Development of Predictive Statistical Model for Gaining Valuable Insights in Pharmaceutical Product Recalls.

Area of Science:

• Pharmaceutical Science
• Artificial Intelligence
• Drug Development

Background:

• Pharmaceutical product recalls pose a significant global challenge, necessitating advanced tools for risk mitigation.
• Optimizing drug development processes is crucial to prevent recalls and ensure product quality.

Purpose of the Study:

• To employ AI and machine learning (ML) to analyze factors influencing pharmaceutical product recalls.
• To develop predictive models for assessing product complexity and forecasting recall likelihood.

Main Methods:

• Utilized FDAZilla and SafeRX tools to construct an open database model.
• Developed predictive statistical models using Multivariate Analysis and the Least Absolute Shrinkage and Selection Operator (LASSO) Approach.
• Analyzed key descriptors including delivery route, dosage form, dose, BCS classification, physicochemical properties, release type, half-life, and manufacturing complexity.

Main Results:

• Identified critical descriptors influencing product recall risk: BCS Class I, dose number, release profile, and drug half-life.
• Assigned risk numbers and computed cumulative risk numbers to assess product complexity and recall likelihood.
• The LASSO model achieved 71% accuracy in confirming key descriptors for recall risk prediction.

Conclusions:

• Presents a holistic AI and ML approach for evaluating and forecasting pharmaceutical product recalls.
• Highlights the importance of formulation complexity and manufacturing processes in mitigating product quality risks.
• Emphasizes the role of key descriptors in predicting and preventing pharmaceutical product recalls.