Modeling, adaptive control, and optimal drug therapy

Drug therapy, its clearly development, and the advent of pharmacokinetic models are described, from the original work of Teorell, through that of Augsberger and Kruger-Thiemer, to the present. Adaptive control of such models, long known in engineering, began in therapeutics with methods for linear and nonlinear least-squares regression, and has progressed to the Maximum Aposteriori Probability (MAP) Bayesian method. Strategies for optimal monitoring of serum concentrations are described and their clinical results briefly evaluated. Lastly, the new method of Approximate Optimal Closed-Loop (AOCL) control is described, in which the therapeutic regimen is used at the same time to probe (learn about) the patient's model approximately optimally. The new method considers the expected values of planned future serum concentrations (or other responses), in addition to the traditional measurement of past serum concentrations. This should optimize the process of learning about a patient's model while treating him at the same time.