Shiny-MAGEC: A Bayesian R shiny application for meta-analysis of censored adverse events

Area of Science:

• Clinical Pharmacology
• Biostatistics
• Drug Safety Evaluation

Background:

• Accurate adverse event (AE) incidence assessment is crucial for drug safety in clinical research.
• Meta-analysis synthesizes evidence but often faces challenges due to incomplete AE reporting, leading to left-censored data.
• Ignoring censored AE data can bias incidence estimates, compromising safety assessments.

Purpose of the Study:

• To present an R Shiny application implementing a Bayesian meta-analysis model designed to incorporate censored AE data.
• To provide researchers with a user-friendly tool for unbiased AE incidence probability estimation.
• To highlight the biases of conventional methods by comparing models that do and do not account for censoring.

Main Methods:

• Development of an R Shiny application featuring a Bayesian meta-analysis model.
• The model specifically incorporates left-censored AE data, addressing underreporting below study thresholds.
• The application allows direct comparison of censored vs. uncensored meta-analysis models.

Main Results:

• The Shiny application facilitates unbiased estimation of AE incidence probability.
• Demonstrated the significant bias introduced by conventional meta-analysis methods that ignore censored data.
• An illustrative example using PD-1/PD-L1 inhibitor safety data showcased the tool's utility.

Conclusions:

• The developed Shiny app, Shiny-MAGEC, improves the accuracy and transparency of AE risk assessment in drug safety.
• Incorporating censored data through Bayesian meta-analysis leads to more reliable AE incidence estimates.
• This tool is vital for robust clinical research and drug safety evaluations.