Near-Infrared Spectroscopy to Determine Residual Moisture in Freeze-Dried Products: Model Generation by Statistical Design of Experiments.

Abstract

Moisture content (MC) is a critical quality attribute of lyophilized biopharmaceuticals and can be determined by near-infrared (NIR) spectroscopy as nondestructive alternative to Karl-Fischer titration. In this study, we create NIR models to determine MC in mAb lyophilisates by use of statistical design of experiments (DoE) and multivariate data analysis. We varied the composition of the formulation as well as lyophilization parameters covering a large range of representative conditions, which is commonly referred to as "robustness testing" according to quality-by-design concepts. We applied principles of chemometrics with partial least squares and principal component analysis. The NIR model excluded samples with complete collapse and MC > 6%. The 2 main components in the principal component analysis were MC (91%) and protein:sugar ratio (6%). The third component amounted to only 3% and remained unspecified but may include variations in process parameters and cake structure. In contrast to traditional approaches for NIR model creation, the DoE-based model can be used to monitor MC during drug product development work including scale-up, and transfer without the need to update the NIR model if protein:sugar ratio and MC stays within the tested limits and cake structure remains macroscopically intact. The use of the DoE approach and multivariate data analysis ensures product consistency and improves understanding of the manufacturing process.