(625g) Surrogate Modeling of Dissolution Behavior Toward Efficient Design of Tablet Manufacturing Processes

Speed is critical in the research and development of drug products and processes. The challenge here is the existence of various process alternatives and product characteristics in tablet manufacturing processes, which makes the decision-makings complicated. As for the process alternatives, the choice of dry or wet granulation is known to affect product quality significantly, while this decision is usually conducted without rigorous comparison. As for the product characteristics, solubility of active pharmaceutical ingredients (APIs) depends on the product, and influences the dissolution behavior of tablets largely. In the literature, first-principle as well as statistical modeling approaches has been researched for tablet manufacturing (e.g., [1] and [2]) for assisting more efficient process design. The power of the first-principle modeling is recognized so well that commercial process simulators, e.g, gPROMS, are available for tablet manufacturing processes. However, there is still a hurdle of computational time to use first-principle models in the actual decision-makings.

This work presents surrogate modeling of dissolution behavior for enhancing the use of first-principle modeling in the design of tablet manufacturing processes. Our research approach consists of the following three steps: (i) collect process input and output data using gPROMS, (ii) perform surrogate modeling, and (iii) test the developed model. In the first step, 200 data are collected per each manufacturing method by simulating flowsheet models. Secondly, random forests are used for developing surrogate models from collected data sets after fitting Weibull models for dissolution behaviors. Finally, in the third step, sensitivity analysis is performed using the developed surrogate model to generate a proposal for the efficient process design.

Case studies of dry granulation and wet granulation methods were performed, where paracetamol was used as an API of the tablets. The simulation results showed the differences in tablet properties between manufacturing methods, e.g., hardness, porosity, and dissolution. After the execution of surrogate modeling, the sensitivity of input parameters was assessed using rank correlation coefficients, which determined critical parameters in the process, and product design phases. In ongoing work, the generality and the effectiveness of the developed models are discussed by performing case studies using a different API.

[1] D. Van Hauwermeiren, M. Verstraeten, P. Doshi, M. T. am Ende, N. Turnbull, K. Lee, T. De Beer, I. Nopens, Powder Technol, 341, 116–125 (2019)

[2] K. Matsunami, T. Nagato, K. Hasegawa, H. Sugiyama, Int J Pharm, 579, 119160 (2020)