Toward Reinforcement Learning–Enabled Model-Informed Precision Dosing: Concepts, Applications, and Implementation

Purpose: Reinforcement learning is suited to sequential decision making in clinical settings, where treatment strategies are adjusted based on patient response. This aligns with therapeutic drug monitoring and model-informed precision dosing, in which dosage is iteratively refined as conditions evolve. Thus, reinforcement learning is a potential dosing optimization framework. This review provides a structured and clinically interpretable overview of these applications and their implementation. Methods: A narrative literature review of studies that applied reinforcement learning to dosing optimization was conducted, using database searches and manual screening. Studies were qualitatively analyzed for drug type, algorithm, training environment, and their state, action, and reward definitions. A simplified implementation example using a deep Q-network illustrates the core workflow. Results: Reinforcement learning has been applied across diverse clinical settings, including anesthesia, oncology, and critical care. Most studies used simulation-based environments (pharmacokinetic, pharmacodynamic, or mathematical models) to enable safe, scalable training. Off-policy methods (eg, Q-learning and deep Q-networks) were commonly used for discrete dosing decisions and retrospective data; on-policy methods (eg, actor–critic approaches) were applied when adaptive dose adjustment was required. Substantial variability was observed in algorithm selection and in state, action, and reward definitions. The simplified example illustrates how deep Q-network–based reinforcement learning can be integrated with pharmacokinetic models to support dose optimization. Conclusions: Reinforcement learning provides a flexible framework that complements pharmacometric models to support dosing optimization. Further research is required to develop best-practice recommendations for study design, algorithm selection, reward design, and model evaluation to facilitate translation into clinical practice.