Bayesian Models for Safety Signal Detection in Ongoing Blinded Studies

The regulatory landscape and approach to safety surveillance has shifted over the last decade from individual case reporting to systematic aggregate review. The FDA Final Rule guidance (2015) recommends regular unblinded aggregate analyses by a Safety Assessment Committee (SAC) to identify serious adverse events (SAEs) occurring at higher rates under active treatment versus control, requiring expedited reporting. Routinely unblinding ongoing studies could raise data integrity concerns. An alternative approach could be to conduct blinded monitoring of ongoing studies and estimate whether the true rates for selected SAEs are higher for active treatment compared to control by a pre-specified margin beyond which a safety signal would be raised. Bayesian models utilizing historical control data are an obvious choice of method given that strong information about the control rate is required to estimate the active group rate in an analysis of blinded data. If a potential signal was identified via blinded monitoring, an unblinded aggregate review by an independent group (eg, SAC) could be conducted. In this talk we will discuss approaches for aggregate safety analyses with a focus on blinded monitoring of ongoing studies aimed at earlier signal detection. We will present results from extensive simulations evaluating the performance of Bayesian models across a range of different scenarios for blinded study analyses, and will discuss results from an application of the models.

About the speaker
Cindy McShea

Cindy received a BS in Mathematics in 1993 from East Carolina University in North Carolina, USA. In 2001 she completed an MPH in Biostatistics from the University of North Carolina at Chapel Hill in North Carolina, USA. Started her career in the pharmaceutical industry in 1993 working with statistical process control as a quality assurance auditor at Burroughs Wellcome and managing product stability studies a Magellan Laboratories and Bayer Biologics. After completing graduate studies in biostatistics, Cindy worked as a senior Biostatistician at Quintiles, Inc. She moved to Schwarz Biosciences in 2005 which became UCB Biosciences in 2008. At UCB Cindy has held different roles within clinical development late phase biostatistics. These include senior biostatistician, principal biostatistician, associate director/US group head, and Mission Lead within the neurology therapeutic area. Cindy is currently a director of biostatistics, serving as the Safety Statistics lead within the Statistical Science and Innovations group at UCB.

Daniel Meddings

Daniel studied at the University of Oxford in the Unit of Health Care- Epidemiology from 2004 – 2008. He received his Msc and PhD in Statistics from University College in London between 2008 and 2014. Daniel started his career in 2014 at Glaxo Smith Kline as a Principal Statistician within the Respiratory therapeutic area. In 2016, he moved to UCB in Slough and served as a Senior Exploratory Statistician within the Global Exploratory Development group until 2018. He is currently a Statistical Methodology Expert within the Center of Excellence in Statistical Innovation (CESI) group at UCB.

Cookies help us deliver our services. By using our services, you agree to our use of cookies. Learn more