Composite Outcomes

A single endpoint combining several component events — e.g. death or re-infarction or stroke or revascularisation. A patient “has the outcome” if any component occurs. It boosts event counts, statistical power and captures a net effect, but can mislead when the components differ in importance or direction.

Several separate events feeding into one combined endpoint. Because any single component counts, the composite collects far more events than death alone — but the headline result hides which component did the work.

Never stop at the composite. Open it up and read each component separately: how big is each effect, and do they all point the same way? Be especially wary when a frequent, subjective (“soft”) component such as revascularisation drives the result while the hard outcomes — death, stroke — sit still.

Composites are common in cardiology and EM trials precisely because they raise power. But an apparently “positive” trial can mean little if the benefit is confined to its least important component, or if components move in opposite directions and cancel a true signal.

• Composite = any one component counts as the event
• Always inspect components individually, especially death
• Beware soft, frequent components carrying the result

PROPPR (JAMA 2015) — 680 severely injured trauma patients, 1:1:1 vs 1:1:2 plasma:platelet:red-cell transfusion. No significant difference in the primary outcomes of 24-hour or 30-day mortality, yet more 1:1:1 patients achieved haemostasis and fewer died of exsanguination by 24 hours. A reminder to read components individually: a softer, more frequent signal (haemostasis, exsanguination death) shifted while the hard all-cause mortality endpoint did not.

• Assuming every component improved because the composite is positive.
• Letting a frequent soft component drive a “positive” headline.
• Missing components moving in opposite directions and cancelling out.