I love your podcast and listen every week. Over the past couple of years I’ve noticed a growing number of trials—especially early DAPT discontinuation trials—using composite endpoints that combine competing outcomes in noninferiority designs.
The most recent example was just last week. On November 27, NEJM published the TARGET-FIRST trial, which declares noninferiority based on a composite endpoint that mixes ischemic events with major bleeding. These outcomes reliably move in opposite directions when you drop aspirin. You cannot statistically lump them together in a noninferiority trial without biasing the entire analysis toward a “no difference” result. A reduction in bleeding can literally cancel out an increase in MI or stent thrombosis within the same composite. That’s not clinical equipoise—it’s a statistical trick.
It’s honestly surprising that NEJM would publish a noninferiority claim built that way.
What makes this even more striking is that other trials are doing it correctly. The NEO-MINDSET trial—published in August—uses the proper approach:
• test ischemic outcomes alone for noninferiority,
• then test bleeding separately for superiority.
No sleight of hand. No burying opposing-direction effects inside one number.
Yes, as cardiologists we absolutely need to weigh bleeding versus ischemic risk. But we can only do that if we know the true risk of each outcome. A composite-driven “no difference” result that mixes the signals together doesn’t help at all.
