Quantifying ternary mixtures of different solid-state forms of indomethacin by Raman and near-infrared spectroscopy.

This study assessed the ability of vibrational spectroscopy combined with multivariate analysis to quantify ternary mixtures of different solid-state forms, including the amorphous form. Raman and near-infrared spectroscopy were used to quantify mixtures of alpha-, gamma-, and amorphous indomethacin. Partial least squares regression was employed to create quantitative models. To improve the model performance various pre-treatment algorithms and scaling methods were applied to the spectral data and different spectral regions were tested. Standard normal variate transformation and scaling by mean centering proved to be the best approaches to pre-process the data. With four partial least squares factors, root mean square errors of prediction ranging from 5.3% to 6.5% for Raman spectroscopy and 4.0% to 5.9% for near-infrared spectroscopy were calculated. In addition, the effects of potential sources of error were investigated. Sample fluorescence predominantly caused by yellow amorphous indomethacin was observed to have a significant impact on the Raman spectra. Nevertheless, this source of error could be minimized in the quantitative models. Sample inhomogeneity, particularly in conjunction with a small sampling area when stationary sample holders were used, introduced the largest variation into both spectroscopic assays. The overall method errors were found to be very similar, resulting in relative standard deviations up to 12.0% for Raman spectroscopy and up to 13.0% for near-infrared spectroscopy. The results show that both spectroscopic techniques in combination with multivariate modeling are well suited to rapidly quantify ternary mixtures of crystalline and amorphous indomethacin. Furthermore, this study shows that quantitative analysis of powder mixtures using Raman spectroscopy can be performed in the presence of limited fluorescence.