Abraham G. Hartzema
University of Florida, USA
Title: Strenghtening the Clinical Evidence provided by Observational Studies
Biography
Biography: Abraham G. Hartzema
Abstract
Why is the clinical evidence from observational studies in the literature often conflicting and delayed? We will address the efforts undertaken in the US FDA to provide more recent evidence on the safety of newly marketed drugs as mandated by US Congress. The Sentinel system is a distributed data system that allow for the time analysis of emerging drug safety questions. The automation of the safety queries and the large population covered (120 male patients), allows for the timely analysis and provide the statistical power to address safety question in the early uptake of drugs in the marketplace. The second important question is the often conflicting evidence presented in the literature. What is the source of this conflicting evidence? Our group has conducted significant research in the influence of biases, confounding and the interaction of statistical methods and structure and quality of data on the clinical evidence presented. Running statistical methods, including high dimensional propensity scores on 12 different databases we found that self-controlled designs, those observational designs in which the patient serves as its own control are the most powerful designs as these provide the opportunity to control for fixed confounders. These methods do not control for variable confounders and are most suitable when the outcomes follow immediately after drug exposure. Other issues considered are case identification in large databases and formulating search algorithm that have high PPVs to make sure that cases are indeed attributable to the exposure. Recent proposed methods include sensitivity analysis to test the robustness of the clinical evidence identified. Adherence to these practices will assure that the clinical evidence presented is valid and can be incorporated in practice guidelines