PROBABILISTIC MODELS FOR FORECASTING PROCESS ROBUSTNESS
José E. Tábora, Jacob W. Albrecht, Brendan C. Mack · 2019
Probabilistic graphical models (PGMs) are a powerful framework to define probabilistic analysis problems. This formalism allows for definition of a large variety of problem statements while also enabling rigorous solution methods for that problem. PGMs represent uncertain probabilities as oval nodes connected by arrows that represent causal influence. The key goal of pharmaceutical process development is to design processes that enable the safe, reliable, and cost-effective manufacture of the active pharmaceutical ingredient (drug substance). This chapter describes some of the statistical tools that are used to quantify process robustness. It provides a background on the regulatory expectations driving the process design and characterization of process robustness, and discusses a probability-based mathematical formalism to model and estimate robustness. The chapter also describes approaches to define and calculate these robustness metrics. Finally, it presents a case study to incorporate the concepts.