Ensuring raw material quality

Nearinfrared (NIR) spectroscopy is a wellknown tool for identifying raw materials in the pharmaceutical industry. It is a very sensitive, nondestructive technique using overtones and combination bands of fundamental vibrations derived from the mid-infrared. Spectra can be used to distinguish closely related materials such as API polymorphs and excipient analogues. Analysis time is only a few seconds and samples can be analysed in disposable glass vials or directly with a fibre optic probe.

Sheelagh Halsey

However, NIR spectroscopy can be used for more than just raw material identification. With current industry trends using PAT and Quality by Design (QbD) initiatives, building knowledge and reducing risk in production processes are becoming increasingly important. Many companies are focusing research on unit operations such as blending and tableting, but there is much to be gained by revisiting the quality of raw materials entering the process.

Raw materials can now be identified and tested for manufacturing suitability. Most unit operations are physical processes. Physical properties of raw materials affect the manufacturing process when used in an inflexible recipe, resulting in poor quality products. Changes in particle size or polymorphism will influence flow properties and moisture uptake, which in turn affect blend and compression behaviour. The result could be poor content uniformity. NIR spectra are sensitive to these physical changes, as well as chemical identification. Figure 1 shows NIR spectra of lactose polymorphs generated using an FT-NIR analyser (Antaris II; Thermo Scientific, UK). These spectra show that the hydrated form has a marked peak at around 5150 cm^-1 (1940 nm), indicating chemical sensitivity to water. The spectra also demonstrate the sensitivity of the technique to morphology changes. The amorphous form shows fewer features than the crystalline form. For example, many sharp peaks can be seen in the 4000–4500 cm^–1 (2200–2500 nm) region in the crystalline forms, whereas the amorphous form shows little detail. These spectral changes allow the material to be identified as lactose, but also to be qualified as the correct morphology required for a particular process.

Figure 1: NIR spectra of lactose polymorphs.

Figure 2 shows the NIR spectra of microcrystalline cellulose of three different particle sizes. Reflectance spectra exhibit baseline shifts because of different light scattering from the samples. This leads to larger particle size samples having a higher baseline than smaller particle sizes. Hence, the 180-μ cellulose sample has a higher absorbance than the 50-μ sample. Furthermore, the reflectance and scattering is not constant across the whole NIR spectrum. Shorter wavelengths (higher frequency) light penetrates more and scatters less than longer wavelength (lower frequency) light. This gives rise to larger baseline shifts at shorter wavenumbers (longer wavelengths) than at higher wavenumbers (shorter wavelengths). A more comprehensive explanation of diffuse reflectance theory can be found in reference 1.

Figure 2: NIR spectra of cellulose of different particles.

Other issues with the quality of raw materials include the source of the supplier. With continuing drives for cost efficiency in the pharmaceutical industry, alternative sources for many materials are being sought; for example, companies are now beginning to source APIs from Asia. However, changes in quality can affect the manufacture of the final product. This is not to say that alternative suppliers are poorer quality, but just a different quality that has not been considered when validating the process. In the example below, the new material sourced from Asia caused blending problems that resulted in poor content uniformity of the product. Figure 3 shows NIR spectra of several batches of an active from a traditional European source compared with a new sample derived from Asia. The samples were scanned in disposable glass vials for this application. This type of sample presentation is better for quality analysis than using a fibre optic probe. When a probe is used, different operators tend to push the probe into the sample at different strengths. This gives different compressions of the sample leading to variations in the spectrum baseline because of sample presentation, not the sample itself. Presenting the sample in a vial gives a much more repeatable surface to the instrument and provides a better indication of the physical properties of the sample.

Figure 3: NIR spectra of APIs from different sources.

The first impression is that the Asian sample has a much smaller particle size than the European samples. There is some variance in the particle size of the European batches, but the Asian sample is well removed from the natural variation. This is probably the first reason why this batch behaved differently in production. On closer inspection, the Asian sample showed less defined peaks compared with the European samples. These differences can be seen more clearly when the spectra are converted to second derivative (Figure 4).

Figure 4: Second derivative NIR spectra of APIs from different sources.

There are quite significant differences in peak position between the two sets of samples, suggesting a polymorphic change between the two sources. The difference in morphology may result in different flow characteristics or water uptake, either of which may affect the way the material behaves in the process. A combination of particle size and morphology changes resulted in the poor performance of the process because of this raw material. NIR spectroscopy is an ideal tool for monitoring the identity and quality simultaneously in a few seconds.

To implement this type of analysis for quality checking of raw materials, a larger number of batches are required than are needed for identification. Typically, 3–5 batches are required for identification, and 10–30 for quality checking. Most software packages available with NIR instrumentation, however, contain features to start method development with small numbers of batches. More batches can be added individually to each material as required. For each type of testing, there are three major algorithms in common use for raw materials. A percentage match or correlation between a library and unknown sample is the most common algorithm used to identify one material from another. For example, there is only a 35% match between the two sources of API discussed above. In this case, the materials are so different that even a simple algorithm examining peak positions is enough to discriminate these materials. As threshold values would normally be set at 95% or higher, these materials are easily identified one from the other. However, for discriminating between different grades of cellulose, for example, a more rigorous algorithm is required. The two most common algorithms used for quality checking are principal component (PC) analysis or spectral distance matching. Indepth descriptions of these algorithms are outside the scope of this article, but PC methods are described in References 2 and 3, and distance matching or conformity index is described in Reference 4. From experience, the distance matching algorithm works well for most discriminations based on particle size. Typical threshold values are between 3 to 6 standard deviations (SD), the maximum distance between the spectra. The distance between the 50 and 100 μ is 8 SD, outside typical threshold values, showing good discrimination. Further guidance on development and testing of NIR methods can be found in References 5–7.

NIR spectroscopy is a wellknown tool for the identification of raw materials, but it can also be used for quality testing of raw materials as part of the armoury for PAT and QbD process improvements. Keeping control of the quality of the raw materials going in to the unit operation will help minimize variations in processes and help ensure conformity of the final product.

Sheelagh Halsey is a NIR Technical Support Specialist at Thermo Fisher Scientific (UK). www.thermo.com/nir

References

J.M. Olinger and P.R. Griffiths, "Theory of diffuse reflectance in the NIR region," in D.A. Burns and E.W. Ciurczak, Eds, Handbook of NearInfrared Analysis (Marcel Dekker, New York, NY, USA, 1992) pp 13–35.

P.J. Gemperline, L.D. Webber and F.O. Cox, Anal. Chem., 61(2), 138–144 (1989).

O. Svensson, M. Josefson and F.W. Langkilde, Appl. Spectros., 51(12), 1826–1835 (1997).

W. Plugge and C. van der Vlies, J. Pharmaceut. Biomed. Anal., 11(6), 435–442 (1993).

Note for guidance on the use of NIR spectroscopy by the pharmaceutical industry and the data requirements for new submissions and variations, 2003. www.emea.europa.eu/pdfs/human/qwp/330901en.pdf

Near-Infrared Spectrophotometry, European Pharmacopoeia, 2.2.40, 2005.

Near-Infrared Spectrophotometry, USP General Chapters <1119>, 2007.