An insight into predictive parameters of tablet capping by machine learning and multivariate tools.

Capping is the frequently observed mechanical defect in tablets arising from the sub-optimal selection of the formulation composition and their robustness of response toward process parameters. Hence, overcoming capping propensity based on the understanding of suitable process and material parameters is of utmost importance to expedite drug product development. In the present work, 26 diverse formulations were characterized at commercial tableting condition to identify key tablet properties influencing capping propensity, and a predictive model based on threshold properties was established using machine learning and multivariate tools. It was found that both the compaction parameters (i.e., compaction pressure, radial stress transmission characteristics, and Poisson's ratio), and the material properties, (i.e., brittleness, yield strength, particle bonding strength and elastic recovery) strongly dictate the capping propensity of a tablet. In addition, ratio of elastic modulus in the orthogonal direction in a tablet and its variation with porosity were notable quantitative metrics of capping occurrence.