Model Predictive Control for Optimal Oral Anticoagulant Drug Administration

n this article an MPC algorithm was proposed for the optimal oral anticoagulant drug administration. The algorithm uses a second-order discrete-time model of the response of coagulation key parameter (INR) to warfarin, which is adapted to the actual patient’s dynamic response during the first period of therapy. A state and disturbance estimator is used to correct the model prediction with feedback information from INR measurement. Then, the steady-state weekly sequence of war- farin is calculated that keeps INR as close to its desired target as possible. Finally, the warfarin doses for the first week (from the day when INR is measured) are computed by a dynamic optimization module, which assumes that the steady-state weekly sequence is used from the second week. 25 clinical tests have been performed with very promising results.