Pseudo-polymorphic conversion by near-infrared spectroscopy

Near-infrared spectroscopy (NIR) is suitable for the analysis of pharmaceutical samples in various solid forms,^1–3 and can be used for determining chemical properties (e.g., content of drug, water), as well as physical properties (e.g., particle size, tablet hardness).^4,5 In some cases, it is even used for very specific tests such as the determination of different polymorphs.^6,7 In this study, NIR was used at-line to qualitatively and quantitatively monitor the rate and extent of conversion of a solvate form of a model API, which is manufactured in the ethanolate form.

Alex Buckhingham/Getty Images

Under elevated temperature and humidity, partial or complete conversion of the ethanolate to hydrate can occur, creating a challenge for formulation scientists attempting to develop a solid dosage form using a fluid-bed wet granulation process. The determination of the rate and extent of ethanolate–hydrate conversion are critical during formulation development. As part of the quality control strategy, Karl Fischer (KF) titration methods have been developed and validated to measure water content, alongside gas chromatographic (GC) methods that measure ethanol content. The combined use of these methods provides the water and ethanol contents in drug substance and granule samples. However, these methods are unsuitable for monitoring the wet granulation process because of the speed of the analysis. In addition, KF and GC methods cannot differentiate the bound versus surface ethanol/water. The adoption of NIR can facilitate formulation development.

Table 1 Composition of liquid binder.

Materials and methods

Diffuse reflectance spectra for the API and granules were obtained with a FOSS XDS Near-Infrared Rapid Content Analyzer (Foss NIRSystems; Silver Spring, MD, USA). The powdered samples (~500 mg) were transferred into 15×45 mm glass vials with rubber-lined caps and then scanned by placing the vial on the sample window. Each sample was scanned 32 times to obtain an averaged spectrum for further processing.

Table 2 NIR versus KF results for total amount of water in granules sampled at different time-points of the fluid-bed granulation process.

Fluid-bed granulation of the API was performed in a Glatt GPCG-1. The processing steps and sampling times include:

• Weighed API was charged into the bowl of the Glatt GPCG-1.
• Preheat samples were taken at the start of the fluidization.
• The API was preheated by fluidizing it for ~2 min.
• Prespray samples were taken after the inlet air temperature was stabilized.
• The granulation fluid was sprayed at a predetermined rate.
• Samples were taken every 20 min during the process and at the end of spray.
• The wet granules were dried until the exhaust temperature reached 40 °C.

• Samples were taken from the dried granules.

The sampling times are depicted as suffixes to the batch identification numbers (eg., 90250H040-x), where 'x' represents the sampling time (i.e., '1' for preheat; '2' for prespray; '3' for 20 min; '4' for 40 min; '5' for end of spray (60 min of granulation) and '6' for dried granules). Part of each sample was tested for loss on drying (LOD) using the Computrac Max 200 Moisture Analyzers (Arizona Instrument LLC; Tempe, AZ, USA). The remaining part of each sample was sealed into scintillation vials for further analysis. The granulating fluid was composed of purified water, absolute ethanol or hydro-alcohols of different composition (Table 2). As the boiling point and latent heat of vaporization of the granulating fluids are different, this experiment is expected to provide an insight into the kinetics of pseudo-polymorphic conversion.

Results and discussion

Differentiation of bound versus surface water/ethanol by NIR. In the NIR region, water has strong absorption bands with peak maxima of ~1440 (overtone) and 1930 nm (combination band). Similarly, because of its hydroxyl functional group, ethanol also has strong absorption bands at 2082 and 2309 nm. Figure 1 shows the spectra of three representative API samples that achieved different degrees of ethanolate–hydrate conversion. The water content of these samples was determined to be 1.15% (spectrum in blue), 3.42% (green) and 6.27% (red).

The expanded spectra of these three samples are shown in Figure 2. Two strong absorption bands with maxima of 1909 and 1933 nm were observed, and the magnitude of these two bands is proportional to the water content in the samples. Therefore, they are identified as the combination bands of water. It is very interesting that the water combination band shows a split doublet, indicating that the two water molecules may have hydrogen-bonding interactions with two different functional groups of the same API molecule. The water combination band at 1909 nm shows a significant blue shift compared with the free water combination band, indicating that hydrogen-bonding interactions between the water and API molecules have replaced the hydrogen-bonding interactions between water molecules.

Figure 1

To differentiate bound versus surface water, the API sample (ethanolate) was spiked with 1.92% of water (surface water). The spiked sample was prepared by weighing ~500 mg of the drug substance into a 15×45 mm glass vial. Water was mixed with the drug substance in the vial using a 10 μL-syringe and the amount of water spiked was determined by weighing. After spiking, the sample immediately was scanned by NIR to minimize the possibility of ethanolate–hydrate conversion. NIR spectra of the spiked and unspiked samples are presented in Figure 3. Two strong absorption bands corresponding to surface water at 1440 and 1930 nm can be seen in the spectrum of the spiked sample (spectrum in purple). The 1930 nm band is a single band (with the ethanol absorption band on its back shoulder) similar to that of free water. This evidence suggests that the API hydrate contains bound water.

Figure 2

The combination band of ethanol, which is a doublet with maxima at 1973 and 1994 nm, can be identified in Figure 2. The magnitude of this band is inversely proportional to the amount of water in the samples and proportional to the ethanol content. In addition, there is a triplet with maxima at 2053, 2068 and 2087 nm (Figure 3), which is also related to the presence of ethanol. Compared with the combination band of neat ethanol (2082 nm), just like the water combination bands of the hydrate form of the API, the ethanol combination band of the ethanolate form of the API is split, indicating that the hydrogen bonding occurs between ethanol and two different API functional groups. This observation is consistent with that of the hydrate. Furthermore, both wavelengths of the maxima show significant blue shifts that strongly support the assumption of bound ethanol in the ethanolate form of API.

Figure 3

Ethanolate–hydrate conversion kinetics of the API. To study the ethanolate–hydrate conversion kinetics, two API samples (open dish) were placed in a humidity chamber at 40 °C/75% relative humidity (RH). Aliquots of samples were pulled at different time-points. To obtain a relatively uniform sample, the contents of the open dishes were mixed using a spatula before the samples were taken. After a sample was pulled, the sample dish was immediately returned to the humidity chamber. For the first dish, samples were pulled between 0–9 h. For the second sample dish, samples were pulled between 14–31 h.

The pulled samples were scanned by NIR, and then tested for water and ethanol using the KF and GC methods. Based on the water and ethanol assay results, calibration models were established for the determination of water and ethanol using NIR.

The conversion study results are presented in Figure 4. The drug substance lot used in the study contained 7.43% ethanol and 0.30% water initially. For the API in the first open dish, the ethanolate–hydrate conversion reached ~91% completion in 9 h. For the sample in the second dish, complete conversion took ~22 h.

The difference in the rate of conversion was attributed to the fact that the samples were handled differently. The sample in the first dish was remixed every time an aliquot was pulled while the sample in the second dish was not disturbed until 14 h. The conversion occurred mostly in the surface layer of the powder directly in contact with moisture and sample remixing accelerated the conversion process.

Figure 4

Analysis of granule samples by NIR

During development of the fluid-bed wet granulation process, purified water and absolute ethanol were used as solvents for preparation of the binding solutions. In an attempt to study the kinetics of pseudo-polymorphic conversion, hydro-alcohols of different composition were used (Table 1). Samples taken from the process were scanned by NIR and then analysed by KF and GC methods. The results were used to build Multiple Linear Regression calibration models for the at-line NIR measurements. The wet and dry granule samples taken have been indicated by the suffixes to the batch number (Tables 2–4).

Table 3 NIR versus GC results for total amount of ethanol in granules sampled at different time-points of the fluid-bed granulation process.

The NIR predicted results for the total moisture content of the granules sampled at different time-points during the wet granulation process had no significant (p<0.01) difference to that obtained by the KF titration method (Table 2). The total ethanol content of the granules determined by the GC method was also found statistically (paired t-test) similar to that of the NIR results (Table 3). During the study, LOD results were also obtained by thermogravimetric analysis (Table 4), which were not consistent with the combined KF and GC results. The discrepancies were attributed to the fact that the ethanolate drug melts (phase change) before reaching 900 °C and the drug substance loses ethanol during further heating to 1000 °C. LOD cannot be used as an at-line test for process control. Water-based granulations resulted in ~50% ethanolate–hydrate conversion (Tables 2 and 3). The rate of conversion was decreasing in the form of first-order (1.6%/min to 0.55%/ min at the end of spray). All other batches retained their bound ethanol until end of spray and then lost 5–9% during drying. This experiment demonstrated that in the presence of alcohol, as high as 50%, water as a granulating fluid didn't affect the pseudo-polymorphic conversion of the ethanolate form of the API.

Conclusion

This work demonstrates the application of NIR in monitoring pseudo-polymorphic conversion at-line and in obtaining suitable granulation parameter, as well as composition of the liquid binder. This nondestructive analytical tool is powerful in determining ethanol and water in wet mass, as well as in dried granules, in a manner that is faster than the standard procedure of Karl Fischer titration and gas chromatography. NIR can also provide answers to questions regarding bound versus surface ethanol and water in drug substance and product.

Table 4 Total solvent estimate in granules sampled at different time-points during the fluid-bed granulation process. LOD versus KF/GC results.

Acknowledgements

The authors would like to thank the technical teams of the Pharmaceutical and Analytical Departments of Johnson & Johnson Pharmaceutical R&D (Spring House, PA, USA) for their contribution in designing and executing the experiments described in this article.

References

1. D.J. Wargo and J. K. Drennen, J. Pharm. Biomed. Anal., 14(11), 1415–1423 (1996).

2. P. Frake et al.,Int. J. Pharm.,151(1), 75–80 (1997).

3. N. W. Broad et al.,Analyst,126(12), 2207–2211 (2001).

4. E. W. Ciurczak et al., Spectrosc., 1(7), 36–39 (1986).

5. J. D. Kirsch et al., J. Pharm. Biomed. Anal.,19(3–4), 351–362 (1999).

6. M. Blanco and A. Villar, Analyst, 125(12), 2311–2314 (2000).

7. W. Li et al., J. Pharm. Sci.,94 (12), 2800–2806 (2005)

Abraham Woldu is Senior Scientist and Team Leader at Johnson & Johnson Pharmaceutical R&D (PA, USA).

Weiyong Li is is a Research Fellow at Johnson & Johnson Pharmaceutical R&D.

Denita Winstead is Director of Johnson & Johnson Pharmaceutical R&D.