Keywords: Blinded safety signal monitoring, FDA IND safety reporting
In the United States, sponsors must report a suspected adverse reaction if an aggregate analysis indicates that a specific event “occurs more frequently in the drug treatment group than in a concurrent or historical control group”, including any clinically important increase over what is listed in the Investigator’s Brochure (IB) or inferred from data in the IB.
Using a Bayesian framework and a collaborative process, we can create continuous safety signals for a two-arm trial with a moderately informative prior on the combined event rate, which is updated with pooled, blinded data. The idea can be extended with a fully Bayesian method that uses a strong prior on the control rate and a separate weak prior on the treatment rate.