A Composite Is Only Its Components

A reanalysis of SURPASS-2 asked how many standard treatment targets each drug got patients to — blood sugar, blood pressure, cholesterol, weight — and tirzepatide won at every dose.^[1] Before that means anything, note what was compared: tirzepatide at up to 15 mg against semaglutide at 1 mg, the diabetes dose. And note which targets actually moved.

The dose question first

That is not a scandal — SURPASS-2 was a diabetes trial, 1 mg was the licensed diabetes dose when it ran, and comparing against an approved comparator at an approved dose is legitimate design. What it means is narrower than the sentence people extract from it. “Tirzepatide gets more patients to target than semaglutide 1 mg in type 2 diabetes” is supported. “Tirzepatide beats semaglutide” is not the same claim.

The trial that does compare them at their weight-management doses is a different study, and its quality-of-life analysis is covered in the mental health score that did not move.

What the analysis did

The authors took 1,879 participants from SURPASS-2 and scored each against four standard targets and four stricter ones, at 40 weeks.

At baseline almost nobody was near all four. On the standard set, 19% met none and only 21% met two or more. On the intensive set, 58% met none and just 4% met two.

Which components carried it

“Three of four targets” sounds like all-round metabolic control. The odds ratios show where the difference actually came from.

Blood sugar and weight do the work, and weight does most of it — nearly four times the odds of losing more than 15%. Blood pressure contributes only at the stricter threshold. LDL cholesterol does not appear at all.

A composite counts its components equally. A drug does not affect them equally.

What is genuinely useful here

Setting the framing aside, the underlying observation matters clinically. Most people with type 2 diabetes are not near their targets — 19% met none of four at baseline — and getting to several at once has historically meant stacking several drugs, each with its own side effects and its own adherence cost.

A single agent moving two of the four substantially is a real simplification, and 57% reaching three or more standard targets is a large number in this population by any historical comparison. The honest version of the finding is that this drug is very good at two of these four things, which turns out to be enough to move a composite a long way.

One structural note: the targets were applied to the trial data afterward rather than being its endpoints. That is standard and reasonable for this kind of question, and it is still a post hoc analysis.