A continued learning approach for model-informed precision dosing: Updating models in clinical practice.

Area of Science:

• Pharmacometrics
• Clinical Pharmacology
• Computational Biology

Background:

• Model-informed precision dosing (MIPD) uses prior knowledge and therapeutic drug monitoring (TDM) for individualized dosing.
• Current MIPD models often rely on limited clinical trial data, potentially not reflecting real-world patient variability.
• Adapting models to diverse patient populations in clinical practice is crucial for effective MIPD.

Purpose of the Study:

• To develop a method for continued learning across patients within MIPD.
• To refine model parameters to better represent real-world patient diversity and variability.
• To enable adaptation of MIPD models using summary data, facilitating cross-center learning.

Main Methods:

• A sequential hierarchical Bayesian framework was proposed for continued learning in MIPD.
• The approach separates individual patient parameter updates from population parameter updates.
• Illustrative application using neutrophil-guided paclitaxel dosing.

Main Results:

• The proposed method allows for ongoing refinement of MIPD models as more patient data becomes available.
• Separating parameter updates facilitates knowledge sharing across different clinical settings without compromising individual patient data privacy.
• Demonstrated successful adaptation of models for paclitaxel dosing.

Conclusions:

• The developed approach enhances the adaptability of MIPD to diverse patient populations encountered in clinical practice.
• Enables continuous learning and model improvement across multiple hospitals or study centers.
• Represents a significant step towards increased confidence and broader adoption of MIPD in routine therapy.