)
Perez et al.2 performed similar experiments on samples of pure chlorpropamide mixed with lactose, with an uncertainty of 5% in the mass of the active pharmaceutical ingredient (API); their calibration plot was not linear, but a much smaller mass of API was present. We used eq 2 for blind tests in which a medicine bottle was prepared with an unknown number of tablets. In these experi-
5576
Analytical Chemistry, Vol. 81, No. 13, July 1, 2009
Figure 4. Integrated signal intensity (I) of the spectral line after echo Fourier transformation of the 35Cl NQR echo signal plotted against the number of 250-mg Diabinese tablets (n).
ments, the spectrometer was first standardized against 10 and 20 tablets and the scans were performed three times. The integrated spectral line intensity (3.031) gave an estimated value of n ) 11.85 for the number of tablets with a confidence interval of ±0.35 tablets, an error of ±3% regarding the number of tablets,12 and a value close to the real value of 12. These results can be compared to our previous 35Cl NQR results1 at 34.2 MHz for the diuretic Furosemide, where the estimated value of n was 3.82 and the actual value was 4. CONCLUSIONS Standard pulsed radio-frequency (RF) spectrometers can be readily adapted to perform noninvasive quantitative nuclear quadrupole resonance (NQR) analyses of medicines in standardsized bottles, provided that care is taken with detection coil design to ensure a uniform B1 across the sample. In blind tests, the number of 250 mg tablets of the antidiabetic medicine Diabinese could be estimated with an uncertainty of ±3%. Further pharmaceutical applications of NQR spectroscopy, such as quality control, shelf life, and the identification of counterfeit materials, will be the subject of future papers. ACKNOWLEDGMENT The authors thank Merck Sharp and Dohme, AstraZeneca, and QRSciences for financial support of this project. Received for review March 30, 2009. Accepted May 12, 2009. AC900656E
