In ligand binding assays (LBA), the concentration to response data is a nonlinear relationship driven by the law of mass action. Four parameter logistic (4PL) and five parameter logistic (5PL) curve fitting models are two widely accepted and validated models for LBA calibration curve data. Selection of the appropriate regression model and weighting function are key components of LBA development. Assessment of selected model and weighting function should be performed during assay development and confirmed later during validation. There has been limited published work on practical approaches to determining an appropriate weighting function and selection of a regression model for ligand binding assays. Herein, a structured scheme is presented to determine both. By applying commonly available software, assay performance data were analyzed to determine weighting functions and associated choice of a curve fitting model in three presented case studies. As a result, assay ranges of quantification were improved by reducing lower limit of quantification (from 1.00 to 0.317 ng/mL in one case study and from 2.06 to 1.37 ng/mL in another) or extending both low and upper limits of quantification(e.g., 1.04 to 48.3 ng/mL improved to 0.602 to 145 ng/mL). In addition, assay calibration curve performance demonstrated improved assay accuracy (%RE) and precision (%CV). Recommendations on decision flow when determining appropriate weighting function and curve fit model are presented.
Keywords: LBA; calibration curve; heteroscedasticity; pharmacokinetic assay; weighting function.