Prefer a table over a stacked bar chart for trial safety review

Choose a table instead of a stacked bar chart when clinicians must judge whether adverse outcomes are concentrated in one treatment group. For example, show treatment-by-outcome counts in rows and columns rather than stacking the outcomes into bars for a stop-or-continue safety review.

A table makes the treatment-outcome counts directly inspectable. That supports a clearer read of whether one treatment is associated with a worse outcome pattern.

Mechanism: The table exposes the cross-classified counts needed for the safety judgment directly, while the stacked bar requires more visual inference about the association.

Evidence: The review reports that, in a simulated patient-safety review using clinical-trial adverse-event data, physicians were more likely to halt the trial when the data were shown in tables than in stacked bar charts (Ancker & Kaufman, 2007).

• User Goal: Decide whether a safety signal is strong enough to stop or flag a trial.
• Task: Compare outcomes across treatment groups for a strong adverse-event association.
• Data: Grouped treatment-by-outcome counts.
• Chart Setting: A static display used in clinical safety review.
• Audience: Physicians or other clinical reviewers.
• Success Criterion: More reliable detection of the adverse-event pattern that matters for the decision.

Break it when: The display is not being used for clinician review of grouped treatment-outcome safety data. Why: The reported advantage is specific to that review setting.

Sacrifice: You give up the stacked bar’s single combined mark. Risk: If reviewers must infer the association from stacked segments, a strong safety signal can be missed. Mitigation: Put the treatment and outcome counts directly into table cells.

Mistake: Using a stacked bar chart for the stop-or-continue safety decision. Why it fails: The treatment-outcome association is harder to judge directly.

Failure Sign: Reviewers must estimate the association from segment sizes instead of reading counts directly. Quick Check: Mock up the same data as a table and as a stacked bar chart, then ask which version makes the adverse concentration in one treatment group easier to judge directly. Stronger Test: Ask a domain reviewer whether they would halt the trial from each version and treat disagreement as a sign to prefer the table.

• Replace the stacked bar chart with a treatment-by-outcome table.
• Move the outcome counts out of stacked segments and into cells.
• Limit the display to the treatment and outcome breakdown needed for the safety decision.