An interaction-based mixing model for predicting porosity and tensile strength of directly compressed ternary blends of pharmaceutical powders

Predicting the mechanical properties of powder mixtures without extensive experimentation is important for model driven design in solid dosage form manufacture. Here, a new binary interaction-based model is proposed for predicting the compressibility and compactability of directly compressed pharmaceutical powder mixtures based on the mixture composition. The model is validated using blends of MCC, lactose and paracetamol or ibuprofen. Both compressibility and compactability profiles are predicted well for a variety of blend compositions of ternary mixtures for the two formulations. The model performs well over a wide range of compositions for both blends and better than either an ideal mixing model or a ternary interaction model. A design of experiments which reduces the amount of API required for fitting the model parameters for a new formulation is proposed to reduce amount of API required. The design requires only three blends containing API. The model gives similar performance to the well-known Reynolds et al. model (2017) when trained using the same data sets. The binary interaction model approach is generalizable to other powder mixture properties. The model presented in this work is limited to curve-fitting of empirical compaction models for mixtures of common pharmaceutical powders and is not intended to provide guidance on the practical operating space (or design space) limits.