Visualizing the treatment effect on kidney hierarchical composite endpoints: From mosaic to maraca plots

TL;DR

This paper introduces a comprehensive visualization framework for kidney hierarchical composite endpoints in clinical trials, including sunset, maraca, mosaic, and Dustin plots, to enhance interpretation of treatment effects.

Contribution

It proposes new visualization methods tailored for hierarchical composite endpoints, improving clarity and interpretability of treatment effects in clinical trial results.

Findings

Maraca plots provide detailed distribution of hierarchical composite endpoints.

Sunset plots visualize all potential treatment effects based on component effects.

Dustin plots assess consistency of treatment effects across components.

Abstract

Visualizations, alongside summary tables and participant-level listings, are essential for presenting clinical trial results transparently and comprehensively. When reporting the results of clinical trials, the goal of visualization is to communicate the results of specific pre-planned analyses with visualization that are tailored to the endpoint and analysis being reported. We are considering the visualization of HCEs, combining multiple time-to-event outcomes, ordered according to a given prioritization and the timing of events, with a single continuous outcome. An illustrative example is the kidney disease progression HCE with a straightforward structure of the composite of clinical events of death and kidney failure and declines in eGFR as surrogates for kidney failure. The HCEs are analyzed by win statistics and visualized using maraca plots. Although maraca plots are very granular and allow for a detailed presentation of the distribution of HCE, researchers are still tasked with explanation of the magnitude of the treatment effect estimated by win odds. In explaining the magnitude of the treatment effect, we propose a comprehensive visualization approach. In the clinical trial design stage, we propose the sunset plots to visualize all possible treatment effects that can be observed based on the treatment effects on components. In reporting the results of the trial, we recommend the maraca plots as the primary method of visualization of the results. While the 2-d mosaic plot with the ordinal dominance graph directly corresponds to the win odds as treatment effect measure and can be used as the primary analysis-specific visualization method. And finally, we propose the Dustin plot to visualize the supportive analysis of the components, added cumulatively from the event of highest priority to assess the consistency of the treatment effect on all outcomes.