Statistical strategies for averaging EC50 from multiple dose–response experiments

Abstract

In most dose–response studies, repeated experiments are conducted to determine the EC50 value for a chemical, requiring averaging EC50 estimates from a series of experiments. Two statistical strategies, the mixed-effect modeling and the meta-analysis approach, can be applied to estimate average behavior of EC50 values over all experiments by considering the variabilities within and among experiments. We investigated these two strategies in two common cases of multiple dose–response experiments in (a) complete and explicit dose–response relationships are observed in all experiments and in (b) only in a subset of experiments. In case (a), the meta-analysis strategy is a simple and robust method to average EC50 estimates. In case (b), all experimental data sets can be first screened using the dose–response screening plot, which allows visualization and comparison of multiple dose–response experimental results. As long as more than three experiments provide information about complete dose–response relationships, the experiments that cover incomplete relationships can be excluded from the meta-analysis strategy of averaging EC50 estimates. If there are only two experiments containing complete dose–response information, the mixed-effects model approach is suggested. We subsequently provided a web application for non-statisticians to implement the proposed meta-analysis strategy of averaging EC50 estimates from multiple dose–response experiments.