We describe the development of an in-line monitoring approach for the fluid-bed drying and coating steps for the production of enteric coated pellets by NIR. Our results show that key pellet characteristics can be monitored in-line. Likewise, the finished product acidic resistance is in excellent agreement to the in-line NIR predictions. Samples were collected at regular intervals and analyzed by several reference methods to characterize both process steps. In-line NIR models for pellets size sieve fractions, residual solvent content, and amount of coating layer have been constructed. Both the pellet coating layer amount and the in-vitro enteric performance demonstrate low variability which represents a challenge to the usual chemometric model development approach. To overcome this challenge a hierarchical PLS model for predicting acidic resistance was successfully constructed using time-evolving spectral data from 22 batches. Moreover, a novel multivariate meta-analysis of the PLS loadings of individual in-line models and the hierarchical PLS model has identified which pellet characteristics correlate most significantly with the observed enteric performance of the finished product. Additionally, the meta-analysis pointed toward the presence of further mechanisms unrelated to studied characteristics that also significantly influence the acidic resistance.
