(754b) Improved Process Understanding Through Implementation of QbD Methodologies | AIChE

(754b) Improved Process Understanding Through Implementation of QbD Methodologies


Grosser, S. T. - Presenter, Merck & Co. Inc.
Moment, A. - Presenter, Merck & Co., Inc.
Spartalis, A. - Presenter, Merck & Co., Inc.
Zhou, G. - Presenter, Merck & Co., Inc.
Fernandez, P. - Presenter, Merck & Co., Inc.

Purpose: The work presented here was conducted to develop a process for the pure step of an ongoing developmental program that simultaneously provides operational flexibility and sufficient process robustness inline with QbD filing strategies. Particular challenges of this process are the strong dependence of final API PSD on seed temperature and the significant impact that solvent composition has on that temperature. Additional steps have been investigated to limit the variability of residual solvents going into the pure step and therefore simplify the seed point determination. Unfortunately these steps have both yield and time-cycle implications and could unnecessarily impact productivity for batches when residual solvent levels may already be low. As a result, the project team has decided to develop a mechanism by which the optimal seed temperature is determined in real-time based upon batch composition assays, therefore enabling increased flexibility in solvent composition without comprising batch quality.

Methods: This work will demonstrate the application of risk assessment strategies for targeted experimental planning and implementation of statistically designed experiments (DoE) for investigation of interacting process parameters. We will also discuss the use of pilot-scale confirmatory runs of DoE results and how large scale experimentation was used to drive further DoE investigation.

We will present the use of PAT technologies for the rapid collection of solubility data to aid in mapping API solubility as a function of solvent composition for a 4-solvent system.

Lastly, we will discuss the use of thermodynamic and empirical modeling tools for the translation of the acquired solubility map into operational parameter ranges and assay specifications to be implemented in the full-scale demonstration and subsequent filing.

Results: Guided by a formal risk assessment, significant DoE experimentation was performed aimed at understanding the effect that 9 operational parameters have upon the API PSD, purity and form. None of the investigated parameters showed evidence of a purity or form impact while PSD was primarily impacted by seed temperature. Although the PSD was not directly influenced by solvent composition considerations, it was readily observable that solvent composition had a profound impact on the optimal seed temperature so by extension, needed to be considered to ensure API PSD within specification.

To facilitate understanding of the relationship between solvent composition and optimal seed temperature, the crystallization platform was employed to construct an API solubility map across the 4- dimensional solvent composition space. With the use of the crystallization platform, the collection of solubility as a function of temperature was automated for each solvent composition. The use of PAT technologies helped to attain the desired data while avoiding the complication of offline sampling of a solution susceptible of crystallization upon cooling, thus greatly reduced the time and effort required to construct this solubility map.

With the relationship between solvent composition and API solubility measured at discrete solvent compositions, the next step was to convert that map into a model that allowed investigation of the continuum of solvent compositions studied. Two strategies were investigated. In the first strategy, a thermodynamic model was used to predict the API solubility. The solubility model employed was based upon the Unifac activity coefficient model and the data obtained through the PAT investigation was used to regress the binary interaction parameters between the API and the solvent subgroups. Secondly, an empirical (polynomial with three variables) model was also developed and found to offer only slightly less accurate predictions of API solubility. Due to the expected ease of implementation on the factory floor and within the filing, the decision was made to utilize the polynomial model long term. With a firm understanding of the relationship between composition and API solubility and an eye to expected process variabilities, the allowable seed temperature range could be determined and ranges on the allowable solvent levels could be specified.

Conclusions: Through efficient application of QbD methodologies, the extended development team has established a process that enables significant operational flexibility and process robustness for an otherwise challenging seed point determination. Although the seed temperature is critical in ensuring the final API PSD, a procedure has been put in place to ensure an appropriate seed temperature while simultaneously leveraging the flexibility in solvent composition afforded through QbD.