Projecting adverse event incidence rates using empirical Bayes methodology

Area of Science:

• Biostatistics
• Clinical Trials
• Pharmacovigilance

Background:

• Clinical trials often have limited durations, necessitating projections of adverse event incidence rates for longer periods.
• Simple modeling of individual adverse event rates overlooks valuable information shared across all adverse event data.

Purpose of the Study:

• To develop and evaluate a novel statistical method for projecting adverse event incidence rates beyond the observed trial duration.
• To improve the accuracy of adverse event incidence rate projections by leveraging data from all adverse events.

Main Methods:

• Proposed an empirical Bayes method incorporating a shrinkage factor to weight projections of adverse event incidence rates.
• Developed a technique to estimate the proportion of true null hypotheses (common area under the density curves) to determine the shrinkage factor.
• Evaluated the method's performance by projecting from interim data and comparing with observed outcomes.

Main Results:

• The proposed shrinkage method demonstrated improved accuracy in projecting adverse event incidence rates compared to naive approaches.
• The estimation of the proportion of true nulls was a critical component in deriving the shrinkage factor.
• The method was successfully illustrated on two distinct clinical trial datasets.

Conclusions:

• The empirical Bayes shrinkage method offers a robust approach for projecting adverse event incidence rates in clinical trials.
• This method enhances the reliability of long-term adverse event forecasting by effectively utilizing all available adverse event data.
• Accurate adverse event projection is crucial for understanding the full safety profile of interventions in clinical practice.