Pharmaceutical industry is strictly regulated to maintain trust in the safety and efficacy of the products. In the last decade quality assurance and Quality by Design (QbD) are preferred over the quality testing methods. But the new approaches require fast, preferably solid-state analytics, capable of qualitative and quantitative analysis of Critical Quality Attributes (CQAs) even in complex, multicomponent processes. Because of these changes, non-destructive instrumental analytical methods are developing. These methods require complicated data analysis to achieve reliable results and to separate the effects of different CQAs on the product. Due to the fast and non-destructive sample preparation larger part of the product can be analyzed, which results in a more confident quality assurance, therefore the new approach can outperform the traditional quality testing.
Process Analytical Technology
Process Analytical Technology (PAT) is the use of real time, in-line applicable process analytical tools to achieve better process understanding. During our research, we have implemented vibrational spectroscopic methods, machine vision and model-based soft sensor approaches during numerous manufacturing steps, from powder blending to electrospinning with real time chemometric evaluations. Process monitoring is crucial in continuous manufacturing to notice disturbances during production. Furthermore, these monitoring applications helps the Quality by Design (QbD) approach, fast optimization, and robust process control.
Raman, FTIR, NIR spectroscopy
The introduction of the PAT guideline gave a boost to the vibrational spectroscopic methods. Numerous applications of in-line FTIR- and NIR-spectroscopic monitoring have been published, and Raman-spectroscopy developed into a reliable and affordable solid-state analyzer. Each technique has advantages and strengths, and the feasibility of the approaches are system dependent.
Chemometrics
Complicated statistical evaluation is necessary to extract chemical information of the aquired spectra. Traditionally, calibration-based univariate data analysis can be used, if the change in the data is obvious, i.e. when a peak intensity is proportional with the chemical information. Advanced, multivariate statistical methods can extract not so obvious changes, and the result has lower dependency on between spectra deviation.
Machine vision
With the integration of cameras into pharmaceutical processes, invaluable data can be obtained regarding several critical process parameters. By equipping these systems with the most up-to-date, AI-based algorithms, image analysis may be used where it was previously unimaginable.