Notes
The preparation of pellets is a complex technological procedure, which requires a large number of experiments for a thorough evaluation. This doctoral thesis provides additional insights into all process steps involved in the preparation of pellets using a range of different approaches. The first chapter is devoted to the retrospective applications of multivariate data analysis (MVDA) and artificial neural networks (ANN), with the aim of identifying critical material attributes (CMAs) and critical process parameters (CPPs) for drug dissolution as the critical quality attribute (CQA) of pellets with a high content of active pharmaceutical ingredient (API). The development of MVDA and ANN model is presented, describing the variability between 49 industrially manufactured batches. Based on the generated models, CMAs and CPPs were successfully identified. The statistical techniques utilize a different approach to data modeling, ANN utilizes nonlinear correlations between factors and responses, whereas MVDA is based on linear correlations. Consequently, we identified other CMAs and CPPs with the ANN model, such as API particle size, which is in line with expectations, as the investigated formulation contains a BCS II class API. Despite the difference in approach to data evaluation, the complementary relationship of MVDA and ANN was elucidated; both play an important role in facilitating a higher level of process understanding, creating opportunities for root cause investigation, and developing control strategies for pharmaceutical products. To the best of our knowledge, this is the first time that MVDA and ANN were used to process such a broad range of data for the evaluation and optimization of pellets, prepared by extrusion and spheronization on an industrial scale. The utilized Quality by design (QbD) approach governed by data modeling is easily transferable also to other pharmaceutical products. The second chapter presents the benefits of using visual analysis for an in-depth understanding of the processes involved in the example of QbD-guided development of pellets with high API content and three coatings. The benefits of PATVIS APA as one of the dynamic image analysis techniques are showcased: - obtaining target or desired pellet size in the process of extrusion and spheronization - at-line yield evaluation of the extrusion and spheronization process - monitoring the effectiveness of the coating process, detection of potential pellet agglomeration monitoring of coating thickness during the coating process with the option of end point determination. Utilization of dynamic image analysis facilitates real-time adaptation of process parameters, which is an added benefit, as it directly influences the quality of the product and efficiency of the process. The third chapter continues with the exploration of the coating process, with a focus on the spraying process, which influences the quality of the film coating. To increase the chance of success when scaling up from laboratory to production scale, the influence of dispersion properties (viscosity, surface tension, and density) and process parameters (dispersion flow rate, atomization air pressure, and microclimate air pressure) on the average droplet size and their distribution by formation with a three-channel nozzle was investigated. Based on a custom-built optical method concept, which enables the monitoring of the speed, direction, size, and distribution of sizes of droplets, two semi-empirical models were built to predict the average droplet size and their distribution. Based on the models constructed, comparable droplet sizes can be achieved on both scales by adjusting the process parameters. To our knowledge, these are the first model built for prediction of droplet size and their distribution, which include Newtonian and non-Newtonian dispersions and simultaneously include also variation of process parameters. The models constructed can be transferred to all coating processes utilizing the investigated three-channel nozzle. Moreover, the significant advantage of this work is, that the approach for construction of the model can be extended to all processes of droplet formation. Therefore, it can also be applied to spray drying, granulation, and coating of other pharmaceutical products. The most important contribution of this doctoral thesis are the insights into new and innovative approaches for the QbD development of pellets. What is more, the approaches presented can be transferred to other pharmaceutical products. As showcased, the use of DoE is more appropriate at the early stages of product development on a laboratory scale, where experiments can be planned in a systematic way (for example, the evaluation of droplet size). By contrast, the use of statistical techniques, such as MVDA and ANN is more appropriate for the evaluation of large data sets obtained on regular production batches during the lifecycle of the product and they can be used as excellent root cause investigation tools. Like the dynamic imaging method presented (PATVIS APA), it can be used as a powerful tool at both stages: during early development to aid process development and understanding as well as during the product lifecycle, as a potential process analytical tool (PAT) tool.