Skip to content
GLP Loss
← Research
Evidence

SURPASS-2 Explained: The Head-to-Head Ran at the Diabetes Dose

Tirzepatide beat semaglutide on glycated hemoglobin and weight at every dose in SURPASS-2. The comparator was semaglutide 1 mg rather than 2.4 mg, the trial was open-label, and every safety column in its posted results runs the other way.

Owen Castellanos10 min read
SURPASS-2: what each column says1,879 adults with type 2 diabetes, 40 weeks, open-labelArmHbA1cWeightSerious AESemaglutide 1 mg-1.86-6.2 kg13 of 469Tirzepatide 5 mg-2.09-7.8 kg33 of 470Tirzepatide 10 mg-2.37-10.3 kg25 of 469Tirzepatide 15 mg-2.46-12.4 kg27 of 470Posted trial results. The published abstract reports a smaller gap.The comparator was the diabetes doseSemaglutide 1 mg, not the 2.4 mg used for weight.Deaths: 1 on semaglutide, 4 in each tirzepatide arm.The trial was not powered to compare either column.

Almost every argument that tirzepatide outperforms semaglutide traces back to one trial, and it is a trial about blood sugar. SURPASS-2 is the only randomized head-to-head the two molecules have ever had, its result is unambiguous in the direction everyone reports, and four separate features of its design limit how far that result travels. What the price difference between the two looks like at the checkout is the subject of the buyer-facing comparison. This page is about the trial itself.

What was run

SURPASS-2 was an open-label, 40-week, phase 3 trial. 1,879 patients with type 2 diabetes were randomly assigned in a 1:1:1:1 ratio to tirzepatide at 5 mg, 10 mg or 15 mg or to semaglutide at 1 mg, all once weekly and all added to metformin. At baseline, mean glycated hemoglobin was 8.28%, mean age 56.6 years and mean weight 93.7 kg. The primary endpoint was the change in glycated hemoglobin from baseline to week 40.[1] Randomization put 471, 469, 470 and 469 participants in the four arms.[2]

Two design facts matter before any number does. Nobody was blinded, in a trial whose secondary endpoints include self-reported treatment satisfaction and whose adverse events are largely symptomatic. And the comparator was semaglutide at 1 mg — the dose used for glycemic control, not the 2.4 mg used for weight management. The difference between those two labels is set out in the two-indications article.

The primary endpoint, reported twice

The published abstract gives the estimated mean change in glycated hemoglobin as −2.01, −2.24 and −2.30 percentage points on tirzepatide 5, 10 and 15 mg against −1.86 on semaglutide. The differences are −0.15 points (95% CI, −0.28 to −0.03; P = 0.02), −0.39 (95% CI, −0.51 to −0.26; P < 0.001) and −0.45 (95% CI, −0.57 to −0.32; P < 0.001). Tirzepatide at all doses was noninferior and superior.[1]

The trial’s posted results give a different set for the same endpoint: −2.09, −2.37 and −2.46 against −1.86, with differences of −0.23 (95% CI, −0.36 to −0.10), −0.51 (95% CI, −0.64 to −0.38) and −0.60 (95% CI, −0.73 to −0.47). The registry states the analysis population explicitly: participants with a baseline and at least one post-baseline value, excluding data collected after starting rescue antihyperglycemic medication or after prematurely stopping the study drug.[2]

That is the same trial measured with and without the people who stopped taking the drug or needed something else added, and the gap between tirzepatide and semaglutide is roughly a third larger once they are removed. Neither figure is wrong. They answer different questions — what happens to everyone assigned a drug, and what happens to those who stayed on it — and a marketing page quoting only the larger one is quoting the narrower question.

The smallest dose wins by an amount nobody would feel

At 5 mg the advantage over semaglutide is 0.15 percentage points of glycated hemoglobin, with a confidence interval whose upper bound is 0.03.[1] That is a real, statistically significant difference and a clinically negligible one: roughly one seventh of a point on a measure that moved by more than two points in both arms. The separation that gets quoted comes from the 10 and 15 mg arms, and a seller advertising tirzepatide without naming a dose is not advertising the arm that produced it.

On target attainment the pattern is steeper. Glycated hemoglobin below 7% was reached by 81.13% on semaglutide against 85.47%, 88.89% and 92.24% on tirzepatide — odds ratios of 1.54 (95% CI, 1.06 to 2.23), 2.14 (95% CI, 1.44 to 3.17) and 3.03 (95% CI, 1.97 to 4.66). Below 5.7%, a threshold in the non-diabetic range, the figures are 19.74% against 29.28%, 44.66% and 50.86%.[2]

Weight moved further, and not uniformly across sources

The abstract reports least-squares mean estimated treatment differences in body weight of −1.9 kg, −3.6 kg and −5.5 kg for the three tirzepatide doses against semaglutide, P < 0.001 for all comparisons.[1] The posted results give per-arm changes of −7.8, −10.3 and −12.4 kg against −6.2 kg, with differences of −1.7 kg (95% CI, −2.6 to −0.7), −4.1 kg (95% CI, −5.0 to −3.2) and −6.2 kg (95% CI, −7.1 to −5.3).[2]

Note the direction of the discrepancy. At 15 mg the registry figure is larger than the abstract’s; at 5 mg it is smaller, 1.7 kg against 1.9 kg. Whatever produces the difference does not simply inflate every number, which is a reason to quote the source alongside the figure rather than picking whichever is more favorable. Weight loss of at least 5% was reached by 58.44% on semaglutide against 68.55%, 82.35% and 86.21% on tirzepatide, odds ratios 1.58, 3.49 and 4.60.[2] What a given percentage tends to mean in practice is what the expected weight-loss tool works through.

A later independent re-analysis of the trial data put the same result in a composite frame. Of participants on semaglutide, 34% met three or more standard therapeutic targets across glycated hemoglobin, blood pressure, LDL cholesterol and weight loss, against 42%, 53% and 57% on the three tirzepatide doses; on an intensive set of targets, 8% against 15%, 20% and 29%. The odds ratio for more than 10% weight loss was 2.72 (95% CI, 2.14 to 3.47) and for more than 15%, 3.86 (95% CI, 2.69 to 5.55).[4]

The columns that run the other way

Serious adverse events occurred in 13 of 469 participants on semaglutide against 33 of 470, 25 of 469 and 27 of 470 on tirzepatide 5, 10 and 15 mg — 2.8% against 7.0%, 5.3% and 5.7%. Deaths were recorded in 1 participant on semaglutide and 4 in each tirzepatide arm.[2] The abstract reports the same shape in rounded terms: serious adverse events in 5 to 7% of tirzepatide patients and 3% of semaglutide patients.[1]

These figures need their limits stated in the same sentence. SURPASS-2 ran 40 weeks in 1,879 people and recorded 13 deaths in total. It was not designed, powered or analyzed to compare mortality or serious events between arms, no causal attribution was made, and a difference of this size in absolute terms is entirely compatible with chance. What is not compatible with the usual summary is the claim that the two drugs differed only in efficacy. They did not. The abstract records hypoglycemia with a blood glucose below 54 mg per deciliter in 0.4% of those on semaglutide against 0.6%, 0.2% and 1.7% on tirzepatide 5, 10 and 15 mg, which is more than four times the semaglutide rate at the top dose and below it at 10 mg.[1]

The gastrointestinal picture is closer than the marketing suggests. Nausea affected 84 of 469 on semaglutide against 82 of 470, 90 of 469 and 104 of 470 on tirzepatide; diarrhea 54 against 62, 77 and 65; vomiting 39 against 27, 39 and 46.[2] Tolerability is the variable that decides most real outcomes, since a drug someone stops produces no result at any dose.

Who the trial was run in

A separate analysis applied SURPASS-2’s own enrollment criteria to the National Health and Nutrition Examination Survey for 2013 to 2018 and found 2,991,003 US adults (95% CI, 2,496,723 to 3,485,282) would qualify — adults with type 2 diabetes, glycated hemoglobin between 7.0% and 10.5%, on metformin and nothing else, with a body-mass index of 25 or above. Permitting other antidiabetes agents raised it to 4,937,063.[3]

The composition differs sharply too. That analysis records the trial cohort as 70.1% Hispanic and 4.2% Black against an eligible US population that is 17.6% Hispanic and 10.5% Black, with a mean weight of 93.7 kg in the trial against 100.6 kg in the eligible population.[3] None of that invalidates the result. It does mean the trial population is not the population most American readers belong to, and metformin monotherapy is a narrower starting point than most people arriving at a telehealth checkout are on.

What SURPASS-2 did not show

It did not compare the weight-management doses of either drug, because it did not use them. It did not enroll anyone without type 2 diabetes. It did not measure cardiovascular outcomes, kidney outcomes or anything beyond 40 weeks. And it was not blinded, which is the weakest point in any trial reporting subjective tolerability.

The obesity head-to-head exists separately. SURMOUNT-5 randomized 751 adults with obesity and without type 2 diabetes to the maximum tolerated dose of tirzepatide or of semaglutide for 72 weeks, and reported least-squares mean weight change of −20.2% (95% CI, −21.4 to −19.1) against −13.7% (95% CI, −14.9 to −12.6), P < 0.001, with waist circumference falling 18.4 cm against 13.0 cm.[5] That trial, and what switching between the molecules involves, is covered in the switching article.

What was in the pen

Every figure here belongs to branded, FDA-approved tirzepatide and semaglutide, supplied free, escalated on a fixed protocol, in people already taking metformin under trial supervision. Sellers on the tirzepatide board largely 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 practical reading is narrow and worth stating exactly. SURPASS-2 establishes that tirzepatide lowers glycated hemoglobin and body weight more than semaglutide at 1 mg over 40 weeks in metformin-treated type 2 diabetes, with more serious adverse events recorded in the tirzepatide arms. It establishes nothing about the compounded versions most sellers actually supply, and nothing about which of the two will work better for any individual — a question no head-to-head trial has ever been able to answer.

Frequently asked

Did SURPASS-2 prove tirzepatide is better than semaglutide for weight loss?
It proved tirzepatide produced more weight loss than semaglutide at 1 mg over 40 weeks in people with type 2 diabetes taking metformin, with estimated treatment differences of 1.9, 3.6 and 5.5 kg in the published abstract. It did not compare the weight-management doses of either drug and enrolled nobody without diabetes. The obesity head-to-head is SURMOUNT-5, which reported 20.2% against 13.7% at 72 weeks.
Why do published and registry figures for SURPASS-2 differ?
They describe different analysis populations. The posted trial results state that their analysis excludes data collected after a participant started rescue antihyperglycemic medication or prematurely stopped the study drug, and report glycated hemoglobin differences of 0.23, 0.51 and 0.60 points. The published abstract reports 0.15, 0.39 and 0.45 points. The gap is roughly a third larger once people who stopped or needed rescue treatment are removed.
Was the lowest tirzepatide dose meaningfully better than semaglutide?
Statistically yes, practically barely. Tirzepatide 5 mg lowered glycated hemoglobin by 0.15 percentage points more than semaglutide 1 mg, with a 95% confidence interval of 0.28 to 0.03. Both arms moved by more than two points from a baseline of 8.28%, so the advantage is about one seventh of a point on a measure that changed substantially in both groups.
Were there more side effects on tirzepatide in SURPASS-2?
The posted results record serious adverse events in 13 of 469 participants on semaglutide against 33 of 470, 25 of 469 and 27 of 470 on tirzepatide 5, 10 and 15 mg, and deaths in 1 participant against 4 in each tirzepatide arm. The trial was not designed or powered to compare either outcome between arms, and with 13 deaths in total over 40 weeks the difference is compatible with chance. Nausea rates were similar across all four arms.
How many people match the SURPASS-2 population?
An analysis applying the trial's enrollment criteria to national survey data for 2013 to 2018 estimated 2,991,003 US adults would qualify — type 2 diabetes, glycated hemoglobin of 7.0% to 10.5%, on metformin alone, body-mass index 25 or above. The trial cohort was also 70.1% Hispanic and 4.2% Black against an eligible US population that is 17.6% Hispanic and 10.5% Black.
Does SURPASS-2 say anything about compounded tirzepatide?
No. Every figure in it belongs to branded, FDA-approved tirzepatide and semaglutide supplied free and escalated on a fixed protocol under trial supervision. Compounded preparations are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed, so a head-to-head between two approved products establishes nothing about what a compounding pharmacy ships.

Sources

  1. [1] Frías JP, Davies MJ, Rosenstock J, et al. (2021). Tirzepatide versus Semaglutide Once Weekly in Patients with Type 2 Diabetes. N Engl J Med. PMID 34170647
  2. [2] Eli Lilly and Company (2021). A Study of Tirzepatide (LY3298176) Versus Semaglutide Once Weekly as Add-on Therapy to Metformin in Participants With Type 2 Diabetes (SURPASS-2): posted study results, NCT03987919. ClinicalTrials.gov. Source
  3. [3] Chiu N, Aggarwal R, Bhatt DL (2022). Generalizability of the SURPASS-2 Trial and Effect of Tirzepatide on US Diabetes and Obesity Control. J Am Heart Assoc. PMID 35929468
  4. [4] Neves JS, Leite AR, Vale C, et al. (2026). Efficacy of tirzepatide versus semaglutide in achieving therapeutic targets in type 2 diabetes: a post hoc analysis of the SURPASS-2 Trial. Diabetologia. PMID 41419618
  5. [5] Aronne LJ, Horn DB, le Roux CW, et al. (2025). Tirzepatide as Compared with Semaglutide for the Treatment of Obesity. N Engl J Med. PMID 40353578

Where to get it

Best GLP-1 injections

Every injectable seller we can verify, with the price each one publishes and an honest read of what the trials measured.

Compare providers →

More in Evidence