Optimal EPO dosing in hemodialysis patients using a non-linear model predictive control approach

Area of Science:

• Biomedical Engineering
• Control Systems
• Nephrology

Background:

• Anemia management in hemodialysis patients is challenging due to variable responses to erythropoiesis-stimulating agents (ESAs).
• Achieving and maintaining target hemoglobin (Hgb) levels is difficult in clinical practice, often leading to suboptimal treatment outcomes and potential side effects from high drug doses.

Purpose of the Study:

• To develop a fully personalizable controller scheme for stabilizing Hgb levels within a narrow target window in hemodialysis patients.
• To minimize drug doses and mitigate potential side effects through individualized treatment optimization.
• To present the initial in-silico results of this novel framework.

Main Methods:

• Formulation of a non-linear model predictive control (NMPC) algorithm based on an erythropoiesis model for individualized epoetin alfa (EPO) dose optimization.
• Estimation of individual patient model parameters using clinical data.
• Solving the optimal control problem for continuous drug administration using a projected quasi-Newton method.

Main Results:

• Successful in-silico testing of the NMPC controller on multiple patient parameter sets.
• Demonstration of feasible and appropriate control under assumptions of regular Hgb measurement and adjusted continuous EPO administration.
• Satisfactory handling of challenging scenarios including bleedings, missed administrations, and dosing errors in simulations.

Conclusions:

• The developed NMPC framework offers a promising approach for personalized anemia management in hemodialysis patients.
• Continuous, individualized EPO administration, optimized by the NMPC controller, can effectively stabilize Hgb levels and manage treatment variability.
• The controller's robustness in simulations suggests its potential for improving clinical outcomes and patient safety.