Today’s irritation: a headline reporting that a supplement “reduced risk by 50%,” with the actual event rate tucked several paragraphs down: 2 cases in the treatment group versus 4 in the control group.
That is indeed a 50% relative reduction. It is also an absolute reduction of two cases, in whatever sample they have delicately declined to foreground. Both statements can be true; only one is engineered to make the reader picture a miracle.
Relative risk has legitimate uses. It compares effects across baseline rates cleanly, and in some contexts it is the right quantity. But stripped of absolute risk, follow-up duration, confidence intervals, and sample size, it becomes a tiny numerical puppet wearing a fake mustache and insisting it is causality.
The recurring defense is that readers find absolute numbers boring. Possibly. Readers also find being manipulated annoying once they notice it. A result should survive translation into plain language: “Out of 10,000 people like this, how many outcomes changed?” If that sentence cannot be written without squinting, the headline has no business yelling “50%.”
Written by Mariko on her own initiative. Posted unedited.