Drug Safety Risk-Benefit Models

点击翻译按钮获取中文摘要

Drug Safety Risk-Benefit ModelsJames C. BenneyanMechanical and Industrial EngineeringNortheastern UniversityBoston, MAAbstractThis grant provides funding to develop new analytic drug safety models to help evaluate risk-benefit tradeoffs of pharmaceutical drugs, a large and growing public and regulatory concern. Three mathematical approaches will be developed and integrated for evaluating drug safety and risk-benefit tradeoffs, based on therapeutic ratio fractional programming models, probabilistic and mean-variance models using moments and distributions of utility-based measures, and probabilistic frontier models to identify stochastically preferential drugs or those maximizing mean net benefits with minimal risk uncertainty. An important and unique focus of this research is that while conventional methods essentially are population expected-utility based, the developed algorithms will provide more comprehensive patient-centered information accounting for risk uncertainty. Each modeling approach will develop solutions to several technical and mathematical issues prevalent in these types of applications and will be developed working with clinical and pharmacovigilence colleagues who will provide project guidance, pilot data, and content area expertise. If successful, the results of this research will contribute improved methods for evaluating the relative risks versus benefits of pharmaceutical drugs while, both under development and undergoing regulatory approval processes. The developed models will be tested on anti-infectives, anticoagulants, prescription analgesics, and other applications identified during this work. These models could be used separately or together to help assess risk-benefit tradeoffs, understand limitations of non-optimal products, and make approval or development decisions. More broadly, while motivated here by drug safety problems, the envisioned results potentially also will offer fundamentally new approaches for evaluating risk-benefit tradeoffs in a wide range of healthcare and non-medical applications. The proposed work also will make important methodological contributions to utility theory, data envelopment, and stochastic frontier analysis methods.