(299d) Estimability Analysis for Improved Parameter Estimation in Deterministic Models: Pharmaceutical Case Studies | AIChE

(299d) Estimability Analysis for Improved Parameter Estimation in Deterministic Models: Pharmaceutical Case Studies

Authors 

Sen, M. - Presenter, Eli Lilly and Company
Garcia-Munoz, S., Eli Lilly and Company
Tandogan Gray, N., Eli Lilly & Co
Borkar, I. V., Eli Lilly and Company
Kolis, S. P., Eli Lilly and Company
Wilson, T. M., Eli Lilly and Company
Buser, J. Y., Eli Lilly and Company
Alt, C. A., Eli Lilly and Company
Often in the early stages of model development, one needs to estimate the model parameters using data with insufficient information. This information insufficiency in the experimental data can arise from either a poor design, or lack of measurements from the system.

This work illustrates the application of Estimability Analysis to aid the parameter estimation of a deterministic mathematical model. This technique has proven useful when the information content of the data available is insufficient to obtain a good estimate of all the unknown model parameters. Estimability Analysis ranks the parameters from the most to least estimable; taking into account the uncertainties in both the initial guesses of the parameters and the data [1]. This ranking enables the modeler to determine which parameters to estimate, and which to fix. The obtained model parameters are a good initial guess to a formal model-based experimental design exercise to determine the necessary experiments to optimally parametrize the model.

Case studies are presented from drug substance process development where a kinetic model of a reactive system is being developed; and a third case study from the modeling for a continuous direct compression line for drug product.

References

[1] Wu et al., 2011, International Journal of Advanced Mechatronic Systems, 3, 188-197.