PARAMETER ESTIMABILITY OF BIPHASIC RESPONSE MODELS

Pharmacodynamics of general anesthetic agents generally exhibit biphas ic concentration-effect relationships (i,e., an activation phase at lo w concentrations and inhibition at higher concentrations), These relat ionships are usually characterized with biphasic models constructed fr om various combinations and modifications oi the nonlinear sigmoid E(M AX) model, We tested and quantified the parameter estimability of the simplest additive biphasic pharmacodynamic models by a Monte Carlo met hod. The estimated model parameters were used to calculate descriptors of the concentration-effect data. Parameters and descriptors were com pared with their true values. When the IC50/EC(50) ratio was low (<10) , E(MAX), EC(50), and IC50 were poorly estimated (high coefficient of variation and pronounced bias). However, the fit to the data was excel lent, and the data descriptors calculated from the estimated model par ameters demonstrated high precision and accuracy. Baseline effect (Eo) was estimated with good precision and accuracy. As the IC50/EC(50) ra tio was increased, the estimability of model parameters and data descr iptors improved, with the data descriptors continuing to be more estim able than model parameters. Thus, model parameters become estimable wh en there is sufficient separation between EC(50) and IC50 to produce a plateauing of peak effect [activation], which can be observed directl y from the data signature, Data descriptors are not subject to this li mitation and thus may serve as better metrics for summarizing concentr ation-effect relationships.