(243a) Use of Bayesian Modeling for Failure Risk Analysis and Control Strategy Design
AIChE Annual Meeting
2021
2021 Annual Meeting
Pharmaceutical Discovery, Development and Manufacturing Forum
Control Strategies in Pharmaceutical Development and Manufacturing I
Tuesday, November 9, 2021 - 8:00am to 8:24am
Process characterization and definition of appropriate control strategies is critical to ensure quality in the
manufacture of active pharmaceutical ingredients. According to the principles of quality by design (QbD),
process development should include identification of critical quality attributes (CQAs), characterization to
understand the relationship between process parameters and these CQAs, and an evaluation of risk for the
process. This presentation provides a case study wherein Bayesian probabilistic modeling was used to evaluate
process risk for a specific critical impurity. Formation and purging of this impurity was studied via DoE and
mechanistic studies spanning the proven acceptable ranges (PAR) for reaction and crystallization parameters.
Probabilistic models were developed to predict process performance and assess and visualize predicted failure
rates across the multivariate parameter space. This analysis informed assignment of critical process parameters
and appropriate multivariate PARs to ensure robust processing. The probabilistic models also enabled
estimation of the impact of common cause plant variability on the anticipated failure rate against specification
levels of the critical impurity. Overall, these various failure rate analyses, enabled by the Bayesian models,
helped to ensure implementation of an effective control strategy and will aid in future process transfers and
process monitoring upon commercialization.
manufacture of active pharmaceutical ingredients. According to the principles of quality by design (QbD),
process development should include identification of critical quality attributes (CQAs), characterization to
understand the relationship between process parameters and these CQAs, and an evaluation of risk for the
process. This presentation provides a case study wherein Bayesian probabilistic modeling was used to evaluate
process risk for a specific critical impurity. Formation and purging of this impurity was studied via DoE and
mechanistic studies spanning the proven acceptable ranges (PAR) for reaction and crystallization parameters.
Probabilistic models were developed to predict process performance and assess and visualize predicted failure
rates across the multivariate parameter space. This analysis informed assignment of critical process parameters
and appropriate multivariate PARs to ensure robust processing. The probabilistic models also enabled
estimation of the impact of common cause plant variability on the anticipated failure rate against specification
levels of the critical impurity. Overall, these various failure rate analyses, enabled by the Bayesian models,
helped to ensure implementation of an effective control strategy and will aid in future process transfers and
process monitoring upon commercialization.