LAUSR

Abstract Material properties are known to have a significant impact on pharmaceutical manufacturing performance, particularly for solid product processes. Evaluating the performance of a specific material, for example an active pharmaceutical ingredient or excipient, is critical during development stages in order to determine the impact of material properties on the process. However, materials may be scarce during the early stages of process development due to high cost, unavailability, import restrictions, etc. Furthermore, research on particular active pharmaceutical ingredients may be difficult given unknown exposure limits, which may delay process development and technology transfer. The purpose of this work was to establish a methodology for finding materials with similar behavior during processing using material property measurements so that a surrogate may be found and may replace the scarce material during process development. This work presents several commercially available material property tests and emphasizes the benefits of compiling material property measurements into libraries. Twenty pharmaceutically relevant materials and seven different powder characterization tests were considered as a case study. A total of 32 measurements were collected for each of the 20 materials, leading to a dataset of 640 measurements. The material property library was utilized to find similarities between materials using two multivariate methods: principal component analysis (PCA) and hierarchical clustering. The similarities between materials were evaluated with the performance of materials on powder feeding, refilling, and continuous blending equipment. Material clusters showed similar behavior in the characterized equipment. Moreover, results from the PCA and clustering analysis were further used to evaluate the level of collinearity and similarity between characterization measurements that can be further investigated to reduce the number of measurements that need to be collected. Material property measurement clusters were established based on the collinearity of the metrics.