(W1230-11-61) In Situ Monitoring of Dissolution and Crystallization Processes for Active Pharmaceutical Ingredients Using Low Frequency Raman Spectroscopy

Purpose: Optimization of the manufacturing process based on scientific evidence is essential for quality control of active pharmaceutical ingredients (API). Real-time monitoring is effective for obtaining crystals of stable quality in the crystallization process. Raman spectroscopy is attractive tool for pharmaceutical quality evaluation and process analytical technology because of its ability to analyze samples nondestructively and rapidly. In particular, Raman spectra in the low-frequency region are expected to be useful for monitoring the crystal polymorphs of API, which could be difficult to identify in the conventional spectral region, because the scattering peaks can reflect the difference in lattice vibration derived from polymorphism. In this study, the process of polymorphic transition of bulk API was monitored using low-frequency Raman spectroscopy to identify crystal polymorphs. In addition, we attempted to develop a model that enables quantification of each polymorph during dissolution and crystallization processes of API using multivariate analysis of obtained spectra.

Methods: Five model drugs, Acetaminophen, Indomethacin, Theophylline, Caffeine and Carbamazepine (CBZ), were selected. For each compound, a methanol-saturated solution was prepared, and a suspension was prepared by adding a fixed amount of the compound to the methanol-saturated solution. The Raman spectra of each API suspension were measured under the same conditions using a probe type Raman spectrometer (Ondax, Inc.). Subsequently, in-situ Raman monitoring of the dissolution and crystallization processes of the compounds considered to have sufficiently high Raman sensitivity was attempted using Crystalline (Technobis B.V.) by changing the temperature of the suspension. Preprocessing and multivariate analysis of Raman spectra were performed using the data analysis software Solo+MIA (Eigenvector Research Inc.). A multivariate calibration model was developed using partial least squares regression (PLSR) analysis with the scattering intensity at each wavenumber of the Raman spectrum as the input variables (X), and the theoretical concentration of CBZ in the suspension as the response variable (Y). The predictive performance of the model was evaluated using the coefficient of determination, root mean square error of cross-validation (RMSECV), and root mean square error of prediction (RMSEP). The spectral preprocessing condition was the standard normal variate (SNV). Both the input variables and response variable were scaled using mean centering.

Results: Comparison of Raman spectra of suspensions of each compound showed that the Raman sensitivity tended to be worse for compounds with larger molecular weights. This was considered to be due to the fact that, in general, the larger the molecular weight of a compound, the poorer its crystallinity. Although some differences in detection sensitivity were observed among the five model APIs selected for this study, the spectra derived from the APIs could be adequately detected by increasing the concentration of the suspension for each compound. Carbamazepine (CBZ) was used for in-situ monitoring during the dissolution and crystallization process. Raman monitoring of a suspension of CBZ form I, the metastable form, during the process of complete dissolution followed by recrystallization　showed that the peaks at 24 cm^-1, 58 cm^-1, and 107 cm^-1 from form I weakened as the temperature increased, and the peaks at 35 cm^-1, 103 cm^-1, and 118 cm^-1 from form III, the stable form, increased as the temperature cooled. This suggests that it is possible to distinguish between the crystal polymorphs of CBZ, form I and form III. To enable simultaneous quantification of form I and form III in the quantitative model, 33 suspension samples were prepared for calibration and 6 samples for validation, with each concentration varying from 0 to 150 mg/mL. A concentration-dependent increase in CBZ peak intensity was observed in each sample group. Furthermore, the calibration model developed using Raman spectra with wavenumbers ranging from 0 to 154 cm^-1 showed RMSEP of 4.38 mg/mL for form III and 5.74 mg/mL for form I, suggesting that the concentration of CBZ form I and form III in the suspension can be determined with practical accuracy. The results of in-situ Raman monitoring with the quantitative model showed that CBZ form I was completely dissolved at 55°C when the suspension was heated from 20°C at a rate of +2°C/min, and that CBZ form III began to crystallize at 30°C when the solution was cooled at a rate of -1°C/min.

Conclusion: In this study, we successfully developed a quantitative model that enables simultaneous determination of CBZ form I/form III content in suspension using low-frequency Raman spectroscopy. It is suggested that in-situ monitoring of crystal polymorph transition and its quantification is possible by applying the developed quantitative model to the crystallization process of active pharmaceutical ingredients.