Raman Chemical Mapping of Low-Content Active Pharmaceutical

Article pubs.acs.org/ac

Raman Chemical Mapping of Low-Content Active Pharmaceutical Ingredient Formulations. III. Statistically Optimized Sampling and Detection of Polymorphic Forms in Tablets on Stability Slobodan Šašić* and Shawn Mehrens Pfizer, Worldwide Research and Development, Groton 06340, Connecticut, United States ABSTRACT: Several tablets of a formulation containing 1% w/w of the desired active pharmaceutical ingredient (API) form are spiked with minimal amounts of two different anhydrous polymorphs and an amorphous form. The amount of contaminant form was 2.5 to 10% of the total API concentration (0.025 to 0.1% w/w in the tablet), with five spiked tablets prepared. The presence of these contaminant particles are then identified using Raman microscopy/ mapping. The entire surface of each of these tablets is Raman-probed through a grid based on our previous proposal (Šašić, S.; Whitlock, M. Appl. Spectrosc. 2008, 62, 916) about the minimal number of spectra to acquire that would guarantee identification of the targeted component (taking into account the limit of detection). All three forms have been clearly identified in the Raman mapping spectra of prepared “calibration” tablets; particularly of note is the 2.5% spike (0.025% w/w in the tablet) of the relatively weakly scattering amorphous form. The same method is then applied to packaged tablets on stability and demonstrates that none of the previously analyzed contaminant forms is detected, hence building confidence that the desired API form does not change during stability testing.

Raman spectroscopy is a very suitable method for dealing with a variety of pharmaceutical products mostly due to the usually high Raman scattering coefficients of the active pharmaceutical ingredients (APIs) relative to excipients.1−6 Thus, even in formulations with low API loadings, strong Raman responses from APIs can be obtained in particular if Raman microscopes/ imaging instruments are used.7−15 In addition, sharp and welldefined Raman peaks also aid in distinguishing the API among other components of a formulation (certainly easier in comparison with the more popular near-infrared spectroscopy). Multivariate analysis of Raman spectra can further improve limits of detection and degree of recognition of the spectra. With regards to APIs, this means identifying not only the desired API molecule but also the possible presence of undesired entities including polymorphs, impurities, isomers, etc. Experience has shown that Raman spectroscopy is particularly useful for monitoring the API form during processing and formulation development and in the final drug product. This study addresses using Raman microscopy for determining the presence of multiple polymorphic or amorphous API forms in a finished drug product (tablet). The premise of the use of Raman microscopy for identifying unwanted forms is to collect spectra across the whole surface of a tablet in equidistant shifts of the microscope stage. If such a particle/domain is present which contains the undesired form, when hit with the excitation laser it should give rise to a Raman spectrum that is distinguishable from that of the desired crystalline form. However, the number of spectra that needs to be collected is usually arbitrarily set. In order to most optimally use the © 2011 American Chemical Society

instrument time and provide a statistically sound sampling (i.e., avoid over- or under-sampling), the sampling grid should not be randomly chosen. This topic is addressed in our previous study12 in which it is stipulated that the minimal number of spectra that should be acquired to statistically guarantee the appearance of at least one spectrum of the polymorph of interest is set as five times the multiple of the intuitively minimal number of spectra while taking into account the suspected level of concentration. For example, if the targeted concentration of a polymorph is about 0.5% of the total weight of the tablet, then intuitively it should suffice to collect 200 spectra across the entire surface of a tablet to detect at least one spectrum of that polymorph. According to the suggested sampling pattern, in order to statistically strengthen this evidence, 1000 spectra are to be acquired, as with this number of spectra the probability to observe the sought form is 99.3% as opposed to 63% in case 200 spectra are collected (Figure 1). All these considerations are subject to reasonable Raman scattering of the targeted polymorph. The present study builds upon the results from the previous one12 in describing the effectiveness of the Raman microscopy approach for analyzing a real example from pharmaceutical practice. Here, we deal with one particular solid dosage formulation and first assess whether the statistical sampling approach as described in Figure 1, as well as the Raman sensitivity of the API in that formulation, suffice to detect Received: October 3, 2011 Accepted: December 6, 2011 Published: December 22, 2011 1019

dx.doi.org/10.1021/ac202396u | Anal. Chem. 2012, 84, 1019−1025

Analytical Chemistry

Article

Figure 1. Probability of observing at least one spectrum of the targeted component with the concentration of 0.5%.

acquisition time per image, they are quite acceptable from an industrial perspective. The longest experiment here took little more than 2 days (mapping the 2.5% spike of the amorphous form) which is suitable for over the weekend runs and is definitely acceptable keeping in mind the applied degree of rigorousness. Chemometrics routines for reducing the noise in the spectra (intentionally ignored here) can be utilized for reducing the acquisition time. The various concentrations of the polymorphs in the tablets dictated how many spectra would be collected, which in turn determined the stage-shift steps (which are noted in the figures below). The sampling grid always spanned across the maximal imageable area on a tablet. The data analysis was carried out using ChemAnalyze (ChemImage, Pittsburgh, PA) and Matlab (Mathworks, Natick, MA) software packages. The former is used for obtaining the images, and the latter for refining them and plotting (all images shown herein are produced in Matlab). The mapping spectra were shortened so as to cover only the bands of interest around 1650 cm−1 or 1350 cm−1, baseline corrected with first order polynomials and then mean-normalized. Mean normalization is necessary for optimal analysis because there is a significant difference in the scattering ability of the crystalline and amorphous forms (but not so much among the various crystalline forms), leading to the amorphous signal to be barely perceivable when compared to the signal from the desired crystalline form. The normalization accounts for this difference in Raman scattering ability and renders the intensities of the spectra comparable although causing the amorphous spectra to be noisier than the crystalline. Although a number of chemical images have been produced and analyzed here and an imaging instrument was used for spectral acquisition, this, in essence, is not an imaging study as the distribution of the API particles is of no interest, only their identity matters. Essentially, Raman mapping was used in this study to collect a large number of Raman spectra, increasing the statistical odds of finding a pixel belonging to the undesired form. Using the hot pixels in the images is a convenient way to detect the API particles and identify their spectra but is not necessary. In principle, the same results could have been achieved by analyzing a very large matrix of Raman mapping spectra without resorting to producing images and looking into hot pixels. The manually made “calibration” samples contain 1% w/w total of API which itself is a mixture of the 10, 5, and 2.5% of

polymorphs in the series of tablets spiked with gradually diminishing concentration of the unwanted polymorph(s). Provided satisfactory results are obtained, the next step is applying the same procedure on the packaged tablets that are exposed to specific stability conditions (elevated temperature and humidity) to establish their polymorphic purity. When combined with our previous study,12 in which the key underlying principles are explained and experimentally confirmed, the methodology presented here is a recipe for rigorous and optimized Raman microscopy testing of polymorphs in a (commercial) pharmaceutical tablet. In this research, the API molecule of interest is a candidate in late stage development with an extremely complex polymorphic landscape. There are greater than 75 known forms of this molecule, the majority of which are solvate forms. The desired form A is the commercial form, and there are three other known anhydrous forms, two of which are relevant to development (forms B and C), as well as an amorphous form that has been seen during aggressive milling. Although thermodynamically there is no rational for a form change during stability, given the complex nature of this system, this research was undertaken to ensure that no changes were seen in the drug product during stability testing.

■

EXPERIMENTAL SECTION All tablets analyzed by Raman microscopy were affixed to specially designed sample holders with a small amount of epoxy resin adhesive. Sample dissection was performed using a specially designed ultrasonic tablet chisel to expose the interiors of tablets (remove the coating) and to produce smooth, flat surfaces for mapping. The cutting depth varied for each tablet but was typically ∼1200 μm. Raman spectra were collected using a Renishaw Invia Raman Microscope and WiRE software (Wotton-under-Edge, UK). Sample excitation was performed using a high-powered NIR diode laser (785 nm) providing approximately 50 mW of laser power at the sample. A 50× microscope objective was utilized for collection of all spectral data. A static Raman spectral window of 1162−1738 cm−1 was used in all measurements, with an acquisition time of 8 s and with 50% laser power, as using 100% power was found to occasionally burn tablets. The 8 s acquisition time is relatively long (regardless of reducing the power to 50%) but had to be employed because of the relatively weak signal of the amorphous form; if only forms B and C were analyzed, shorter acquisitions should suffice. In terms of total 1020



Article

Figure 2. Comparison of the Raman bands of the four existing forms of the API in the two regions of interest (pointed to by the arrows). Form A spectrum is shown in thick gray, the amorphous in black, form B in blue, and form C in red. The spectra illustrate that confining the analysis only to the bands marked by the arrows should be sufficient to distinguish the forms and produce chemical images in univariate fashion.

detected or hot pixels in the images. In the most simplistic terms, the outcome of the analysis can conceptually be defined as “yes” or “no” with regards to the presence of unwanted forms with the actual number of detected spectra of unwanted forms, if any, being of no particular interest. The spectra of the desired form A are compared with the spectra of the three undesired forms in Figure 2. The emphasis is mostly placed on the 1600−1700 cm−1 region where the strongest API peaks of interest reside. The bands are overlapping but still with large enough differences to be comfortably discerned without the use of any multivariate tools. One of the unwanted forms (C) overlaps quite heavily, however, with the form of interest (A), and its analysis is hence conducted in the 1300−1400 cm−1 region where the spectral differences allow for relatively simple discernment of these two forms. The key analytical outcome of this study can be presented by going directly to the 2.5% spiked tablet, but the description of all the cases matters for understanding the effectiveness and robustness of the proposed methodology. The first addressed experiment is hence that conducted with the 10% amorphous impurity present. Figure 3 illustrates the Raman chemical map of the 10% amorphous API obtained from the mean-normalized spectra at 1638 cm−1. The selected position is not the peak apex of the amorphous band but the Raman response of the amorphous form there is significantly stronger than that of form A. We have often witnessed cases of successful univariate imaging at nonoverlapped wavenumbers on the side of the peak of the band of the imaged component. The map features a prominent domain of API as well as a number of hot pixels that can be straightforwardly thresholded due to their sharp intensities. The spectra that correspond to those hot pixels are also shown in Figure 3 alongside the spectra from a number of randomly chosen “cold” pixels that only feature the spectra of the desired form A. Figure 3 clearly shows the difference between the spectra from the hot and the cold pixels, demonstrating thus that the amorphous form has been unambiguously detected. Figure 4 shows the spectra with 5% and 2.5% contaminations of the amorphous form. In both cases, there are a fair number of hot pixels (not shown in the figure) each of which are clearly associated with the spectrum of the amorphous form (Figure 4). We also noticed that as the concentration of the amorphous

amorphous form or 5% of forms B or C, respectively, with the rest being the desired form A. Only the amorphous form is tested at three levels of concentration due to its Raman signal being the weakest and hence most demanding to analyze. If the amorphous form is detected in all three concentrations, it is deemed likely that the forms B and C could also be detected. Overall, the concentrations of the unwanted form account for 0.1, 0.05, or 0.025% w/w in the respective tablets. In terms of the number of spectra to collect, this transfers from 20 000 spectra from an individual tablet for the 2.5% impurity spike to 5000 spectra for the 10% spike. The tablets were prepared using multiple blending, mixing, and sieving steps with iterative additions of excipients and API (both form A and the contaminant forms). The materials were sieved after mixing to ensure particle size uniformity between the excipients and the API (amorphous particle size was uncontrolled compared to form A). After all the components were blended, mixed, and sieved, the entire blend was roller compacted, and the resulting ribbons were milled with a comill. The blend was then compacted into round tablets using an Fpress. The initial experiments started with higher concentrations of contaminant form due to the significant acquisition times of the experiments; any problem during the acquisitions and in particular with determining the sensitivity is easier to detect in much less time-consuming maps of the 10% spiked tablet rather than in the four times longer experiments with the 2.5% spiked tablet.

■

RESULTS Calibration Tablets. The statistical sampling approach described in the previous study12 is applied here on a real tablet sample with the goal of determining the sensitivity of the Raman mapping instrument for detecting polymorphs or amorphous form in the inspected formulation, as well as to more rigorously test the proposed sampling scheme. The first part of this study can also be understood as a crude calibration in which a tentative limit of detection for a particular form is determined, under the optimized Raman mapping conditions. The tablets at various stages of stability conditions are tested afterward by comparison to these “calibration” tablets. It should be borne in mind here that the goal of the analysis is to merely detect the presence of an unwanted form and not to quantify it in any way through the number of spectra 1021



Article

Figure 3. Raman chemical map at 1638 cm−1 (reflecting the amorphous content) of the tablet spiked with 10% of the amorphous form. The hot pixels obtained after thresholding the image are associated with the Raman band of the amorphous form as demonstrated in the bottom part of this figure in which the spectra of the amorphous form are shown alongside the few spectra of the desired crystalline form A (thick gray traces). The distance between pixels is 50 μm; the images cover a 3550 × 3550 μm grid. The axes in this and all other images stand for pixels.

Figure 4. Similar to the preceding Figure 3 except that these spectra are obtained from the hot pixels in the images (not shown here) of the tablets spiked with 5% (a) and 2.5% (b) of the amorphous form. In both cases, the spectra of the amorphous form clearly differ from the crystalline form A spectra (thick gray). The distance between pixels in the images from which these spectra are obtained is 35 μm in (a) and 25 μm in (b).

form decreases and the number of collected spectra correspondingly increases so as to maintain the above cited theoretical probability of detecting the API, the number of hot pixels referring to the amorphous form actually somewhat increases in comparison to the 10% contamination. This fact supports the statistical methodology proposed here. The 2.5% spiked tablet is apparently the most demanding to analyze because of the low concentration of the amorphous form but in fact the outcome of its analysis is very comparable to the previous two. A number of hot pixels are detected (>50), and all of them represent the amorphous form, as the shown spectra illustrate (Figure 4b). Hence, the test with the 2.5% amorphous spike is quite successful, and as such, it can be considered the tentative LOD for the applied methodology. The same analysis of the amorphous form is replicated with the Raman maps of the tablets with 5% of unwanted polymorphs B and C. The 5% case is shown in Figure 5 for form B and Figure 6 for form C. The Raman map at 1638 cm−1 in Figure 5 reflects the appearance of the 5% form B spike. Form B can be identified at the same wavenumber as the amorphous form (but it is comparatively a much stronger Raman scatterer) so the imaging at 1638 cm−1 reflects appearance of either of these forms. This is a fortunate instance in the analysis because a lack of Raman signal at that wavenumber would indicate absence of these two forms. There are a number of hot pixels in Figure 5

which can easily be identified in the 1D presentation of the image. 1D presentations are found to be very useful in this application for gaining information on relative intensity differences between the hot pixels, the applicability of which will be demonstrated below. The spectra associated with the pixels from the thresholded image clearly demonstrate form B has been identified. Interestingly, in some cases, the spectra in Figure 5 prominently feature the shoulder at 1648 cm−1 due to form A which indicates the close vicinity of form A particles. Form C is identified through the peak at 1349 cm−1. Form A scatters at 1342 cm−1 so that, despite a significant overlap of the bands of these two forms, one still easily distinguishes form C. Figure 6 repeats the previous pattern and shows the spectra associated with the hot pixels in the thresholded image of the 5% form C spike. The shoulder/peak at 1342 cm−1 due to form A appears more notably here but, as in the previous cases, the 5% form C spike is unambiguously identified. The conclusion of the three depicted calibration runs is hence that all three unwanted polymorphs have been reliably detected through the optimized Raman mapping pattern. A tentative limit of detection of 2.5% w/w of the API content or 0.025% w/w in a tablet can be assigned on the basis of the ability to detect the amorphous form at these low levels. Stability Samples. Packaged tablets from several time points in the stability protocol are analyzed for the presence of the unwanted forms, starting with the initial t = 0 to 24 months 1022



Article

Figure 5. Top row: Univariate image at 1638 cm−1 of a tablet with a 5% spike of form B and its one-dimensional presentation, the purpose of which is to demonstrate the appearance of the spikes due to form B. Bottom row: The thresholded image and the spectra associated with the hot pixels. All the spectra demonstrate the peaks of form B at 1638 cm−1. The images cover a grid spanning 3550 × 3550 μm, with each pixel being 35 μm apart.

situation seen in the spiked examples in Figures 3−5, as there are no shoulders on the side of the key band at 1650 cm−1. The quality of the spectra is poorer in comparison to those seen previously for the simple reason that it is only the noise in these spectra that produces the hot pixels and not a band of an unwanted form. Figure 8 shows the spectra associated with the hot pixels from the Raman chemical image at 1348 cm−1 from the same stability tablet in the spectral region indicative of form C. The spectra are obviously different from those in Figure 6 that reveal the form C particles, and the lack of response at 1349 cm−1 indicates absence of form C. The overall comparison is less clear than in the cases involving form B and the amorphous form due to some interference of the Raman signal from the excipients. While Figure 6 features the Raman signal from the pixels referring to API forms A and C, Figure 8 shows signal from the pixels associated to the noisy spectra that do not arise from form C but have some spectral interference from the excipients in the formulation. The Raman scattering ability of the placebo is significantly lower than that of the API forms, but it is artificially amplified through the mean-normalization. Taking into account the calibration data above, it can hence be concluded that, strictly speaking, the inspected 6 months 30 °C/75% RH 1% loading API tablet contains at least 95% of the desired API form. The acceptance criteria for this testing is not more than 5% of any contaminant form being present in the tablet after stability, so all of the tablets tested “passed” the stability criteria for the presence of undesired polymorphic or amorphous forms.

Figure 6. Spectra associated with the hot pixels of the thresholded univariate image at 1349 cm−1 of a tablet with 5% spike of form C. All the spectra demonstrate the peaks of form C at 1349 cm−1 with some interference of the form A band at 1342 cm−1. The images (not shown) cover a 3550 × 3550 μm grid, with each pixel being spaced by 35 μm.

at 30 °C/75% RH. Only the results obtained from the packaged tablets exposed to 30 °C/75% RH for 6 months are shown here, as the results from the other tablets on stability are quite similar. Figure 7 shows the Raman chemical image at 1638 cm−1 from a 6 months at 30 °C/75% RH tablet. It needs to be compared with what is shown in Figures 3−6 for the calibration tablets. The image in Figure 7 contains a smaller number of hot pixels that are much harder to discern. The corresponding 1D presentation is considerably different from that seen in Figure 5 in that it shows a significantly reduced number of intensity spikes. The hot pixels from the thresholded image are tested on the presence of the unwanted form B and the amorphous form that both scatter at 1638 cm−1 where the image is produced. The associated spectra (Figure 7) evidently do not reflect the

■

DISCUSSION Our previous study that presented a statistical approach for selecting the number of spectra to collect focused on defining 1023



Article

Figure 7. Top row: Univariate image at 1638 cm−1 and its 1D representation of a 6 months at 30 °C/75% RH tablet. Bottom row: thresholded image and the spectra that correspond to the hot pixels. The spectra demonstrate the lack of signal at 1638 cm−1 which indicates the absence of form B and the amorphous form. The hot pixels appear due to the noisiness of the spectra and not Raman bands of the unwanted forms. These images cover 3550 × 3550 μm2, with each pixel being spaced by 35 μm.

thresholding images like those shown here typically results in ambiguities that lead to rather large deviations in results. Hence, we manually thresholded and limited the analysis merely at establishing that a satisfactory number of spectra have been detected in each of the calibration tablets. Additionally, as mentioned throughout this paper, the main focus of this study was only to detect the presence of a form and not to quantify it. Raman chemical imaging is not a method that lends itself to quantitative analysis due to numerous spectroscopic and imaging details that have to be addressed. One important consideration here is the homogeneity of the distributions of the API contaminant forms in the calibration tablets. Those tablets are manually prepared, only five in total, and without the use of the equipment normally used in the mass-production of the tablets. Although sieving was used in making these tablets, there is the possibility of particle size differences in the calibration tablets compared to the tablets made with large-scale manufacturing equipment. Commercial tablets usually contain API particles of controlled size and are unlikely to have agglomerates as that seen in Figure 3, which is especially important for low loading API formulations. Except for that case, however, we find that the API distribution in the calibration tablets reasonably well corresponds to what can be expected in a commercial product so that the inevitable differences in manufacturing of the tablets do not seem to significantly affect the viability of the applied method. The successful detection of such small amounts of the undesired forms can be assigned to the extraordinary Raman scattering ability of all the polymorphs of this particular API (with some exception for amorphous). Raman scattering from the contaminant particles seems to be excited with whatever small amount of the excitation light that reaches them. Another

Figure 8. Spectra that correspond to the hot pixels in the thresholded image at 1349 cm−1 of a 6 months 30 °C/75% RH tablet. There are no peaks at 1349 cm−1 that correlate to those seen in Figure 6 which would indicate the presence of form C. The noisy features detected at 1349 cm−1 in these spectra are due to the excipients.

and applying statistical concepts suitable for this subject and covered several tablets without systematically varying concentrations. The analysis of the calibration tablets here reveals that the significant reduction of the contaminant in the calibration tablets is being offset by the increased number of collected spectra so, overall, ample spectral evidence for all the forms is found. This confirms the soundness of the proposed approach for utilizing a specific sampling grid. The best estimate of how well is the reduction of the amount of an unwanted polymorph countered with increasing the number of collected spectra, or whether the number of the identified spectra of the amorphous form remains relatively constant for all three concentrations, would have been achieved by counting the spectra or the hot pixels in the image. This is not done here due to uncertainties with thresholding the pixels. Applying a common algorithm for 1024



Article

universally defined, so studies addressing that matter will have thresholds set arbitrarily.

illustration of this excellent Raman efficiency is in the signal of the predominant form A being detected to some extent in almost every single pixel in the image. Such a response hugely exceeds the nominal concentration of that form, which is only 1% w/w, and also points out to the ramification of Raman scattering of form A particles. The results with this formulation can probably be considered exceptional and not likely to be as impressive in working with other formulations. We expect the methodology presented here to certainly differ case-by-case depending on Raman scattering efficiency: the stronger the scatterer, the better the sensitivity and the lower the limit of detection. That said, the general principle of increasing the number of spectra collected offsetting a decrease in concentration should still hold true. With regards to this, the laser spot size may also play a role although we do not foresee it to be that influential given the experiments are conducted with a microscope on a solid sample which implies high power density on a very dense sample. In theory, the higher the magnification, the smaller the dissipation of the light and the sampling area which leads to the experiment better correlating with the sampling theory. Interestingly, if one is to produce a quantitative chemical image of form A in which the area coverage is to match its nominal concentration, one would have to massively suppress the signal of form A in the images by mathematical means which would, in effect, look like altering the experimental results. The threshold to be set in such a case could not be based on the difference between the signal and noise as seen in the 1D representations above but would have to be “imposed” by not allowing form A to exceed more than 1% of image pixels. Altogether, that would be a somewhat confounding case in which it is logical to limit form A to only 1% of the pixels (based on its w/w percentage), but the experimental data indicate form A in a much larger quantity so the experiment would need to be “corrected”. Such a situation is by no means uncommon in Raman chemical imaging of tablets, but the present case is a drastic illustration due to the outstanding Raman scattering coefficient of the API. It is said above that in principle this analysis does not consider the number of spectra of an unwanted form but merely their presence. Does it really come down to detecting a few polymorph spectra from a tablet and how meaningful (for practical applications) and robust could such a result be? In this case, if an unwanted form is really present in several percent (relative to the desired API form), according to the results of this study, it is unlikely that only a very small number of the spectra of that form will be detected. As the calibration tablets show, following the suggested number of spectra to collect and with a favorable scattering efficiency of the investigated molecule, a large enough number of spectra should emerge to unambiguously indicate that form. The meaningfulness of the limit of detection which dictates the calibration and the length of the experiment with the real sample are arbitrarily set before the whole experiment is run. On the basis of the results for this formulation, it is reasonable to expect detection of even a 1% polymorphic impurity or 0.01% w/w in a tablet. Ordinarily, it is assumed that there is only one API form in the formulation and that form is stable for an extended period of time. The experiments described here are only in place in cases where the API molecule is known to exist in a number of polymorphic forms of non-negligible probability of existing or being produced in time in the commercial tablets. However, the thresholds for polymorphic impurity of an API are not

■

SUMMARY Raman microscopy is successfully employed to detect contaminant forms in tablets with a total of 1% w/w API that, respectively, contained 10%, 5%, and 2.5% spikes of the amorphous form and 5% of two other polymorphs. These results show that, for the spectroscopically most demanding amorphous form, 0.025% w/w of it in a tablet has been reliably detected. A sampling grid previously proposed by us that optimizes the number of spectra to be taken from the surface of the tablet is employed in all the acquisitions. Its effectiveness is demonstrated in the relatively similar number of the spectra of unwanted polymorphs found in the tablets with significantly different concentrations of polymorph spikes; the reduction in the amount of spiking is countered with the increase in the number of acquired spectra, in line with the concept behind the applied sampling grid. The exact same approach is then applied to multiple packaged stability tablets. The analysis of these chemical images and associated spectra reveals that none of them resembles the spectra and images from the spiked tablets. This fact confirms the absence of the undesired forms and very high polymorphic purity of the final product. The excellent sensitivity of Raman microscopy to the API in this formulation is certainly due in large part to the exceptionally high Raman scattering coefficient of this particular API.

■ ■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected].

REFERENCES

(1) Šašić, S., Ed. Pharmaceutical Applications of Raman Spectroscopy; Wiley: Hoboken, NJ, 2007. (2) Pivonka, D. E., Chalmers, J. M., Griffiths, P. R., Eds. Applications of Vibrational Spectroscopy in Pharmaceutical Research and Environment; Wiley: Chichester, UK, 2007. (3) De Beer, T. R. M.; Baeyens, W. R. G.; Ouyang, J.; Vervaet, C.; Remon, J. P. Analyst 2006, 131, 1137. (4) Finni, G. J. Raman Spectrosc. 2004, 35, 335. (5) Shah, R. B.; Tawakkul, M. A.; Khan, M. A. J. Pharm. Sci. 2007, 96, 1356. (6) Hausman, D. S.; Cambron, R. T.; Sakr, A. Intl. J. Pharm. 2005, 80, 298. (7) Šašić, S. Pharm. Res. 2007, 24, 58. (8) Henson, M.; Zhang, L. Appl. Spectrosc. 2006, 60, 1247. (9) Bell, S. E. J.; Beattie, J. R.; McGarvey, J. J.; Laota Peters, K.; Sirimuthu, N. M. S.; Speers, S. J. J. Raman Spectrosc. 2004, 35, 409. (10) Bell, S. E. J.; Barrett, L. J.; Burns, D. T.; Dennis, A. C.; Speers, S. J. Analyst 2003, 128, 1331. (11) Kauffman, J. F.; Dellibovi, K.; Cunningham, C. R. J. Pharm. Biomed. Anal. 2007, 43, 39. (12) Šašić, S.; Whitlock, M. Appl. Spectrosc. 2008, 62, 916. (13) Aaltonen, J.; Gordon, K. C.; Strachan, C. J.; Rades, T. Int. J. Pharm. 2008, 364, 159. (14) Widjaja, E.; Kanaujia, P.; Lau, G.; Ng, W. K.; Garland, M.; Saal, C.; Hanefeld, A.; Fischbach, M.; Maio, M.; Tan, R. B. H. Eur. J. Pharm. Sci. 2011, 42, 45. (15) Widjaja, E.; Seah, R. K. H. J. Pharm. Biomed. Anal. 2008, 46, 274.

1025


Raman Chemical Mapping of Low-Content Active Pharmaceutical

Recommend Documents