Reporting cumulative proportion of subjects with an adverse event based on data from multiple studies

Area of Science:

• Biostatistics
• Clinical Trial Analysis
• Pharmacovigilance

Background:

• Naïvely pooling data from multiple studies can lead to Simpson's Paradox, a statistical phenomenon that can distort findings.
• This paradox particularly affects the analysis of proportions, such as the incidence of adverse events (AEs).
• Current meta-analytic approaches focus on comparing safety data but lack standardized methods for reporting cumulative AE proportions.

Purpose of the Study:

• To demonstrate the impact of Simpson's Paradox on estimating and comparing adverse event proportions from pooled data.
• To propose novel approaches for reporting cumulative proportions of patients experiencing target adverse events.
• To advocate for regulatory guidelines for scientifically defensible reporting of cumulative safety data.

Main Methods:

• Illustrating Simpson's Paradox using the example of target adverse event proportions in pooled study data.
• Developing and presenting two distinct methodologies for reporting cumulative adverse event proportions.
• Reviewing existing meta-analytic practices for safety data comparison.

Main Results:

• Simpson's Paradox can significantly alter the perceived risk of adverse events when data are pooled without appropriate methods.
• The proposed approaches offer statistically sound ways to report cumulative adverse event proportions.
• Lack of standardized reporting guidelines can lead to misinterpretation of safety data.

Conclusions:

• Careful consideration of statistical methods is crucial when pooling safety data to avoid Simpson's Paradox.
• Standardized reporting of cumulative adverse event proportions is needed for accurate risk assessment.
• Regulatory bodies should establish guidelines for reporting pooled safety data to ensure clarity and scientific rigor.