(417b) Applications of Quality by Design Principles Utilizing Process Modeling and PAT in the Development of An API Process

Authors: 
LaPack, M. A., Eli Lilly and Company
Martinelli, J. R., Eli Lilly and Company


Process analytical technology (PAT) can be used as an online control tool to determine a process endpoint or to change a setpoint if installed as part of a feedback control loop. During the development of a process, it can be used to better elucidate the governing physics of a process, allowing for an understanding of the mechanism of the process. In the later example, this can lead to the development of process models and allows for the development of a process using the principle of ?Quality by Design?.

In this presentation, PAT is used in the development of a first principles process model. The use of the process model will be presented in the context of developing the synthetic process to make semagacestat, a compound currently in Phase III clinical trials for the treatment of Alzheimer's Disease.

The penultimate synthetic step in the semagacestat route involves the removal of a tert-butyloxycarbonyl (BOC) protecting group followed by the coupling of a carboxylic acid to the deprotected amine. During the deprotection, carbon dioxide (CO2) is evolved. In the subsequent coupling reaction, residual CO2 reacts with the deprotected amine, forming a urea dimer, if the CO2 has not been sufficiently purged. Control of the urea dimer impurity is accomplished by the thorough removal of CO2 prior to the initiation of the coupling reaction. A mass transfer model for the partitioning of CO2 in the reaction matrix and the reactor headspace was derived to predict the concentration of CO2 in the reaction matrix, as it not measurable at the low levels needed for suitable control of the impurity. The model requires a mass transfer coefficient and the Henry's law partitioning coefficient as inputs. These parameters were determined using online mid-IR, to measure CO2 in solution (at levels above that needed for control) and mass spectrometry to measure CO2 in the headspace. The model was used for optimize the processing parameters needed to purge residual CO2 prior to the onset of the coupling reaction. The model allowed for an understanding of the effect of scale-up to commercial scale on the purge process parameters, eliminating the risk of failure with scale-up. The model was verified at production scale, by monitoring the concentration of CO2 in solution by ATR IR and in the headspace by GC mass spec, resulting in excellent agreement between the measured results and expectations. The subsequent product exhibited no measurable amount of the urea impurity.