Dose-Response Modelling and Optimization of Quantitative High-Throughput Screening Assays

Human health risk assessment is a process designed to characterize potential health risks associated with exposure to environmental agents. Classically, it involves four main steps: (1) hazard identification, (2) dose-response assessment, (3) exposure assessment, and (4) risk characterization. Traditionally, toxicological testing has relied heavily on experimental animals to predict potential human health risk. Motivated in part by the 2007 U.S. National Research Council report, Toxicity Testing in the 21st Century: A Vision and a Strategy, there has been a shift towards new approach methodologies using in vitro, in silico, and in chemico techniques. Quantitative high-throughput screening (qHTS) is one such methodology that can rapidly produce in vitro assays for thousands of chemicals that can be analyzed for hazard identification and dose-response assessment purposes. This thesis provides a statistical foundation for effectively designing qHTS assays and extends the value of information (VOI) framework to allow for more realistic comparisons in practice. Using over 8,000 qHTS assays from the Tox21 program, we develop optimal designs that maximize the efficiency of BMD estimates. Further, we extend the framework for a VOI analysis to compare the benefits realized by collecting information to aid in health risk decision-making, including variables such as timeliness, cost, and reduction in uncertainty. Initially, the statistical properties of maximum likelihood estimators (MLEs) of the BMDs and underlying Hill model parameters are studied. A simulation study is performed to investigate the relationship between the properties of the MLEs and the observed concentration response curve (CRC). The development of optimal designs based on BMD estimates for qHTS assays and the comparison of efficiencies for six fixed designs (FDs) is considered next. The relatively low median relative efficiencies for the FDs suggests that without some knowledge of the model parameters, no fixed design can consistently deliver high efficiency across all shapes of CRCs. Finally, the VOI framework is extended by calculating the total social cost (TSC) over a specified time horizon instead of the annualized social cost (ASC) traditionally used in practice. The TSC uses the timeliness of information collected, thereby allowing for a more complete comparison of alternative toxicological testing methodologies.