Machine Learning-Assisted Predictive Formulation Modeling for Segmented Combination Vaginal Rings: Construction and Validation

The vaginal ring represents an innovative sustained-release drug delivery system in the field of women’s reproductive health, demonstrating significant clinical value. However, its formulation development remains constrained by a traditional “trial-and-error” empirical approach, which often requires substantial cost investment for studies involving multiple factors and levels. To address evolving patient needs and diverse drug properties, this research investigates the application of machine learning algorithms in pharmaceutical development, with a specific focus on the development and optimization of vaginal ring formulations. A segmented combination(ATZ/LNG) IVR for the treatment of endometriosis is successfully fabricated and subjected to pre-formulation studies. Furthermore, the XGBoost machine learning model is selected and trained for the vaginal ring formulation. The feasibility of the developed machine learning-assisted predictive formulation framework was experimentally validated. In response to the high costs associated with the conventional trial-and-error method in formulation development, this study leveraged research on the complex formulation of the intravaginal ring (IVR) dosage form to develop an AI-based predictive model. It suggests the feasibility of AI-assisted pharmaceutical development and proposes a novel methodology for formulation research.