Data-driven tiered procedure for enhancing yield in drug product manufacturing

• Generic, data-driven yield enhancement procedure for drug product manufacturing.
• A four-step procedure, combining heuristics and statistics, is presented.
• Multivariate and Descriptive Statistical Modelling is applied as a core technique.
• Process changes are to be based on validated loss cause models.
• Industrial case study resulted in threefold decrease of an investigated loss cause.

Enhancing efficiency of pharmaceutical batch production processes is an important challenge in times of increasing public pressure on healthcare costs and decreasing research productivity. This study presents a data-based procedure for systematic yield enhancements in drug product manufacturing, based on four steps. On the first step, production is reviewed to select relevant loss causes, which are assessed on the second step deductively with the goal of assigning measurable parameters. Descriptive Statistical Modelling of loss causes is then performed on the third step, enabling model-based enhancements of processes on the fourth step or, if necessary, a loop-back review of a given loss cause.An industrial case study was performed on production data of 88 batches and demonstrated the applicability of the procedure by prioritizing relevant loss causes, reducing required sample quantities by up to 8% and a cosmetic defect by about 70% by a process change.