International Journal of Pharmaceutics 417 (2011) 32��� 47 Contents lists available at ScienceDirect International Journal of Pharmaceutics j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / i j p h a r m Near infrared and Raman spectroscopy for the in-process monitoring of pharmaceutical production processes T. De Beer a,���, A. Burggraeve a, M. Fonteyne a, L. Saerens a, J.P. Remon b, C. Vervaet b a Laboratory of Pharmaceutical Process Analytical Technology, Department of Pharmaceutical Analysis, Ghent University, Harelbekestraat 72, B-9000 Gent, Belgium b Laboratory of Pharmaceutical Technology, Department of Pharmaceutics, Ghent University, Harelbekestraat 72, B-9000 Gent, Belgium a r t i c l e i n f o Article history: Received 18 October 2010 Received in revised form 3 December 2010 Accepted 9 December 2010 Available online 15 December 2010 Keywords: Raman NIR Process Analytical Technology In-process monitoring a b s t r a c t Within the Process Analytical Technology (PAT) framework, it is of utmost importance to obtain crit- ical process and formulation information during pharmaceutical processing. Process analyzers are the essential PAT tools for real-time process monitoring and control as they supply the data from which relevant process and product information and conclusions are to be extracted. Since the last decade, near infrared (NIR) and Raman spectroscopy have been increasingly used for real-time measurements of crit- ical process and product attributes, as these techniques allow rapid and nondestructive measurements without sample preparations. Furthermore, both techniques provide chemical and physical information leading to increased process understanding. Probes coupled to the spectrometers by fiber optic cables can be implemented directly into the process streams allowing continuous in-process measurements. This paper aims at reviewing the use of Raman and NIR spectroscopy in the PAT setting, i.e., during processing, with special emphasis in pharmaceutics and dosage forms. �� 2010 Elsevier B.V. All rights reserved. 1. Introduction The Food and Drug Administration���s (FDA) Process Analyti- cal Technology (PAT) initiative (FDA, 2004) forms the basis of the pharmaceutical Good Manufacturing Practice (GMP) rules for the 21st century (Hinz, 2006). Because the pharmaceutical indus- try is highly regulated, final products must meet very stringent specifications. However, this does not mean that pharmaceutical processes are optimized. Conventional pharmaceutical manufac- turing is generally accomplished using batch processing with off-line time-consuming and less efficient laboratory testing con- ducted on randomly collected samples to evaluate quality. The processes themselves are not fully understood and are often ineffi- cient black-boxes. Limited relevant information is mainly obtained after the process, making process control difficult which can result in batch losses. The ultimate goal of PAT is a better fundamen- tal scientific understanding of manufacturing processes (i.e., the process should not be a black box system). One of the most impor- tant statements within the PAT concept is that ���quality should not be tested into products it should be built in���. PAT should therefore play a crucial role in design, analysis, and control of manufacturing processes based on timely in-line, on-line and at-line measure- ments (i.e., during processing) of critical quality and performance ��� Corresponding author. Tel.: +32 9 2648097 fax: +32 9 2228236. E-mail address: Thomas.DeBeer@UGent.be (T. De Beer). attributes of raw and in-process materials, with the goal of ensur- ing final product quality (EUFEPS QbD and PAT Sciences Network, 2010). It is thus aimed to obtain real-time information of all criti- cal process aspects and to guide processes towards their desired state, hence ensuring the quality of each end product and pos- sibly allowing real-time release. Continuous gathering of critical process information during production should allow real-time pro- cess adjustments to keep product specifications within predefined limits, hence avoiding batch loss. For a long time, innovations in pharmaceutical manufacturing and quality assurance have been slowed down by the stringent regulatory constraints within the pharmaceutical industry which allowed little room for change and which significantly contributed to the aversion to bring new manufacturing technologies and quality assurance methods to the attention of the regulators (in order to avoid delaying regula- tory approval). The PAT initiative to move towards a risk- and science based approach for pharmaceutical processing is now strongly encouraged by the most important pharmaceutical reg- ulatory authorities (e.g., FDA and the European Medicines Agency, EMA). Process analyzers are the essential PAT tools for real-time pro- cess monitoring and control as they supply the data from which relevant process and product information and conclusions are to be extracted. Available tools have evolved from those that pre- dominantly take univariate process measurements, such as pH, temperature, and pressure, to those that provide multivariate infor- mation related to biological, physical, and chemical attributes of the 0378-5173/$ ��� see front matter �� 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpharm.2010.12.012

T. De Beer et al. / International Journal of Pharmaceutics 417 (2011) 32��� 47 33 materials being processed (e.g., Raman and NIR spectrometers). In a PAT environment, real-time process measurements can be - At-line: measurements where the sample is removed, isolated from, and analyzed in close proximity to the process stream. - On-line: measurements where the sample is diverted from the manufacturing process, and may be returned to the process stream. - In-line: measurements (invasive or noninvasive) where the sam- ple is not removed from the process stream (FDA, 2004). The following aspects should be considered when implementing process analyzers into process streams: - selection of a suitable process analyzer or combination of com- plementary process analyzers able to monitor the desired critical process and product information, - determination of the locations in the process streams where and how process analyzers should be and can be implemented to monitor the required information, - determination of the optimal measurement conditions for the process analyzer to obtain useful data, - validation of the performance of process stream analyzers (over time). Since the last decade, near infrared (NIR) and Raman spec- troscopy have been increasingly used for real-time measurements of critical process and product attributes during pharmaceuti- cal processing, as these spectroscopic techniques allow rapid and nondestructive measurements without sample preparations. Fur- thermore, probes coupled to the spectrometers by fiber optic cables can be implemented directly into the process streams allowing continuous real-time in-process measurements. A quick search in Web of Sciences�� shows the strongly increased number of publi- cations during the last decade about the use of Raman and/or NIR for pharmaceutical in-process measurements. Besides the practical advantages, an important reason for the high interest for Raman and NIR spectroscopy as process analyzers is their ability to supply versatile and multivariate information. One single spectrum may contain qualitative and quantitative physical and chemical infor- mation. This paper aims at reviewing the use of Raman and NIR spec- troscopy in the PAT setting, i.e., during processing, with special emphasis in pharmaceutics and dosage forms. Hence, there will be focused on applications where NIR and Raman spectroscopy have been used in-line, at-line and on-line to monitor and control phar- maceutical production processes in real-time. Many studies have been performed using Raman and/or NIR spectroscopy as off-line tools for chemical and physical analysis of pharmaceuticals (e.g., API quantification in tables and co-crystal screening). Evidently, these studies also contribute to the increased understanding of the production of pharmaceuticals. Several reviews describe and eval- uate these off-line Raman and NIR applications (Aaltonen et al., 2008, 2009 Fevotte, 2007 Heinz et al., 2009 Luypaert et al., 2007 McGoverin et al., 2008 Pinzaru et al., 2004 Radtke et al., 1999 R��s��nen and Sandler, 2007 Reich, 2005 Strachan et al., 2007 Tantipolphan et al., 2008 Tummala et al., 2005 Wen, 2007 Yu et al., 2007). 2. NIR versus Raman spectroscopy NIR and Raman spectroscopy are both molecular vibra- tional spectroscopic techniques studying vibrational transitions in molecules. 2.1. Raman spectroscopy The Raman effect was first observed in 1928 by Sir C.V. Raman (Raman and Krishnan, 1928). However, this effect was already predicted in 1923 by Smekal, based on theoretical calculations (Smekal, 1923). The Raman effect is the inelastic scattering of electromagnetic radiation (EMR) as a result of energy exchange between the radiation and molecular vibrations. The theoretical background of the Raman effect is well described in the literature (Gardiner, 1989 Long, 1977 Nakamoto, 1986 Stoicheff, 1959). In Raman spectroscopy, the samples are irradiated with monochromatic laser light. Typically, lasers producing laser light in the visible (e.g., 532 nm) or near-infrared (e.g., 785 nm and 1064 nm) range are used. The energy of this light is higher than the energy needed to bring molecules to a higher vibra- tional state (Fig. 1). Most of the incident radiation is scattered by the sample molecules at the same frequency (=energy). This identical scattered light is called Raleigh radiation. Only 10���8 is scattered inelastically by the sample molecules, indicating that energy exchange occurred between the incident light and the sample. This inelastic scattering is called the Raman effect. This inelastic scattered radiation can have a lower energy (lower fre- quency) than the incident radiation or a higher energy (higher frequency) than the incident radiation. The first type of inelas- tic scattering is called Stokes radiation, the latter anti-Stokes radiation (Fig. 1). At room temperature, mainly Stokes radia- tion will occur. At high temperatures (e.g., 500 ���C), many sample molecules are already at a higher vibrational state as can be derived from the Boltzmann distribution, hence favoring anti-Stokes radiation. The selection rule for molecules to be Raman active is that a change in polarizability of the molecule occurs during its normal modes. The polarizability of a molecule is the ease with which the electron cloud of a molecule can be distorted after bringing the molecule in an electromagnetic field (i.e., by light irradia- tion). However, molecules are no static entities, but continuously in motion by vibrations, rotations and translations. The actual vibration of a molecule seems arbitrary, but is in fact the super- position of some simple vibration modes, called normal modes. Each of these normal modes has its own frequency. An atom has 3 degrees of freedom (x, y and z direction). Hence, a molecule containing N atoms has 3N degrees of freedom: 3 translations, 3 rotations (2 for a linear molecule) and 3N-6 normal modes (3N-5 for linear molecules). For a diatomic molecule, e.g., the ease to distort the electron cloud will be different for the three vibration states of its harmonic vibration having frequency v , as schematically shown in Fig. 2a. The larger the bond distance, the further the electrons are apart from the nuclei an thus the easier they can be moved. This change in ease with which the electron cloud of a molecule can be distorted (i.e. change in polarizabil- ity) is needed for Raman activity. Complicated quantum mechanics and the group theory are needed to determine for which of the normal modes of complex molecules a change in polarizability occurs. A Raman spectrum displays the frequency difference between the incident radiation and the scattered radiation, expressed as wavenumbers, versus the intensity (I) of the scattered radiation (Fig. 3). Placzek developed the following equation which allows the calculation of the Raman intensity during a particular vibration: I = cte v0 + vv vv 4 NI0 1 ��� e(���hvv /kT ) 45 ( ��S )2 + 13 ( ��a )2 (1) where v0 is the frequency of the incident radiation, vv is the vibra- tion frequency of the molecule, N is the number of irradiated Raman active molecules, I0 is the power of the light source, h is Planck���s constant (=6.6260 �� 10���34 J s), k is the Boltzman���s constant, T is the

34 T. De Beer et al. / International Journal of Pharmaceutics 417 (2011) 32��� 47 Fig. 1. IR and NIR absorption, the Raman effect and fluorescence. temperature, ��S is the polarizability from the molecules causing Stokes radiation, and ��a is the polarizability from the molecules causing anti-Stokes radiation. However, besides these parameters the Raman intensity is also determined by instrumental (e.g., detector and length of glass fiber when using probes) and sample parameters (e.g., sampling volume, sample particle size, refractive index of sample and concentration of the considered compound). Furthermore, as scattering occurs in all directions, only a fraction is detected. When probes are imple- mented into process streams, only the light which is scattered into the same direction as where the incident light comes from is mea- sured. The irradiation of materials can result in different phenom- ena: scattering, absorption and fluorescence. As the Raman effect is inherently weak, the other phenomena can interfere strongly. Moreover, heating, photodecomposition and laser ablation can occur. The incident laser light can be absorbed when the wave- length of this light corresponds with the absorption band in the spectrum of the molecule, resulting in the transition of the molecule to the excited state (dependent of the laser to the excited vibra- tional or electronic state). The absorbed energy is then transformed via radiationless transitions into thermic energy. This absorption interferes quite strongly with the Raman effect as the intensity of Raman scattering is proportional to the intensity of the laser light. This interference can be avoided by using another wavelength laser. Besides the absorption of the laser light, absorption of the scattered light can likewise occur. Fluorescence can produce sub- stantial interferences in Raman spectroscopy when the molecule is excited to an electronic excited state. The molecule decays to a lower energy level by a radiation-less transition (the vibrational ground state in the electronic excited state), followed by the decay to the electronic ground state (Fig. 1). During this last decay, flu- orescent radiation is emitted, interfering with the Raman signal. Interference by fluorescence can be avoided employing lasers with Fig. 2. (a) Vibration states of diatomic molecule. (b) No change in dipole moment during the stretch vibration of an X2 molecule. (c) Change in dipole moment during the stretch vibration of an XY molecule.