Fast Non-Destructive Detection of Low Level Crystalline Forms in Amorphous Spray Dried Dispersion Using Transmission Raman Spectroscopy and Comparison to Solid-State NMR Spectroscopy

Abstract

The enhancement of oral bioavailability of poorly water-soluble drugs is an increasingly common challenge facing the pharmaceutical industry. This may sometimes be overcome by converting the stable crystalline form to a metastable amorphous form, usually stabilized by a polymer in an amorphous solid dispersion. However, amorphous forms are still thermodynamically driven to crystallize to the less soluble forms during processing or long-term storage. Sensitive quantitation of crystallinity is critical to ensure a stable and bioavailable amorphous drug product. There are many technologies known for solid-state characterization and each technique is suitable for a particular purpose. Here we compare various technologies to quantitate low levels of crystalline material in an amorphous spray-dried dispersion with a moderate 20% drug load, particularly transmission Raman spectroscopy (TRS), X-ray powder diffraction (XRPD) and solid-state NMR spectroscopy (ssNMR). TRS is being demonstrated here, for the first time, as a potential alternate to detect and quantitate the crystalline form of an active pharmaceutical ingredient (API) among the bulk of an amorphous solid dispersion. It is a rapid, non-destructive, automated technique, where the data is processed using a multivariate approach. We have shown that TRS has better sensitivity compared to traditional methods such as XRPD and comparable sensitivity to ssNMR; however TRS has the added benefits of faster acquisition time.

Introduction

Aqueous solubility and intestinal permeability are two parameters predicting intestinal absorption.1,2 A number of newly developed drugs belong to Biopharmaceutics Classification System (BCS) Class II or IV and exhibit poor or no solubility in aqueous environments resulting in poor bioavailability. Forming an amorphous phase, which can exhibit much higher dissolution rates in biological fluids compared with its crystalline counterpart, may enhance bioavailability.3,4 Amorphous drugs are typically stabilized by molecularly dispersing the API in a pharmaceutically acceptable polymer, at moderate drug loads of 10- 30%, which allows for the drug to be bioavailable for several hours prior to getting crystallized. Such solid dispersions can be prepared via a number of methods including spray drying, hot melt extrusion, or microprecipitation.5-13 Due to high melting point and poor solubility of drugs in organic solvents used in microprecipitation, spray drying was chosen for the current study to prepare the amorphous dispersion.

The drug in an amorphous dispersion is in a higher-energy state relative to a crystalline drug since enthalpy, entropy and free energy of an amorphous form are much higher compared to its crystalline counterpart.14,15 Due to higher energy, amorphous materials (drug as well as polymeric excipient) are known to absorb water when exposed to humid air, increasing the mobility.16,17 Higher mobility may facilitate nucleation and finally crystallization of the drug. In these products, higher potential for crystallization (upon manufacture and/or upon storage) will adversely affect bioavailability; thus, this is a critical quality attribute (CQA).

In such formulations it is highly desirable to be able to detect nondestructively residual crystalline forms, which have no or low bioactivity at low concentration levels. There are many techniques available for characterization of crystalline materials such as X-ray powder diffraction (XRPD),18,19 differential scanning calorimetry (DSC),20 infrared spectroscopy (IR),21,22 Raman spectroscopy,23,24 microscopy,25,26 and solid-state nuclear magnetic resonance spectroscopy (ssNMR).27-31 Each of these techniques is suitable for a particular purpose and can be used to supplement other techniques orthogonally. The sensitivity of the commonly used techniques is usually adequate for drug substances, but generally not for drug product due to the low levels of API and interference from excipient signals. DSC is a relatively simple analytical technique, which may be used to determine the homogeneity of the samples and heat flow rate difference between a sample and reference material. However, in this context the technique could not be used to quantify the amount of crystalline material as the glass transition temperature (Tg) for HPMCAS-L is ~120 °C, which is significantly below the melting point of the crystalline drug substance (273 °C). At elevated temperatures the HPMCAS-L will melt and degrade before the drug substance melts and may impact the data at higher temperatures. XRPD is generally considered as the “gold standard” for identifying and quantifying crystalline forms, however, this technique suffers from limited levels of detection (~5% w/w), orientation and statistical effects, and interference from both crystalline and amorphous excipients in formulated products.32 Detection and quantitation tasks become considerably more challenging when the API is diluted in excipients, as is the case in an amorphous solid dispersion and virtually all solid drug products. Additionally, the preparation of samples requires extensive, highenergy (mechanical grinding, micronization), which could facilitate nucleation of crystalline domains in the amorphous form.

Recently, Transmission Raman Spectroscopy (TRS) has been introduced for rapid, non-invasive and non-destructive quantification of pharmaceutical tablets or capsules constituents33,34 The technique offers several advantages including volumetric, non-destructive analysis with high chemical specificity.35-38 The analysis is currently performed typically at rates of seconds per sample although an advanced version of TRS has been recently demonstrated capable of reaching rates of tens of milliseconds paving the way for the use of this technique for in-line analysis.39 Raman spectroscopy is inherently sensitive to polymorphic and crystal structure changes and significant spectral differences can be observed in the low wavenumber region. By operating in transmission geometry the subsampling issue of (typically) back-scattering Raman microscopes is negated, yielding the potential for a truly volumetric technique for polymorph quantification in finished dosage forms.

In this study an hypromellose acetate succinate (HPMCAS)-based spray dried dispersion (SDD) has been prepared and various analytical techniques have been evaluated in regards to sensitivity for quantitation of low-level crystalline forms and for practicality as a routine tool for measuring this CQA. The drug load in the present study was set at 20%, thus the challenge was to quantitate relative amounts of crystalline and amorphous API when diluted in 80% polymer, a challenge regularly faced by pharmaceutical scientists today. To do this, a method needs sufficient specificity to distinguish API signal in the midst of significant excipient signal, since the total API represents only one-fifth of the total sample mass. Here we demonstrate TRS as a potential alternative for the first time as its ability to detect and quantitate the crystalline form of an active pharmaceutical ingredient (API) among the bulk of an amorphous solid dispersion, and compare results to those obtained via two different modes of quantitation using ssNMR. We also aim to evaluate strengths and weaknesses of TRS, particularly in terms of speed, practicality, and levels of detection in comparison to other methods, including X-ray powder diffraction and solid-state NMR spectroscopy.

Experimental

Instrumentation

Transmission Raman Spectrometer (TRS)

X-Ray Diffractometer

NMR Spectrometer.

Chemicals

20% API: HPMCAS SDD

API

HPMCAS-L (AQOAT AS-LG)

Dichloromethane and methanol

Samples

Amorphous API: polymer SDD granules: Drug substance and HPMCAS-L in a ratio of 20:80 (w/w) were dissolved in a DCM/ methanol mixture and spray dried using a mini spray dryer to yield the amorphous solid dispersion.

Crystalline API: polymer mixture: Crystalline drug substance and HPMCAS-L were mixed in a ratio of 20:80 using mortar and pestle. The homogeneous mixing was achieved by mixing the compounds threedimensionally using a scraper.

Crystalline admixtures preparation: Accurately weighed 20% crystalline API: HPMCAS mixture was spiked in 20% amorphous API: HPMCAS SDD at various concentration levels. The proportion of crystalline API ranged from 0.2% to 20% (w/w) relative to overall sample mass. Table 1 lists the details about the various sample preparations and percent of crystalline amount per sample.

Table 1. Sample Preparation: Admixtures crystalline (%w/w)

Methods

Transmission Raman Spectroscopy (TRS)

The system uses a fiber-coupled 830 nm laser, delivering a 650 mW, 4 mm diameter beam to the sample via zoom optics. Collection optics on the opposite side of the sample to the incident laser captures energy transmitted through the sample. Intense Rayleigh (elastically-scattered) radiation is filtered using a pair of long-pass filters. Raman radiation (inelastically-scattered) is delivered by fiber to a high efficiency spectrometer and converted to a Raman spectrum by the TRS’ control and analysis software. Glass vials were chosen as the containers for the powder samples as they allowed a larger volume of sample (~35 mg at ~1 cm deep) to be analyzed in a single measurement. Glass vials were placed in the system’s automated sample tray for analysis. Acquisition parameters were adjusted for each vial. Samples 1 to 3 were scanned for 21 seconds (30 accumulations x 0.7s) and samples 4 to 10 for 30 seconds (30 accumulations x 1s). Each sample was then rescanned five times, vials were shaken and tapped to settle between scans. Multivariate modeling was performed using a data discovery and modeling software. Due to the large laser illumination area and relatively low incident laser power, insufficient energy was supplied to modify the crystalline structure or induce crystalline transformation

X-Ray Powder Diffraction (XRPD)

X-ray Powder diffraction (XRPD) patterns were obtained using a powder X-ray diffractometer, with Ni filtered Cu Kα radiation. The tube voltage and current were 30 kV and 15 mA, respectively. Each sample was placed onto a zero background silicon sample holder and flattened prior to measurement. Samples were rotated during acquisition and diffraction was measured from 2-40° 2θ with a scan rate of 1°/min and a step size of 0.02°. Data analysis software was used.

Solid-State Nuclear Magnetic Resonance (ssNMR)

Solid-state NMR spectra were acquired using a spectrometer operating at 500.13 MHz for 1H and 125.77 MHz for 13C. Magic-angle spinning (MAS) was performed at 8 kHz for all experiments. Samples were packed as powders into 4 mm zirconia rotors and sealed with Kel-F drive caps. Pulse sequences for 13C spectra employed ramped cross polarization (CP),40-42 5-π total sideband suppression (TOSS)43,44 with a 243-step phase cycle, and high power 1H decoupling with a SPINAL64 scheme and field strength of 89 kHz.45 The 1H 90° pulse width was 2.8 μs and the TOSS sequence employed 13C 180° pulses of 6.5 μs. Experiments for the calibration curves utilized a CP contact time of 4 ms, a recycle delay of 2.5 s, and a total of 31104 scans were collected for each spectrum (21.6 h experiment time). Variable contact time quantitation experiments utilized a recycle delay of 8 s (>5x 1H T1 of crystalline component) and a 4-point contact time array of 4, 6, 8, and 10 ms, with 5120 transients acquired for each slice (45.5 h total experiment time). 1H T1 values were measured using a 13C-detected CP-based saturation recovery experiment with TOSS, using a 16-point variable delay array from 0.05-20 s, and 32-128 scans per slice. All data were collected at 293 K. 13C chemical shifts were externally referenced to tetramethylsilane by setting the methyl peak of 3-methylglutaric acid to 18.84 ppm.46 Deconvolutions for carbon 6 (an aliphatic quaternary carbon) fitted peaks at 46.3 ppm (amorphous) and 45.3 ppm (crystalline), carbon 10 (an aliphatic methylene carbon) fitted peaks at 49.9 ppm (crystalline) and 49.4 ppm (amorphous), and carbon 13 (a heteroaromatic carbon) fitted peaks at 162.9 ppm (crystalline) and 161.5 ppm (amorphous). Peaks for carbons 6 and 10 were fitted together in a single four-site deconvolution due to overlap among these peaks. 1000 iterations were performed to give the best fit for chemical shift, intensity, and line width, while Gaussian/Lorentzian ratios were fixed at 0.1 for crystalline peaks and 0.9 for amorphous peaks. Data were analyzed and calibration curves were plotted with Microsoft Excel.

Results and Discussion

Transmission Raman Spectroscopy (TRS)

TRS is an automated, non-destructive, non-invasive technique, which measures Raman signal on the opposite side of the sample to laser illumination region. The diffusedly scattered photons have a long effective path-length through the sample; hence, the method is well suited for the bulk analysis of material. The spectral range of the TRS system used is ~50 cm-1 to 2500 cm-1. The extension into the lowwavenumber “phonon region” (<500 cm-1) provides an advantage for structural and polymorph analysis since the intermolecular modes of the lattice are probed. Transmission Raman spectra were collected for SDD amorphous drug (20:80 API: HPMCAS-L), crystalline drug Form A and HPMCAS-L (polymer excipient) to evaluate selectivity. Figure 1 shows the representative spectra; significant differences were observed across the entire spectral range. Peaks for the crystalline material appear sharper and well defined in comparison to the amorphous drug mixture. Specifically peaks ~400 cm-1, ~800 cm-1 as well as ~1000 cm-1 to 1600 cm-1 for crystalline vs. amorphous drug can be readily identified. Moderate auto-fluorescence contributions from both the excipient and API is apparent, represented by the gently sloping background signal. Background of this type is typical in pharmaceutical drug product and is treated by straightforward spectral processing algorithms.

 Figure 1. Overlay of representative raw TRS spectra of Amorphous SDD, Crystalline drug and polymeric excipient

Figure 2 shows overlay of average TRS spectra of each sample, subjected to baseline correction and normalization to assist visualization of spectral changes. From visual inspection of the data, differences in the spectra can be observed starting from sample no. 2 with 1% relative crystallinity (0.2% w/w absolute crystallinity). Peaks associated with crystalline API such as peaks at ~255 cm-1, 390 cm-1 and 789 cm-1 become sharper and more intense as the %w/w of relative crystalline material increases from 1% to 100%. Additionally, the peaks between 1000 cm-1 and 1633 cm-1 resolve and start to appear while the peak shoulder at 1581 cm-1 disappears. The crystalline weight % values are with respect to only the API portion of the samples, which represents only 20% of the total sample mass.

 Figure 2. Overlay of Normalized TRS Spectra: 20:80 (w/w) API: HPMCAS samples with varying amounts of crystalline API spiked in. Significant peaks, which change with increase in %w/w crystallinity are marked with*

Raw spectra were analyzed via multivariate Partial Least Squares (PLS) modeling using data analysis software. Optimization of model parameters lead to the choice of spectral pre-processing across the wavelength range 1000 cm-1 and 1700 cm-1. The baseline was subtracted using an automatic Whittaker filter, and the spectra were normalized and mean centered. Sample no.10 (100% w/w crystalline: amorphous) was excluded from the model based on model statistics. The absence of amorphous material and the large jump between 42.9% and 100% w/w crystalline: amorphous may explain why the model was unable to fit a satisfactory trend.

PLS has been able to determine a linear fit with a high correlation coefficient (R2 = 0.991) from 0% to 42.9% w/w relative crystallinity in the presence of HPMCAS-L polymer excipient (0 to 8.6% w/w absolute crystallinity) as shown in Figure 3.

 Figure 3. PLS model showing linearity: 0% - 42.9% (w/w) crystalline: amorphous (0 to 8.6% w/w of overall sample mass)

Limit of detection (LOD) for this data set can be calculated following standard procedure (3.3 x StDev of residual error)/slope of the regression47 to give a value of 0.9 % absolute crystallinity. This fits extremely well with the visual inspection of the spectra where changes are observed between 0.2 and 1% absolute crystallinity.

X-Ray Powder Diffraction (XRPD)

Each sample was also analyzed using XRPD, and resulting diffraction patterns are shown in Figure 4. A reference pattern for crystalline API Form A is shown at the top of the overlay, followed by 20:80 API: HPMCAS dispersions with varying amounts of crystalline Form A spiked in. At the bottom of the overlay is the pattern of fully amorphous API: HPMCAS solid dispersion, with a characteristic broad amorphous halo. In the 9.3% crystalline sample, the most intense crystalline Form A peaks at 15.2° and 17.3° are barely visible against the noise, indicating that the detection limit of crystallinity in the dispersions falls at about 10% crystalline API, which equates to ~2% of the overall sample mass (absolute crystallinity) due to the 20% drug load. However, in some cases residual crystallinity is severely detrimental to dissolution, and can act as seed material to accelerate crystallization in the dispersion, or to aid precipitation in the gut following oral delivery. Only a small amount of material is exposed to radiation due to relatively low penetration of X-ray beam into the sample. Also, elevation of the background in the range of 15 to 25o 2θ for the amorphous SDD (“halo”) is not constant and intensities may change impacting the S/N ratio for the peak corresponding to crystalline material. Due to variations in relative peak heights in the XRPD experiments, quantitation of percent crystalline was not attempted for these samples and LOD was estimated visually directly from the diffraction patterns. XRPD is generally used as a limit test due to complexity in developing and validating a quantitative method.

 Figure 4. Overlay of XRPD patterns of 20:80 (w/w) API:HPMCAS samples with varying amounts of crystalline API spiked in. Pure crystalline API Form A is shown for reference at the top of the overlay

Solid-State Nuclear Magnetic Resonance (ssNMR)

While NMR is inherently a quantitative technique, experiments performed using cross polarization (CP) are not strictly quantitative, in that one cannot simply integrate peaks due to different phases and use peak areas without consideration of a number of factors. Both relaxation dynamics and CP dynamics of each component of the sample must be considered for accurate quantitation results, and quantitation can be accomplished multiple ways. Here we demonstrate two methods of quantitative solid-state NMR, including variable contact time experiments, which can give accurate quantitation results without the need for known standards, as well as a more traditional standard curve preparation method.

Figure 5 shows the 13C CP-TOSS ssNMR spectra of HPMCAS-L (5a), amorphous API (5b), crystalline API (5g), and representative spectra of the 20:80 solid dispersions with varying amounts of crystalline API (5c-5f ). Crystalline API may be easily distinguished from amorphous material based on line width. Highly ordered materials exhibit sharp, Lorentzian peaks, as seen in the spectrum of crystalline API (Figure 5g). Amorphous materials result in very broad, Gaussian peaks, as seen in the spectra for HPMCAS-L (5a), amorphous API (5b), and the amorphous solid dispersion of the two (5c). As can be observed in Figure 5, selectivity in ssNMR is excellent, as most of the API peaks are unobstructed by HPMCAS peaks, in particular the aromatic and carbonyl peaks from 165-110 ppm, and aliphatic carbon peaks from 55-35 ppm. However, with respect to crystalline and amorphous API, all peaks are at least partially overlapping, as is usually the case for crystalline/amorphous mixtures due to the broad peaks of the amorphous phase. This simply means that peaks for individual carbon nuclei must be deconvoluted to obtain true peak areas for each phase. It is also evident in Figure 5 that 1% crystalline content (w/w of total sample mass) is clearly detectable (Figure 5d), manifested as small, narrow spikes rising above the broad peaks of the amorphous component.

 Figure 5. 13C solid-state NMR spectra of a) HPMCAS-L, b) amorphous API, c) spray dried dispersion containing 20% API: 80% HPMCAS-L, d) 20:80 solid dispersion with 1% crystalline API (w/w% of overall sample mass), e) 20:80 solid dispersion with 3.3% crystalline API (w/w% of overall sample mass), f) 20:80 physical mixture of 20% crystalline API and 80% HPMCAS-L, g) crystalline API. All samples employed 8 kHz MAS and were collected at 293 K.

Three carbons were selected for quantitative evaluation, including the peaks ranging from 165-160 ppm and 52-45 ppm. These peaks were chosen for optimal balance of sensitivity and ease of peak deconvolution. In theory, any peak is usable for quantitation, and the peak selected is a matter of choice of the analyst. In practice, only one carbon peak from each phase is necessary for quantitative analysis, though the peak representing the same carbon must be used for each phase. For standard curve preparation, samples were run with a non-quantitative recycle delay time to achieve good signal-to-noise in reasonable amounts of time (21.6 h), and an optimal contact time of 4 ms was used. The 1H T1 for the amorphous phase was found to be 0.84 s, while the crystalline phase was 1.44 s, and a recycle delay of 2.5 s was utilized. Peaks for each phase were deconvoluted, and the ratio of crystalline: amorphous peak area was plotted against theoretical weight % of the crystalline component. Expansions of the peaks used for quantitation are shown in Figure 6, with percentage of crystalline API indicated with each spectrum. These numbers correspond to just the relative amount of crystalline API, which itself comprises only 20% of the overall sample mass. As seen in Figure 6, 1.0% crystalline (0.2% absolute crystallinity) is essentially indistinguishable from the fully amorphous sample, however 4.9% crystalline (1.0% absolute crystallinity) clearly shows the presence of crystalline API, with peaks at 162.9 ppm (C13), 49.9 ppm (C10), and 45.3 ppm (C6). The signal-to-noise ratio of C6 in the sample containing 4.9% crystalline API (1.0% absolute crystallinity) in the data set collected here was ~10.5, thus the limit of detection is estimated to be 0.3-0.5% crystalline relative to the overall sample mass (absolute crystallinity). This is in line with the observation that 0.2% crystalline was not detectable in these experiments.

 Figure 6. 13C solid-state NMR spectra of 20:80 API: HPMCAS solid dispersions containing various amounts of crystalline API. Spectra are expanded from 167-153 ppm and 55-35 ppm to show the API peaks of interest, and the regions selected for peak deconvolution and quantitation calculations are is shaded in gray. The % crystalline values shown are the relative amounts of crystalline to amorphous API, while the total amount of API remained constant at 20% in each sample. All samples employed 8 kHz MAS and were collected at 293 K.

Calibration plots of crystalline:amorphous peak area ratios versus crystalline weight % are shown in Figure 7. The crystalline weight % values are with respect to only the API portion of the samples, which represents only 20% of the total sample mass. As can be seen in Figure 7, excellent results were obtained regardless of the carbon peak chosen for quantitation. Carbon 6 (48-45 ppm) yielded the best results and is the recommended peak to use for absolute quantitation in these formulations. The correlation, shown in Figure 7c, was highly linear form 1.0-33.3%, giving an R2 value of 0.994. Carbon 10 (Figure 7b) also gave good results, with a linear fit from 1.0-33.3% and an R2 value of 0.976. Aromatic carbon 13 (Figure 7a) gave the second best linear fit, with an R2 of 0.985.

 Figure 7. Calibration curves for ssNMR quantitation of crystalline API in an amorphous solid dispersion containing 20% API and 80% HPMCAS-L. Peaks representing atoms a) C13 (165-160 ppm), b) C10 (52-48 ppm), and c) C6 (48-45 ppm) in respective crystalline and amorphous phases were deconvoluted to obtain peak areas and peak area ratios. Crystalline-to-amorphous peak area ratios are plotted against weighed % crystalline values (% of API co

The generally accepted method for quantitation in cross polarization experiments without the need for standard samples is via a variable contact time experiment.29 In these experiments, spectra are collected at varied CP contact times, and must be collected with a quantitative recycle delay, which is >5 times the 1H T1 of the slowest relaxing component of the sample. Since crystalline API has a longer T1 relaxation time (1.44 s) than amorphous API (0.84 s), the recycle delay chosen for these experiments was 8 s. Contact times chosen were 4, 6, 8, and 10 ms. Relatively long contact times are typically required, as quantitation relies on measuring peak areas at each contact time point, then extrapolating back to a theoretical contact time of 0 ms to obtain the natural peak area in the absence of CP dynamics (TIS and T). These experiments can be time consuming (45.5 h here), but also invaluable as no other technique has the ability to quantitate relative amounts of different phases in a sample in the absence of pure phase standards. Table 2 shows the results of variable contact times experiments performed on four of the API:HPMCAS solid dispersions with 1.0, 1.9, 3.3, and 6.7% crystalline API spiked in (overall sample w/w%). With respect to only the API component (20% drug load), these values correspond to 4.9, 9.3, 16.4, and 33.3% crystalline API, respectively. Carbon 6 was used for peak deconvolution and area calculations in these experiments. Once again, these quantitation experiments are self-sufficient and do not require known standards for quantitation

Table 2. Quantitation results from ssNMR variable contact time experiments

Conclusion

DSC has some usefulness in describing differences between samples, however, it has been shown here to be impractical for the purpose of investigating the proportion of crystalline API in a matrix of amorphous material. XRPD is a widely used analytical technique in this application; it is well understood and relatively low cost to implement. Since the signal intensity correlates with crystalline content, the data suffer from very poor signal to noise and must compete with the amorphous ‘halo’. Crystalline features in the spectra are apparent at around 2% w/w crystalline API (absolute crystallinity). Furthermore, XRPD requires significant sample preparation and cannot be considered a true volumetric method when the penetration depth of the x-ray beam is so shallow. There is also a potential of converting amorphous API to crystalline during sample preparation.

Both ssNMR and TRS share the similarity in that loss of crystalline structure is reflected in a broadening and weakening of spectral features. As such, both show spectral changes between samples 2 and 3 (1% and 4.9% relative crystallinity or 0.2 and 1 % absolute crystallinity). With high signal to noise spectra, these spectral differences may be apparent at half the concentration of crystalline material, i.e. 0.5% w/w (absolute). Indeed, the 0.2% w/w (absolute) spectra are indistinguishable from the 0% crystalline API spectra both in ssNMR and TRS data. Table 3 compares different technologies evaluated for the analysis of crystalline amount in amorphous solid dispersion containing 20% drug load.

Table 3. Comparison of Various Technologies

TRS has been shown to be selective, linear and sensitive to measuring crystalline API in an SDD of amorphous API in HPMCAS-L via PLS analysis with and R2 of 0.991 and a LOD estimated at 0.9% absolute crystallinity. This is by no means an extensive study, as further work would include validation samples and repeatability measurements to test the robustness for this specific application. A secondary analysis technique could be used to validate the system for future measurements. Despite this, this study has demonstrated that TRS is a practical alternative to the other methods described here; nondestructive and fast speed of measurement. Additionally, TRS has the potential to be used as an at-line in-process check (IPC) test during drug product process development and manufacturing.

Acknowledgements

Authors would like to thank Hong Lin for helpful discussions, Dr. Nik Chetwyn for reviewing the manuscript and Genentech Inc. for financial support.

References

  1. A. Dahan, J. Miller and G. Amidon, AAPS J, 11 (2009) 740-746. (DOI: 10.1208/s12248-009- 9144-x)
  2. G. Amidon, H. Lennernäs, V. Shah and J. Crison, Pharmaceutical Research, 12 (1995) 413-420. (DOI: 10.1023/a:1016212804288)
  3. L. Yu, Advanced Drug Delivery Reviews, 48 (2001) 27-42. (DOI: 10.1016/s0169- 409x(01)00098-9)
  4. B. C. Hancock and G. Zografi, J. Pharm. Sci., 86 (1997) 1-12. (DOI: 10.1021/js9601896)
  5. J. Baumann, D. Dobry and R. Ray, Drug Development & Delivery, 13 (2013) 32-37.
  6. P.-C. Chiang, Y. Cui, Y. Ran, J. Lubach, K.-J. Chou, L. Bao, W. Jia, H. La, J. Hau, A. Sambrone, A. Qin, Y. Deng and H. Wong, AAPS Journal, 15 (2013) 608-617. (DOI: 10.1208/s12248-013- 9469-3)
  7. Q. Hu, D. S. Choi, H. Chokshi, N. Shah and H. Sandhu, International Journal of Pharmaceutics (Amsterdam, Netherlands), 450 (2013) 53-62. (DOI: 10.1016/j.ijpharm.2013.04.040)
  8. R. Iyer, S. Hegde, Y.-E. Zhang, J. Dinunzio, D. Singhal, A. Malick and G. Amidon, J. Pharm. Sci., 102 (2013) 3604-3613. (DOI: 10.1002/jps.23661)
  9. J. M. Keen, C. Martin, A. Machado, H. Sandhu, J. W. McGinity and J. C. DiNunzio, Journal of Pharmacy and Pharmacology, 66 (2014) 204-217. (10.1111/jphp.12106)
  10. J. R. Patel, R. A. Carlton, F. Yuniatine, T. E. Needham, L. Wu and F. G. Vogt, J. Pharm. Sci., 101 (2012) 641-663. (DOI: 10.1002/jps.22800)
  11. A. Paudel, Z. A. Worku, J. Meeus, S. Guns and G. Van den Mooter, International Journal of Pharmaceutics (Amsterdam, Netherlands), 453 (2013) 253-284. (DOI: 10.1016/j. ijpharm.2012.07.015)
  12. A. L. Sarode, H. Sandhu, N. Shah, W. Malick and H. Zia, Mol. Pharmaceutics, 10 (2013) 3665- 3675. (DOI: 10.1021/mp400165b)
  13. N. Shah, H. Sandhu, W. Phuapradit, R. Pinal, R. Iyer, A. Albano, A. Chatterji, S. Anand, D. S. Choi, K. Tang, H. Tian, H. Chokshi, D. Singhal and W. Malick, International Journal of Pharmaceutics (Amsterdam, Netherlands), 438 (2012) 53-60. (DOI: 10.1016/j.ijpharm.2012.08.031)
  14. W. Kauzmann, Chemical Reviews, 43 (1948) 219-256. (DOI: 10.1021/cr60135a002)
  15. J. W. P. Schmelzer, Journal of Non-Crystalline Solids, 354 (2008) 269-278. (DOI: http://dx.doi. org/10.1016/j.jnoncrysol.2007.06.103)
  16. R. Surana, L. Randall, A. Pyne, N. M. Vemuri and R. Suryanarayanan, Pharmaceutical Research, 20 (2003) 1647-1654. (DOI: 10.1023/A:1026199604374)
  17. L. S. Taylor, F. W. Langkilde and G. Zografi, J. Pharm. Sci., 90 (2001) 888-901. (DOI: 10.1002/ jps.1041)
  18. R. Surana and R. Suryanarayanan, Powder Diffraction, 15 (2000) 2-6. DOI: 10.1017/ s0885715600010757
  19. R. Suryanarayanan, Pharmaceutical Research, 6 (1989) 1017-1024. (DOI: 10.1023/A:1015970218980)
  20. M. L. P. Leitao, J. Canotilho, M. S. C. Cruz, J. C. Pereira, A. T. Sousa and J. S. Redinha, Journal of Thermal Analysis and Calorimetry, 68 (2002) 397-412. (DOI: 10.1023/A:1016023315613)
  21. D. E. Bugay, A. W. Newman and W. P. Findlay, J Pharm Biomed Anal, 15 (1996) 49-61. (DOI: 10.1016/0731-7085(96)01796-7)
  22. C. M. Deeley, R. A. Spragg and T. L. Threlfall, Spectrochimica Acta, Part A: Molecular and Biomolecular Spectroscopy, 47A (1991) 1217-1223. (DOI: 10.1016/0584-8539(91)80208-z)
  23. P. Chakravarty, S. P. Bhardwaj, L. King and R. Suryanarayanan, AAPS PharmSciTech, 10 (2009) 1420-1426. (DOI: 10.1208/s12249-009-9337-8)
  24. W. P. Findlay and D. E. Bugay, J Pharm Biomed Anal, 16 (1998) 921-930. (DOI: 10.1016/S0731- 7085(97)00058-7)
  25. T. L. Threlfall, Analyst, 120 (1995) 2435-2460. (DOI: 10.1039/an9952002435)
  26. V. A. Windram and T. L. Threlfall, Analytical Proceedings, 29 (1992) 108-110.
  27. R. K. Harris, Analyst, 131 (2006) 351-373. (DOI: 10.1039/B516057J)
  28. J. W. Lubach and E. J. Munson, Polymorphism in the Pharmaceutical Industry, ed. R. Hilfiker, Wiley-VCH, Weinheim, Germany, 2006, pp. 81-93.
  29. T. J. Offerdahl, J. S. Salsbury, Z. Dong, D. J. W. Grant, S. A. Schroeder, I. Prakash, E. M. Gorman, D. H. Barich and E. J. Munson, J. Pharm. Sci., 94 (2005) 2591-2605. (DOI: 10.1002/jps.20469)
  30. B. E. Padden, M. T. Zell, Z. Dong, S. A. Schroeder, D. J. W. Grant and E. J. Munson, Analytical Chemistry, 71 (1999) 3325-3331. (DOI: 10.1021/ac9902446)
  31. F. G. Vogt, Future Medicinal Chemistry, 2 (2010) 915-922. (DOI: 10.4155/fmc.10.200)
  32. P. Varlashkin, American Pharmaceutical Review, 14 (2011) 22-28.
  33. P. Matousek and A. W. Parker, Appl. Spectrosc., 60 (2006) 1353-1357. (DOI: http://dx.doi. org/10.1366/000370206779321463)
  34. B. Schrader and G. Bergmann, Z. Anal. Chem., 225 (1967) 230-247. (DOI: 10.1007/ bf00983673)
  35. J. Johansson, A. SparÈn, O. Svensson, S. Folestad and M. Claybourn, Appl. Spectrosc., 61 (2007) 1211-1218. (DOI: http://dx.doi.org/10.1366/000370207782597085)
  36. C. Eliasson, N. A. Macleod, L. C. Jayes, F. C. Clarke, S. V. Hammond, M. R. Smith and P. Matousek, J Pharm Biomed Anal, 47 (2008) 221-229. (DOI: 10.1016/j.jpba.2008.01.013)
  37. M. D. Hargreaves, N. A. Macleod, M. R. Smith, D. Andrews, S. V. Hammond and P. Matousek, Journal of Pharmaceutical and Biomedical Analysis, 54 (2011) 463-468. (DOI: v http://dx.doi. org/10.1016/j.jpba.2010.09.015)
  38. M. Fransson, J. Johansson, A. Sparén and O. Svensson, Journal of Chemometrics, 24 (2010) 674-680. (DOI: 10.1002/cem.1330)
  39. J. A. Griffen, A. W. Owen and P. Matousek, Analyst, 140 (2015) 107-112. (DOI: 10.1039/ c4an01798f)
  40. G. Metz, X. Wu and S. O. Smith, J. Magn. Reson., Ser. A, 110 (1994) 219-227. (DOI: 10.1006/ jmra.1994.1208)
  41. A. Pines, M. G. Gibby and J. S. Waugh, J. Chem. Phys., 56 (1972) 1776-1777. (DOI: http:// dx.doi.org/10.1063/1.1677439)
  42. E. O. Stejskal, J. Schaefer and J. S. Waugh, J. Magn. Reson., 28 (1977) 105-112. (DOI: 10.1016/0022-2364(77)90260-8)
  43. W. T. Dixon, J. Schaefer, M. D. Sefcik, E. O. Stejskal and R. A. McKay, J. Magn. Reson., 49 (1982) 341-345. (DOI: 10.1016/0022-2364(82)90199-8)
  44. Z. Song, O. N. Antzutkin, X. Feng and M. H. Levitt, Solid State Nuclear Magnetic Resonance, 2 (1993) 143-146. (DOI: 10.1016/0926-2040(93)90032-I)
  45. B. M. Fung, A. K. Khitrin and K. Ermolaev, J. Magn. Reson., 142 (2000) 97-101. (DOI: 10.1006/jmre.1999.1896)
  46. D. H. Barich, E. M. Gorman, M. T. Zell and E. J. Munson, Solid State Nuclear Magnetic Resonance, 30 (2006) 125-129. (DOI: 10.1016/j.ssnmr.2006.06.001)
  47. International Committee on Harmonization, “Validation of Analytical Procedures: Text and Methodology,” Q2(R1), Nov 2005.

Author Biographies

Archana Kumar is a Scientist in the Small Molecule Analytical Chemistry and Quality Control (SMACQC) department at Genentech Inc. Archana holds a Ph.D. degree in organic chemistry from Syracuse University and worked as an organic chemist for 7+ years. She started her career as analytical chemist in 2006 when she joined Elan Pharamceuticals and then Genentech in 2009. She lead analytical and quality support for numerous projects from preclinical to phase 2 and then technical transfer to Roche. Most recently, she is responsible for leading process chemistry analytics group and development of process analytical technologies (PAT) for process understanding during early stage development process.

Joseph Lubach is a Scientist in the Small Molecule Pharmaceutical Sciences department at Genentech, Inc., where he has been since 2007. In his current role, he leads pharmaceutics efforts on numerous projects in discovery and early development, working on physicochemical characterization as well as preclinical and clinical formulations. Additionally, he provides solid-state NMR analysis for Genentech and Roche. Joe holds a Ph.D. in Pharmaceutical Chemistry from the University of Kansas.

Julia Griffen PhD is application scientist at Cobalt Light Systems Ltd. Over the last 2 years at Cobalt her primary focus has been Transmission Raman Spectroscopy and pharmaceutical applications. Julia works on projects such as content uniformity, crystallinity and polymorph stability analysis. Working alongside customers as well as independently generating novel research for publication. Julia received her PhD in Chemistry from the University of Bath, UK.

Jenny Wang MS is a Research Associate in the Small Molecule Analytical Chemistry and Quality Control group at Genentech. Prior to joining Genentech in 2012, she was formerly a senior professional at Allergan, associate scientist at Arena Pharmaceutical and scientist at Pfizer. She holds a Master’s degree in Chemistry from the University of Scranton.

Chi Y. Tsang is a Scientific Researcher with SMACQC group at Genentech in South San Francisco, CA, USA. She earned a B.S. in Biochemistry from California State University, East Bay, CA, USA. Chi has more than 10 years experience in the pharmaceutical industry and recently received her M.B.A. from the University of San Francisco. Her research interests lie in the area of analytical method development for chiral and achiral pharmaceutical compounds, ranging from HPLC, 2DLC, GC method development, to screening platform development.

Jonathan Hau is Scientific Researcher at Genentech Inc. (South San Francisco, CA) where he has worked in the Small Molecule Pharmaceutics Group for over 7 years. His work has ranged from discovery projects focusing on parenteral formulations to development projects focused on spray-dried dispersions. He obtained a Bachelor of Science in Biochemistry from the California Polytechnic State University – San Luis Obispo in 2005 and has worked in the field of formulations and pharmaceutics for over 9 years.

Matthew Bloomfield joined Cobalt Light Systems, Ltd in 2010 as Applications Manager for their transmission Raman and SORS products following 8 years at Renishaw’s Spectroscopy Products Division in the same role. There, he worked extensively in hyphenated Raman techniques, including, NIR-, SEM- AFM-Raman and Tip Enhanced Raman spectroscopy (TERS). Matthew’s Bachelor’s degree is in Geology, and he holds an MSc in Industrial Mineralogy. His PhD studies at the University of Leicester explored mineralogical applications of Raman spectroscopy. Matthew relocated to the USA to start Cobalt’s operations there in November 2014.

Professor Pavel Matousek is an STFC Senior Fellow at the Rutherford Appleton Laboratory (Oxford, UK) where he has worked in the area of vibrational spectroscopy for 25 years. He pioneered the concepts of Kerr gated and Spatially Offset Raman Spectroscopy (SORS). Pavel published over 200 peer-reviewed articles and filed over 10 patents. His honours include the premier Royal Academy of Engineering’s 2014 MacRobert Award, 2009 Charles Mann Award and the 2002 & 2006 Meggers Awards. Pavel is also an Associate Editor of Applied Spectroscopy, a Fellow of the Royal Society of Chemistry, a Fellow of the Society for Applied Spectroscopy, an honorary professor at the University College London and a founding Director of Cobalt Light Systems Ltd.

Larry Wigman is an Analytical Chemist by training with his doctorate from Duke University under the direction of the late Charles Lochmuller; his master’s from San Diego State University under the direction of Reilly Jensen; and his bachelor’s from State University College of New York at Buffalo. He has held various Pharmaceutical Development positions including: Senior Research Scientist at Pfizer, Manager at Mylan, Associate Director at Sanofi, Principal Consultant at Regulitics; and, most recently Principal Scientific Manager of the Small Molecule Analytical Chemistry and Quality Control (SMACQC) Group at Genentech. He joined the relatively new SMACQC Group to lead the analytical develop of small molecule therapeutics. He was attracted by Genentech’s exceptionally strong discovery biology initiative and the ability to translate these discoveries into new drugs that make a difference to patient’s health.

  • <<
  • >>

Join the Discussion