(478g) Integrated Filtration and Washing Modelling to Predict Continuous Isolation Performance

The pharmaceutical industry is starting to adopt continuous active pharmaceutical ingredient (API) manufacturing in order to reduce production costs, improve manufacturing flexibility, reduce infrastructure costs, reduce manufacturing lead time (from typically 6 months to 10 days) and to improve sustainability. A further driver is reduction of variance in API quality critical attributes. To facilitate the complete transition from batch to continuous manufacturing it is necessary to “smartly” integrate single continuous unit operations to achieve a continuous material flow from synthesis to formulation. To achieve this smart integration of unit operations, a combination of modelling, online measurement and advanced control techniques are vital to predict product property outcomes, to monitor and control processes and to reduce the risk of non-conforming products.

Another challenge the pharmaceutical industry faces is to reduce the quantity of material consumed during process development. A stretch goal is to consume just 100g of API (and the corresponding precursors) and to complete development in 100 days. Digital design of continuous API manufacturing offers a path to achieving this goal. This includes modelling and predicting process performance as a function of the operating conditions for both individual continuous unit operations and for the integrated processes with the aim of optimizing process design and reducing the laboratory time and cost needed to develop new products. Whilst a few examples of modelling integrated continuous unit operations using flowsheet models, have been published, these are mainly focused on secondary drug product manufacture rather than API synthesis and isolation.

The team in the Continuous Manufacturing and Advanced Crystallization (CMAC) Future Manufacturing Research Hub have addressed this gap by modelling primary drug manufacturing using an integrated unit operations approach. A digital tool capable of transferring material property information between operations to predict the product attributes in integrated synthesis and purification processes has been developed. The focus of the work reported here combines filtration and washing operations used in API purification and isolation by combining predicted and experimental data generated during upstream crystallization process. The model approach described is subdivided into two levels of increasing system complexity where the filtration model approach is held constant, while different washing modelling approaches are used for the two levels in order to describe different washing scenarios. The filtration model approach considers the case where the different components of the input suspension are separated between the retained filter cake with residual impure mother liquor and the impure liquid filtrate removed during filtration. The assumptions considered for the washing modelling described in level 1 are that the impure mother liquor is fully miscible with the wash solvent selected and no changes in solid phase are considered (no particle dissolution or growth). Level 1 thus can describe three different washing mechanisms: pure displacement (level 1a), dilution with perfect liquid mixing (level 1b) and dilution with axial dispersion (level 1c). The more complex level 2 allows, additionally, the possibility of cake and impurity dissolution or precipitation during washing. The level 2 washing model considers cake dissolution due to solubility variation during washing, the possibility of API or impurity precipitation from solution due to a drastic solubility drop (the antisolvent effect occurring during washing) as an instantaneous process. This level presumes that kinetic aspects can be neglected and equilibrium is reached instantaneously. In detail; the level 2a model describes a washing process where the system is assumed to be homogeneous and the same solvent composition is assumed for the entire cake. In level 2b, a more realistic washing mechanism is modelled in which the system shows vertical heterogeneity and there is a composition gradient along the cake height, which is described in the model

The integrated modelling tool uses information on the product crystal suspension characteristics predicted using gPROMS/Matlab to predict filtration time, filtrate flow rate and the composition of the filter cake and filtrate generated during filtration. The washing of the wet filtered cake is then simulated to predict; washing efficiency and to generate washing curves, cake and filtrate composition, indicate the probability of particle size variation caused by cake dissolution, and residual cake moisture content and composition. To validate the scenarios described using the integrated models a design of experiments (DoE) approach was used and two experimental case studies were investigated using mefenamic acid and paracetamol as representative isolation processes. In each case the objective of the work was to meet a product purity specification and minimize changes to the crystalline particle attributes occurring during the isolation process. The validation is also useful to identify which model level best reflects the experimental data, with the aim of identifying the minimum criteria a filtration and washing model needs to have to simulate experimentally verified isolation outcomes.