A model-free approach to estimation of relative potency in dose-response curve analysis

Abstract

We have developed a new, general approach to analysis of dose-response curves from bioassay, immunoassay (including radioimmunoassay, immunoradiouretic assay, enzyme-linked immunosorbent assay), and other experimental procedures. It provides a test for parallelism, similarity of shape, and a measure of relative potency for any set of two or more curves. The method uses a constrained smoothing spline function to estimate the curve shape, together with a nonlinear least-squares fitting technique to estimate parameters for relative potency and slope. The use of "constrained splines" permits the analysis of nonlinear dose-response curves that cannot be described by a simple model or equation such as the symmetric four-parameter logistic. A microcomputer program is used for the analysis, providing relative potencies and their SE and evaluation of goodness of fit.