A comparative study of two data-driven modeling approaches to predict drug release from ER matrix tablets

Area of Science:

• Pharmaceutical Sciences
• Drug Delivery Systems
• Computational Chemistry

Background:

• Extended release (ER) oral formulations offer therapeutic benefits but face challenges in achieving consistent drug release.
• In vitro dissolution testing is crucial but time-consuming for evaluating ER formulations.
• Quality by Design (QbD) principles are increasingly adopted in pharmaceutical development.

Purpose of the Study:

• To develop and compare predictive models for drug dissolution in polyethylene oxide (PEO)-based ER oral hydrophilic matrix tablets.
• To integrate data-driven modeling within a QbD framework for enhanced formulation development.
• To improve the accuracy and efficiency of predicting drug release profiles.

Main Methods:

• Model screening and comparison of machine learning (ML) models, specifically artificial neural networks (ANN).
• Application of functional data analysis (FDA) combined with design of experiments (DoE) for continuous dissolution curve modeling (FDOE).
• Analysis of a dataset of 91 ER matrix tablet formulations using training, validation, and test sets.

Main Results:

• Both ANN and FDOE models demonstrated high similarity to experimental dissolution profiles (f2 values 48-88 for FDOE, 52-88 for ANN).
• The study successfully validated the predictive capabilities of the developed modeling approaches.
• Data-driven techniques were shown to be effective in characterizing and predicting dissolution behavior.

Conclusions:

• Advanced data-driven modeling techniques, including ANN and FDOE, can significantly enhance dissolution prediction accuracy for ER oral formulations.
• Integrating these modeling approaches into QbD-based development streamlines the formulation process, reducing development time and costs.
• This work supports the adoption of computational tools for more efficient pharmaceutical product development.