(445d) Estimation of Active Pharmaceutical Ingredients Content In Blending Process for Drug Products Manufacturing

In recent years, a lot of research in online quality estimation and control technology has been carried out to achieve Real-Time Release Testing (RTRT) of drug products in the pharmaceutical industry. Non-destructive and rapid Near-Infrared (NIR) spectroscopy is expected to be a powerful online monitoring method. However, NIR spectrum measured by NIR spectroscopy is affected by water content and particle properties in real time manufacturing process. In addition, unknown factors, which cannot be detected by existing sensing technologies, may affect NIR spectrum. Therefore it is very difficult to extract the quality information from the spectrum data. This study focuses on the content uniformity estimation of Active Pharmaceutical Ingredients (API) in the blending process. In the proposed method, particle size data are used as inputs in addition to spectrum data to take account of the influence of particle size distribution on NIR spectrum, and locally weighted partial least squares (LW-PLS) is applied to estimate API content. LW-PLS is a Just-In-Time modeling method that builds a local linear regression model on-line whenever output prediction is required. Furthermore, a statistical wavelength selection method is proposed. This method can select a suitable set of wavelengths from experimental data; the selected wavelengths are found to include those corresponding to spectrum peak specific for API. LW-PLS and the proposed wavelength selection method were applied to real process data, and the estimation accuracy was improved by 18.7 % in Root Mean Square Error of Prediction (RMSEP) compared with the conventional PLS and wavelength selection method base on spectrum peak positions. The results clearly show that the proposed method is useful for the API content estimation and is superior to the conventional method.