Two of today's findings are composites, and in both the useful information is in how the arms differed rather than in the summary number.
In the atrial fibrillation trial, pace-and-ablate and pulmonary vein isolation did not differ by one arm having fewer events of the same kind. The ablation arm generated atrial fibrillation admissions and cardioversions; the pace-and-ablate arm generated heart failure admissions. A composite hazard ratio of 0.45 conceals that entirely, and the trade it describes — fewer arrhythmia episodes, more heart failure — is exactly what a patient of 82 needs to weigh.
The habit is quick to apply. When a trial reports a composite, look for the component table before the forest plot, and ask which component is driving the result and whether it is the one your patient cares about. A composite driven by cardioversions is a different clinical claim from one driven by death.
- Find the component breakdown before quoting a composite hazard ratio to a patient
- Ask which component moved, and whether it is the one that matters to them
- Watch for composites where a soft component supplies most of the events
- Two strategies can share an endpoint rate while failing in opposite ways
- Say the trade out loud in consent: fewer arrhythmia admissions, more heart failure ones
The statistics, in plain English
A composite endpoint gathers several outcomes so a trial can reach significance with fewer patients. The cost is interpretive: the summary estimate weights a cardioversion the same as a heart failure admission, though no patient does. The component counts are where the clinical meaning sits, which is why good reports print them.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for cardiology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free