Analysis of pharmaceuticals by NIR spectroscopy without a reference method

Abstract

Using near-infrared (NIR) spectroscopy for quantitative analysis requires the application of multivariate-calibration techniques and careful selection of representative samples in order to construct accurate calibration models from NIR spectra using partial least squares (PLS). Developing an appropriate set of samples for calibration is difficult and complex, since the quality of models obtained and their prediction ability depends on the selection of these samples. For this work, we developed a new approach to constructing calibration sets for quantifying the active pharmaceutical ingredient (API) and excipients in pharmaceutical tablets. The operating procedure involves using an appropriate experimental design to prepare a set of laboratory samples and recording a series of NIR spectra during the pharmaceutical-production process. Process spectra are calculated by difference between the NIR spectra for production tablets and those for laboratory samples containing API and excipients at their nominal concentrations. Adding the matrix of process spectra to that of NIR spectra for powder samples prepared by weighing the appropriate amount of each component on an analytical balance provides a spectral set encompassing the experimental domain of the production tablets, as confirmed by the scores plot of a principal-component analysis of the results thus obtained. All the quantitative information required to construct the calibration model is the concentration of each component, as calculated from its weight in each mixture; this dispenses with the need to use a reference method and results in increased accuracy in the predicted values. The combined matrix containing process and sample spectra was used in this work to construct five PLS models in order to quantify API and four excipients in the formulation studied. The predictions obtained were quite accurate, and their root mean square error of prediction never exceeded 1.0%. Based on the results, the proposed approach provides a simple, expeditious tool for constructing appropriate sample sets for the development of PLS-calibration models without a reference method.