Content uniformity analysis by near-infrared (NIR) spectroscopy represents one of the most successful applications of process analytical technology (PAT) in tablet manufacturing in the pharmaceutical industry. A key component of a NIR content uniformity analysis application is a multivariate calibration model that predicts the contents of the active pharmaceutical ingredient (API) in tablets. Content uniformity is a regulated product quality attribute for tablets and often requires a sophisticated NIR calibration model for the analysis.
Apparently after the deployment of a NIR method to the manufacturing environment, transferring the multivariate calibration model is unavoidable because of changes in instruments and sampling accessories, etc. Many near infrared calibration transfer algorithms have been developed, such as spectral standardization, robust calibration, instrument standardization, etc. , and these methods typically required the same set of tablets being collected in the primary and secondary spectrometers. However, since many pharmaceutical tablets degrade during their shelf-lives, it could be difficult to keep a set of tablets at the same condition as the time the calibration set was measured.
In this work, we developed a NIR calibration transfer method specifically for the content uniformity analysis in tablet manufacturing. This method contains three standard rare earth oxides and glass tablets and a spectral conversion algorithm based on Fourier transform operations. NIR spectra of these stable rare earth oxide and glass tablets were used to characterize spectral feature changes resulting from instrument and accessory changes. Calculations based on instrument design principles were then conducted to match the spectra before and after the instrument change. In the final step, chemometric preprocessing methods were used to filter out unmatched spectral features and the converted and filtered spectra were used to build a calibration model for the new measurement condition. In this work, the method was demonstrated by transferring a manufacture scale NIR calibration model from a grating spectrometer to a Fourier transform near infrared (FT-NIR) spectrometer.
Standard Tablets: Three standard tablets were developed for near infrared calibration transfer. Each tablet contained one rare earth oxide and glass power. The three rare earth oxides were Dy2O3, Er2O3, and Ho2O3, which were components of the NIST 1920a and KemTek 1920x standards for near infrared spectrometer wavelength calibration. The standard tablets had approximately the same diameter and thickness as the pharmaceutical tablet product.
To make the standard tablet, a stainless steel tablet mold about the same size as the tablet product was made in-house. The stainless steel mold was filled with power mixture of a rare earth oxide and glass and baked in a Muffle furnace. After the power melted, the stainless steel mold was removed from the furnace and allowed to cool to room temperature and the tablet was separated from the mold for use.
Pharmaceutical Tablets:A total of 196 tablets were used to develop a content uniformity analysis method for a tablet product. The tablets were collected in the beginning, middle, and end of several manufacture scale batches during the setup of the manufacture process. The actual content of the active pharmaceutical ingredient (API) varied from 85% to 115% of the API strength. The accurate API content in each tablet was analyzed by a UV-Visible method. The weight of each tablet was also measured before the UV-Visible analysis.
Near Infrared Spectra: Near infrared spectra of the pharmaceutical tablets and the three rare earth oxides and glass tablets were measured both in a grating near infrared spectrometer and a FT-NIR spectrometer for the study of the near infrared calibration transfer method. In this study, the grating spectrometer was treated as the primary spectrometer and the FT-NIR spectrometer was treated as the secondary spectrometer. The infrared spectra were collected in the transmission mode in both spectrometers. The grating spectra were collected with 10nm nominal resolution and the FT-NIR spectra were collected at 4cm-1 resolution.
Calibration Model: Partial least squares (PLS) regression was used to build the calibration models. Independent validation was used to validate the calibration model from the primary spectrometer. About 60% of the tablets were randomly selected as the calibration set and the rest 40% spectra were used as the validation set. Second derivative (SD), multiplicative scattering correction (MSC), and standard normal variate (SNV) or their combinations were used preprocess the NIR spectra before building the calibration model.