The influence of the way of regression on the results obtained by the receptorial responsiveness method (RRM), a procedure to estimate a change in the concentration of a pharmacological agonist near the receptor

Abstract

The receptorial responsiveness method (RRM) enables the estimation of a change in the concentration of a degradable agonist, near its receptor, by fitting its model to (at least) two concentration-effect (E/c) curves of a stable agonist of this receptor. One curve should be generated before this change in concentration, while the other one after this change, in the same (or in identical) biological system(s). It follows that RRM yields a surrogate parameter ("cx"), the concentration of the stable agonist that is equieffective with the change in the concentration of the degradable agonist. However, the curve fitting can be implemented several ways, which can affect accuracy, precision and convenience of use. This study utilized data of previous ex vivo investigations. Known concentrations of stable agonists were estimated with RRM by performing individual (local) or global fitting (with one or two model(s)), combined with the use of a logarithmic (logcx) or non-logarithmic parameter (cx), and with ordinary least-squares or robust regression. We found that the individual regression, the most complicated option, was the most accurate, followed closely by the moderately complicated two-model global regression and then by the easy-to-perform one-model global regression. The two-model global fitting was the most precise, followed by the individual fitting (closely) and by the one-model global fitting (from afar). The use of cx and robust regression did not, whereas pairwise fitting (i.e. fitting only two E/c curves at once) did improve the quality of estimation. Thus, the two-model global fitting, performed pairwise, is recommended for RRM, but the individual fitting is a good alternative.