(417a) Model-Based Quality by Design Applied to An API Process | AIChE

(417a) Model-Based Quality by Design Applied to An API Process


Burt, J. L. - Presenter, Bristol-Myers Squibb
Bergum, J. - Presenter, Bristol-Myers Squibb Company
Braem, A. - Presenter, Bristol-Myers Squibb
Ramirez, A. - Presenter, Bristol-Myers Squibb
Tabora, J. - Presenter, Bristol-Myers Squibb Company
Tummala, S. - Presenter, Bristol-Myers Squibb

In this case study, an API process was investigated within the Quality by Design framework.  Employing a multi-staged experimental design as the basis of investigation, process chemistry models were developed using both mechanistic and empirical approaches.  Statistical analysis of crystallization purging data facilitated the establishment of in-process limits for key impurities, and our chemical models were employed to generate design spaces based on these specifications.  The predictive capacity of the models was explored both on the laboratory scale, and via comparison with production-scale batches.  By employing a scientific and risk-based approach, an API design space was defined that ensures product quality to a high degree.

In broad terms, this API process can be considered as three steps: (1) chemical dehydration, (2) a sequence of reversible reactions that affords the API, and (3) reaction quench and isolation.  Reaction modeling focused on the conversion of Intermediate Z, along with the formation of Impurities A and B arising from API degradation.  The presence of Impurity C in the starting material introduces further complexity, since it is transformed to Impurities A, B, and D in proportions that vary as a function of API processing conditions.  Concentrations of Intermediate Z and Impurity A must be controlled to set specifications in the isolated product, whereas Impurities B and D purge near-quantitatively.

Points that will be highlighted include: (1) comparison of the design spaces generated by the mechanistic and empirical models; (2) key insight gained via the mechanistic approach; (3) impact of upstream process challenges on API design space definition; (4) opportunities for combined empirical and mechanistic modeling approaches.