(277f) Sensitivity Analysis and Optimization Study of Pharmaceutical Crystallization Process

Crystallization is a prevalently applied technique in many industries including food, bulk and fine chemicals as well as manufacturing of active pharmaceutical ingredients (APIs). In the pharmaceutical industry, crystallization processes are employed mainly for separation,, purification as well as for formulation of APIs. Moreover, the process should guarantee the performance of the delivered APIs suitable for downstream processing, handling and drug efficacy.

The performance of an API crystal is strongly affected by the product quality attributes such as crystal size distribution, morphology, shape and purity, which are the main targets of the process output to be achieved within specific range. However, in a crystallization process, complex feedback interactions take place between the formed crystals and the growth and nucleation kinetics. These are at the same time dependent on the supersaturation and hydrodynamic conditions to which the crystals are subjected. Consequently, the quality characteristics of the crystal product have a complex dependency on multiple operational variables and input parameters.

In order to evaluate and quantify which process factors or variables have the most impact on the final quality attributes of the crystal product, a comprehensive sensitivity analysis of the system model is required. This study therefore focuses on a global sensitivity analysis of a predictive pharmaceutical crystallization model through assessing the sources of uncertainty and their influence on crystal size distribution as the measured process outcome/performance criteria in order to guide design and suggest the optimized operation conditions.

This process model of pharmaceutical batch cooling crystallization is developed based on compartmentalization technique which takes into account mixing phenomena as well. Several global sensitivity analysis methods including Morris screening, global derivative based sensitivity analysis (DGSM), variance decomposition methods such as Sobol’s index and linear regression of Monte Carlo outputs are used to perform the sensitivity analysis. The developed crystallization process model is analyzed for a range of values for parameters like cooling rate, growth and nucleation kinetics etc. to rank their contributions on the measured outcome of the system “crystal size distribution”. Comparison of two sensitivity analysis methods and evaluation of the results are used as a guide for process optimization and robust system design.

Acknowledgment 

We would like to thank the Danish Council for Independent Research (DFF) for financing the project with grant ID: DFF-6111600077B.