Visualizing adverse events in clinical trials using correspondence analysis with R-package visae

Area of Science:

• Statistical analysis
• Data visualization
• Multivariate techniques

Background:

• Graphical displays and data visualization are crucial for understanding clinical trial adverse event (AE) data.
• Correspondence analysis (CA) is a multivariate technique for visualizing AE contingency tables in 2D plots.
• CA quantifies information loss, similar to principal components and factor analysis.

Purpose of the Study:

• To apply stacked CA using contribution biplots for exploring AE data differences across clinical trial treatments.
• To refine AE analysis using five levels based on Common Terminology Criteria for Adverse Events (CTCAE) data.
• To develop an interactive R-package (visae) for investigating CA configurations.

Main Methods:

• Stacked CA with contribution biplots applied to AE data.
• Analysis refined across five CTCAE levels (grades, domains, terms, combinations).
• Development of the 'visae' R-package for interactive CA exploration.
• Illustration using data from NSABP R-04 and NSABP B-35 randomized controlled trials (RCTs).

Main Results:

• CA biplots in R04 trial revealed discrepancies between single agent treatments and oxaliplatin combinations, explaining significant variability.
• An interaction effect was identified in R04 when adding oxaliplatin, indicated by distinct CA biplot quadrants.
• In the B35 trial, CA biplots showed differing patterns for non-adherent versus adherent Anastrozole and Tamoxifen groups.

Conclusions:

• CA with contribution biplots is an effective tool for summarizing AE data.
• This method provides a two-dimensional display that minimizes information loss and aids interpretation.
• The 'visae' R-package offers an interactive platform for AE data exploration using CA.