Process analytical methods for monitoring of the film coating on pharmaceutical pellets and tablets using NIR spectrometry and image analysis

Abstract: Pharmaceutical tablets and pellets are often covered with a coating. There is a need for methods that enable monitoring of the growth of such films during processing. Such analyses can be performed using process analytical chemical methods based on near infrared (NIR) diffuse reflectance spectrometry. It provides with analyses that take only ca. 8 seconds in total analysis time, including physical sampling. The use of chemometrics is the key to handle the complexity of the NIR spectra. The chemometric methods of principal component analysis (PCA) and partial least squares (PLS) were used to study and to calibrate NIR-spectra to the amount of coating. An at-line method using NIR-spectra for determination of the amount of coating of film applied on pharmaceutical tablets was developed. Limitations for quantitation of the film were determined by the 'information depth', ca 0.1 - 0.2 mm. Novel tools to determine the coating thickness and its variations on pharmaceutical pellets were based on image analysis and NIR diffuse reflectance spectrometry. Data from the images were used to predict variations in the release rate of the active substance due to geometrical variations. The image analysis was further used as a reference method for thickness measurements. Together with a theoretical film thickness growth model with physical relevance to underlying process mechanisms, an in-line process analytical chemical method based on NIR was enabled for analysis of the coating thickness on pharmaceutical pellets (RMSEP = 2.2 µm, range tested 0 - 50 µm). Indeed, such tools have not been combined previously, and through use of a multivariate batch calibration, the thickness values delivered from the NIR spectrometer became less disturbed by commonly varied process conditions. Additionally, variations in the coating thickness could be indicated by NIR . The spectral artefacts encountered for samples moving during acquisition of a scan were characterised in a separate study using multivariate methods. These artefacts could be detected by PCA.