279149 Prediction of Post-Relaxation Radial Tensile Strength of Pharmaceutical Tablets Using At-Line Collected near Infrared Data
The prediction of relaxed radial tensile strength (RTS) of pharmaceutical tablets in real-time using near infrared (NIR) spectroscopic data collected at-line was investigated. Radial tensile strength is a critical quality attribute (CQA) of tablets which may correlate to drug product performance in vivo. However; post-compaction, the tablets undergo axial and radial relaxation, which influences dosage form characteristics such as disintegration time, dissolution time, and mechanical strength. It is the RTS of completely relaxed tablets that is of interest, as it is a completely relaxed tablet that will be administered to a patient. The existing work relating NIR spectra to RTS relies on the use of data collected from relaxed tablets (off-line NIR data) to predict the relaxed (off-line) RTS. However, the NIR analysis of tablets in a manufacturing setting, as desired in the current Quality by Design philosophy and accomplished in a modern real-time release environment, necessitates that tablet quality attributes are determined immediately following compression and prior to complete relaxation.
A six component system was compressed with variability in excipient ratio, disintegrant level, and compression force (total of sixteen design points). Eighteen tablets were collected per design point and scanned at-line using a commercial NIR spectrometer immediately after compression. Tablets were left to relax for two weeks. Their physical dimensions were then measured and they were crushed to calculate RTS values. A multiple linear regression (MLR) was used to predict RTS of relaxed tablets from at-line spectra. In a first step, partial least squares models were developed to predict variables with a significant relationship to RTS (tablet density, tablet volume). Outputs from the PLS models were then used along with spectral slope and formulation parameters as input in an MLR to predict RTS. Radial tensile strength predictions on an independent test set had an error of 0.25 MPa with a R2 of 0.8. The MLR model had calibration and cross validation error between 0.2 – 0.3 MPa. Updating spectral information with physical features and chemical information through MLR allowed for predicting RTS of relaxed tablets using at-line information while compensating for the effect of tablet relaxation over time. This approach has the potential to be implemented on-line since all the information needed by the MLR model can be calculated from the same near infrared spectrum. Automating spectral collection will allow the real-time prediction of relaxed RTS at the tablet press level.
See more of this Group/Topical: Topical I: Comprehensive Quality by Design in Pharmaceutical Development and Manufacture