The obesity trials that produced the famous percentages all excluded diabetes, which left the most common reason a person is prescribed one of these drugs sitting outside the evidence. SURMOUNT-2 was built to close that gap for tirzepatide, and it did: a weight trial, in people with type 2 diabetes, with weight-based coprimary endpoints. The number it returned is lower than the one from the trial without diabetes, and a post hoc analysis of its own participants suggests the shortfall is not evenly distributed at all — it belongs to a subgroup the trial could already identify eight weeks in.
The funnel, and the three arms
SURMOUNT-2 was a phase 3, double-blind, randomized, placebo-controlled trial conducted in seven countries between 29 March 2021 and 10 April 2023. Adults aged 18 or over with a body-mass index of 27 or higher and a glycated hemoglobin of 7% to 10% were assigned 1:1:1 to once-weekly subcutaneous tirzepatide at 10 mg or 15 mg, or to placebo, for 72 weeks.[1]
Of 1,514 adults assessed for eligibility, 938 were randomized and received at least one dose: tirzepatide 10 mg (n = 312), tirzepatide 15 mg (n = 311) and placebo (n = 315).[1] Nearly two of every five people screened did not enter, which is the usual shape of a trial with a glycemic window in its eligibility criteria and is worth holding when the results are read as a population estimate.
The estimand a diabetes trial needs
The coprimary endpoints were the percent change in body weight from baseline and a body-weight reduction of 5% or higher. The treatment-regimen estimand assessed effects regardless of treatment discontinuation or initiation of antihyperglycemic rescue therapy, analyzed in the intention-to-treat population.[1]
That second clause does not appear in the trials without diabetes, and it carries real weight. A participant whose glucose control deteriorated was given rescue medication, and glucose-lowering drugs are not neutral on the scale: some add weight and some remove it. Counting those participants anyway is the conservative choice, and it means the headline percentage already contains people whose regimen changed underneath them. The distinction between treating diabetes and treating weight is set out in the two-indications article.
Who this trial enrolled
The 938 participants had a mean age of 54.2 years (SD 10.6). Four hundred and seventy-six (51%) were female, 710 (76%) were White and 561 (60%) were Hispanic or Latino. Baseline mean body weight was 100.7 kg (SD 21.1), body-mass index 36.1 (SD 6.6) and glycated hemoglobin 8.02% (SD 0.89).[1]
Set that beside the pivotal trial without diabetes, which randomized 2,539 adults with a mean baseline weight of 104.8 kg and a mean body-mass index of 38.0.[2] SURMOUNT-2's participants were about eight years older, lighter, less obese by body-mass index, evenly split by sex rather than predominantly women, and majority Hispanic or Latino. Almost nothing about the two cohorts is interchangeable except the molecule.
They also arrived on treatment. In the SURMOUNT-2 population, 54% were receiving one oral antihyperglycemic medication at baseline and 39% were receiving two or more.[3] Tirzepatide was added on top of that, not substituted for it.
What the trial reported
Least-squares mean change in body weight at week 72 was −12.8% (SE 0.6) with tirzepatide 10 mg and −14.7% (0.5) with 15 mg, against −3.2% (0.5) with placebo. The estimated treatment differences were −9.6 percentage points (95% CI, −11.1 to −8.1) at 10 mg and −11.6 percentage points (95% CI, −13.0 to −10.1) at 15 mg, both p < 0.0001. A reduction of 5% or more was reached by 79% to 83% of tirzepatide participants against 32% on placebo.[1]
The step from 10 mg to 15 mg is worth naming: a 50% larger dose bought 1.9 percentage points. The dose-by-dose picture across both tirzepatide obesity trials is in the tirzepatide weight article, and the escalation schedule that reaches those doses is in the titration article.
The split the average is hiding
A 2026 post hoc analysis divided tirzepatide-treated participants by whether they had lost at least 5% of body weight at week 8, the point at which everyone had spent four weeks at 5 mg and was due to escalate. In SURMOUNT-2, 247 of 609 analyzed participants — 40.6% — met that bar. In the trial without diabetes, 1,103 of 1,775 did, or 62.1%.[3]
At week 8 the two SURMOUNT-2 groups stood at −7.1% and −2.5% (p < 0.001). At week 72 they stood at −20.0% and −10.8% (p < 0.001). Within the 15 mg arm alone, early responders reached −21.2% and the rest reached −11.2%.[3]
That upper figure is the one that complicates every simple story about diabetes and these drugs. Early responders on 15 mg who had diabetes finished at −21.2%, which is the territory the same dose reached in people without diabetes.[2] The population-level shortfall is carried almost entirely by the 59.4% who had not moved much by week 8.
Two cautions before that travels anywhere. These figures use the efficacy estimand over tirzepatide-treated participants who had week-8 data, whereas the trial's −14.7% uses the treatment-regimen estimand over everyone randomized; the two numbers have different denominators and different assumptions and cannot be stacked. And the split is post hoc, without multiplicity adjustment, which its authors describe as hypothesis-generating.[3]
What separated the two groups at baseline
Early responders had a lower baseline glycated hemoglobin (7.9 ± 0.83% against 8.1 ± 0.96%) and were far less likely to be on more than one antihyperglycemic medication — 28.3% against 44.2%. They were also more likely to be female (54.3% against 47.8%) and more likely to be White (84.2% against 67.7%).[3]
Those characteristics move together, and a post hoc split cannot separate them. What the gradient suggests is that the relevant variable is not the diagnosis but its depth: more established diabetes, on more medication, with higher glycated hemoglobin, responded less. The same attenuation appears with the other molecule — semaglutide 2.4 mg produced −9.6% against −3.4% in 1,210 adults with type 2 diabetes, an estimated treatment difference of −6.2 percentage points (95% CI, −7.3 to −5.2)[5] — and why that happens is the subject of the STEP 2 article.
Tolerability, and what the primary report did not break out
The most frequent adverse events with tirzepatide were gastrointestinal, including nausea, diarrhea and vomiting, mostly mild to moderate in severity, with fewer than 5% of events leading to treatment discontinuation. Serious adverse events were reported by 68 participants (7%) overall, and two deaths occurred in the tirzepatide 10 mg group, neither considered related to study treatment by the investigator.[1]
The primary publication reports that discontinuation figure as a ceiling rather than arm by arm. The comparison trial without diabetes did break it out, at 4.3%, 7.1% and 6.2% for 5, 10 and 15 mg against 2.6% on placebo.[2] The post hoc analysis adds one useful reassurance: the pattern, severity and time course of gastrointestinal events were similar in early responders and in everyone else, so a poor week-8 result was not explained by people tolerating the drug worse.[3]
Weight alone is not the bar clinicians set
A 2026 post hoc analysis across the SURMOUNT program asked how many participants cleared three targets at once: a weight-reduction threshold, a systolic blood pressure reduction of at least 5 mmHg, and non-HDL cholesterol below 130 mg/dL. At the 5% weight threshold, 32% to 38% of tirzepatide participants cleared all three against 2% to 8% on placebo; at the 15% threshold, 22% to 34% against 1% to 3%, all p < 0.001.[4]
The ratio between the arms is large and the absolute number is not. Under a third to a bit over a third of treated participants met a composite that a cardiologist would regard as unremarkable, which is a more honest summary of what 72 weeks delivers than the weight column on its own.
The comparison SURMOUNT-2 could not make
Its comparator was placebo added to whatever the participant was already taking. Nobody was randomized to the alternative a prescriber actually weighs, which is a different glucose-lowering drug, so the trial cannot say whether tirzepatide is the better next addition for a person with a glycated hemoglobin of 8%. It also counted no cardiovascular events, ran for 72 weeks, and enrolled only within a 7% to 10% glycemic window, which excludes both well-controlled diabetes and the least controlled.
Nothing here was measured after the drug stopped, either. The only randomized withdrawal evidence for this molecule comes from the SURMOUNT-4 article, and it enrolled people without diabetes.
What was in the syringe
Every figure belongs to branded, FDA-approved tirzepatide at labeled doses, escalated under supervision, in people whose diabetes medications were being monitored throughout. Services on the tirzepatide board mostly dispense compounded preparations, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed — the distinction drawn in the compounded-versus-brand article.
A person with type 2 diabetes adding a weekly injection to an existing regimen is changing a medication list that somebody has to hold. That is a prescriber's job rather than an intake form's, and how each figure on this page was checked is described in the methodology.