Abstract
This is Part II of a sequential review on Twin-screw extrusion (TSE). Whereas Part I addresses twin-screw granulation, this review focuses on hot-melt extrusion (HME). Specifically, it examines HME for producing amorphous solid dispersions (ASDs) to enhance the bioavailability of poorly water-soluble drugs, and focuses on the shift from empirical development to model-based, rational ASD design. This shift is framed through a multi-scale computational framework. The framework integrates molecular dynamics (MD) for drug-polymer miscibility prediction with computational fluid dynamics (CFD) and population balance models (PBM) for process digital twins. It also incorporates physiologically based pharmacokinetic (PBPK) modelling to predict in vivo performance. As a manufacturing platform, HME is solvent-free, continuous, and scalable. However, successful application requires tight control of thermal, shear, and residence-time profiles within the thermo-mechanical limits of the API-polymer system. Rational selection of polymers, surfactants, and plasticisers is therefore central to balancing processability, long-term physical stability, and recrystallization resistance. Moreover, this selection should be guided by miscibility tools and glass-transition criteria. Case studies on solubility enhancement, taste masking, controlled or modified release, and 3D-printed dosage forms confirm the broad capability of HME. They also reveal trade-offs in drug loading, extrudate mechanical robustness, and compatibility with downstream unit operations. From a modelling perspective, strategic coupling across scales can enhance the predictive power of integrated in silico frameworks. However, this potential depends on rigorous validation of cross-scale model interfaces. It also requires high-fidelity standardised data and lower computational barriers for high-throughput application. Ultimately, this review highlights the power of integrating artificial intelligence (AI) and machine learning (ML) with mechanistic modelling. It further argues that closed-loop AI systems and robust digital twins are key enablers of efficient, well-controlled manufacturing of advanced therapeutics.
[Display omitted]