Almost every trial on this site uses placebo. That answers whether a drug does something. It does not answer the question a patient and a doctor face: is this better than what we would have done anyway.
Four kinds of comparator
| Compared against | Example | What it tells you |
|---|---|---|
| Placebo | Most trials on this site | Whether the drug works at all |
| Another drug | SURPASS-2 | Which of two drugs works better |
| Real practice | SURPASS-EARLY | Whether it beats what you would otherwise get |
| A whole population | SURMOUNT-REAL UK, running | What it does to a health system |
What SURPASS-EARLY did
It randomized 794 adults with early type 2 diabetes at 78 sites in 10 countries [1]. One arm got tirzepatide. The other got intensified conventional care, delivered as clinicians actually deliver it and guided by local guidelines.
That comparator arm could include other GLP-1 drugs. Only tirzepatide itself was excluded. So this is a drug against real practice, including practice that reaches for the same family.
Over two years HbA1c fell 1.99 points against 1.32. The estimated treatment difference was 0.68 points, with an interval from 0.84 to 0.51. Weight differed by 8.0 kg. Waist circumference differed by 6.2 cm.
An HbA1c below 5.7% was reached by 60.2% against 24.0%.
The trial was open-label. Everyone knew their arm. For a lab value and a scale reading that matters less than it would for a symptom score. It still colors how hard each arm was intensified.
The rung below: a drug against a drug
SURPASS-2 compared tirzepatide against semaglutide directly [2]. HbA1c fell 2.01, 2.24 and 2.30 points at 5, 10 and 15 mg, against 1.86 on semaglutide. The differences were 0.15, 0.39 and 0.45 points. Weight differences ran 1.9 kg, 3.6 kg and 5.5 kg.
Note the comparator dose. It was semaglutide 1 mg, the diabetes dose, not the 2.4 mg used for weight management. A reader setting those figures beside the STEP trials is comparing different doses. That mismatch keeps recurring here, set out in the tirzepatide network comparison.
Read the three margins in order. Against placebo, arms fall by two points or more. Against another drug, 0.45. Against real practice, 0.68. The harder the comparator, the smaller the claim.
The rung above, still running
SURMOUNT-REAL UK is a five-year pragmatic trial in UK primary care [3]. It plans to enroll about 3,000 adults with class I obesity and no diabetes. They are randomized to tirzepatide plus standard care, or standard care alone.
Its endpoints go past the scale. Time to onset of type 2 diabetes at month 60. Healthcare resource use. Employment and sickness absence.
⚠ It has reported nothing. The primary endpoint is month 24 and the key secondary is month 60. Treat it as a design that now exists, not as evidence. That is the distinction this site applies to every unreported study, as in the review that found no outcome data at all.
For a reader, the rule is simple. Check what a figure was measured against before comparing it to anything. A 0.68-point margin over real care is smaller than the placebo-controlled arms this site usually reports. It is also more useful. The same reasoning applies to the older drugs in where metformin lands when ranked.