In this paper we examine several of the fundamental foundations and relevant clinical issues in adaptive control of drug dosage regimens for patients. Truly individualized therapy with drugs having narrow margins of safety first requires a practical pharmacokinetic/dynamic model of the behavior of a drug, Past experience with a drug is stored in the form of a population model. Next, using the information in such a model and its relationship to the incidence of adverse reactions, a specific, explicit therapeutic goal must be selected by the responsible clinician, based on the patient's need for the drug and the risk of adverse reactions felt to be justified by each patient's need, small, moderate, or great. Individualized drug therapy thus begins with the selection of individualized therapeutic goals (low, moderate, or high) for each patient. Using subsequent feedback from the patient's serum drug levels, and using Bayesian fitting, the model is then linked to each patient as a patient-specific model. Control of the model by the dosage regimen increasingly controls the patient, to better obtain the desired explicit therapeutic goals. This process is essentially similar to that of a flight control or missile guidance system.
