SURPASS-1 is the cleanest experiment in the tirzepatide program and the one that produced its smallest numbers. Every other trial in the series gave the drug a job on top of something else — metformin, a sulfonylurea, basal insulin, or a competing injectable — which means the comparison always contains someone else’s drug. Here there is nothing underneath: people with type 2 diabetes managed on diet and exercise alone were randomized to tirzepatide or to a matching placebo, and the columns show what the molecule does by itself. The head-to-head against semaglutide is in the SURPASS-2 article, and the ladder as a prescribing decision is in the tirzepatide dose article. This page is about what happens when the ladder is watched in isolation, because in two of the trial’s own endpoints it does not climb.
What was run
A 40-week, double-blind, randomized, placebo-controlled phase 3 trial at 52 medical research centers and hospitals in India, Japan, Mexico and the United States assessed 705 people and randomly assigned 478 in a 1:1:1:1 ratio by computer-generated sequence to once-weekly tirzepatide 5, 10 or 15 mg or to placebo: 121, 121, 121 and 115. Participants, investigators and the sponsor were all masked.[1]
Entry required type 2 diabetes inadequately controlled by diet and exercise alone, no injectable diabetes therapy ever and no oral antihyperglycemic medication in the three months before screening, a glycated hemoglobin of 7.0% to 9.5%, stable weight within 5% for three months, and a body-mass index of at least 23.[2] Mean baseline glycated hemoglobin was 7.9%, mean age 54.1 years (SD 11.9), diabetes duration 4.7 years, mean body-mass index 31.9, and 231 of 478 participants (48%) were women.[1] Enrollment ran 164 in Mexico, 140 in the United States, 89 in Japan, 73 in India and 12 in Puerto Rico; 22 of the 478 were Black or African American.[2]
The preregistered primary endpoint was a single quantity: the mean change in glycated hemoglobin from baseline at 40 weeks.[1] Weight is a secondary endpoint here, which is worth holding on to, because weight is the only thing most people quoting this trial care about.
The primary endpoint, and the gap between the first two doses
Glycated hemoglobin fell 1.87 points (SE 0.094), 1.89 (0.096) and 2.07 (0.098) across the 5, 10 and 15 mg arms, while the placebo arm rose 0.04 (0.105). The estimated treatment differences were −1.91 (95% CI, −2.18 to −1.63), −1.93 (−2.21 to −1.65) and −2.11 (−2.39 to −1.83), each at p < 0.001.[2]
Those are large effects and two of them are indistinguishable from each other. Doubling the dose from 5 to 10 mg moved the primary endpoint by 0.02 points, inside intervals more than half a point wide. Tripling it to 15 mg bought 0.20 points over the starting dose. On the trial’s own primary measure, most of what tirzepatide does to blood sugar is already done at the lowest maintenance dose.
The threshold endpoints make the same point more sharply. Reaching below 7.0% ran 86.78%, 91.53% and 87.93% against 19.64% on placebo, with odds ratios of 49.00 (95% CI, 21.12 to 113.67), 80.39 (31.80 to 203.19) and 52.95 (22.30 to 125.73) — the 15 mg arm below the 10 mg arm. Reaching a nondiabetic 5.7% ran 33.88%, 30.51% and 51.72% against 0.89%.[2] The middle dose scored below the lowest one. None of those gaps is large relative to the intervals around them, which is the honest reading: within a 121-person arm the ordering of adjacent doses is not reliably resolved.
Weight, the column everyone quotes
Body weight fell 7.0 kg (SE 0.52), 7.8 (0.53) and 9.5 (0.54) against 0.7 kg on placebo, with estimated treatment differences of −6.3 (95% CI, −7.8 to −4.7), −7.1 (−8.6 to −5.5) and −8.8 (−10.3 to −7.2).[2] This is the one headline column that climbs cleanly with dose, and it is the reason the trial gets cited at all outside diabetes care.
It does not climb everywhere either. The proportion losing at least 5% of body weight was 66.94%, 77.97% and 76.72% against 14.29% on placebo — the top dose again marginally below the middle one.[2] And the trial cannot be quoted as a percentage at all: neither the abstract nor the posted results reports a baseline body weight or a percentage weight change, so 9.5 kg is the whole of what this trial establishes about magnitude. What the obesity trials of the same molecule found, in people selected for weight rather than for glucose, is in the tirzepatide weight-loss article.
The placebo comparison is unusually informative here because the placebo arm barely moved: 0.7 kg over forty weeks, in people receiving no diabetes medication of any kind. There is no lifestyle program inflating the control arm and no competing drug deflating it. Whatever the reader makes of 9.5 kg, almost all of it belongs to the injection.
The safety table runs the other way
Serious adverse events occurred in 5 of 121, 2 of 121 and 1 of 121 participants across the ascending tirzepatide doses, against 3 of 115 on placebo, and the trial’s single death was in the placebo group.[2][1] That is the opposite of a dose-response and it should not be read as one: these are single-digit counts in a trial powered for a glucose endpoint, not a safety comparison, and an arm with one serious event and an arm with five are not distinguishable at this size.
The more interesting number is the total. Participants reporting at least one adverse event above the 5% reporting threshold ran 50, 55, 53 and 54 across the four arms — essentially flat, including placebo.[2] The symptom mix is what separates them. Nausea was 14, 16 and 22 against 7; diarrhea 14, 17 and 14 against 9; vomiting 4, 3 and 7 against 2; constipation 7, 6 and 8 against 1; decreased appetite 5, 8 and 10 against 1. The abstract renders the same picture as nausea 12–18% against 6%, diarrhea 12–14% against 8%, and vomiting 2–6% against 2%.[1]
And the placebo arm’s leading adverse event was hyperglycemia: 31 of 115, against 4, 5 and 3 on tirzepatide.[2] More than a quarter of the control group had their own untreated diabetes recorded as an adverse event. A comparison that reports only the gastrointestinal rows describes one arm’s side effects and none of the other’s.
Hypoglycemia, where the two sources disagree
The abstract states that no clinically significant hypoglycemia below 54 mg/dL and no severe hypoglycemia were reported with tirzepatide.[1] The posted results give a rate of 0.02 episodes per participant-year (SE 0.002) in each of the three tirzepatide arms against 0.04 (0.028) on placebo.[2] Those two statements are not compatible as written, and a reader should quote one source or the other rather than a figure from each.
What both accounts agree on is the direction and the order of magnitude: low blood sugar was rare in every arm, and no arm looked worse than placebo. That is the expected behavior for a drug whose insulin stimulation is glucose-dependent, given without a sulfonylurea or insulin underneath it, and the general picture is in the hypoglycemia article. It is also a claim that does not survive being carried into a trial with background insulin, where the arithmetic changes entirely.
Who left, and from which arm
The abstract reports that 66 participants (14%) discontinued the study drug and 50 (10%) discontinued the study prematurely, without breaking the first figure out by arm.[1] The registry posts the second: 7, 9, 18 and 16 did not complete, across 5 mg, 10 mg, 15 mg and placebo.[2]
The 15 mg arm has the worst retention of the three active arms, and the reason is not adverse events. Departures attributed to an adverse event were 1, 3, 1 and 1. Withdrawal by the participant — leaving by choice, with no clinical reason recorded — was 2, 3, 13 and 6.[2] The dose with the best glucose and weight columns is the dose the most people walked away from, and the trial records no reason why. Thirteen people is a small number and no test is attached to it; it is also more than the other two tirzepatide arms combined.
A posted analysis that contradicts its own table
One row in the registry should be read with care. Fasting serum glucose fell 43.6, 45.9 and 49.3 mg/dL across the tirzepatide arms and rose 12.9 on placebo — a gap of well over fifty points. The posted analyses against placebo for that same outcome report least-squares mean differences of −1.5, −2.6 and −0.6 at p = 0.776, 0.622 and 0.908.[2] Those two halves of the same table cannot both be right, and the abstract states that all three doses were superior to placebo on fasting serum glucose.[1] The arm means are the figures to use; the analysis rows look like a posting error and are reported here so that nobody mistakes them for a null result.
What SURPASS-1 was not built to show
It ran 40 weeks. Nothing in it speaks to a second year, to maintenance, or to what happens when the injections stop. It adjudicated no cardiovascular outcome and reported no event composite, and with 478 participants and eleven serious events in total it could not have. It measured no body composition: no imaging, no lean mass, no bone. It ran no active comparator — no metformin arm, no GLP-1 receptor agonist, nothing but placebo — so it establishes that tirzepatide beats nothing, not that it beats anything.
The population is the larger limit. Everyone enrolled had type 2 diabetes with a glycated hemoglobin between 7.0% and 9.5%, had taken no diabetes drug of any kind for at least three months, and had a mean body-mass index of 31.9 at a threshold that began at 23. Two thirds were enrolled outside the mainland United States. The trial contains no one without diabetes, no one on metformin, and no one selected for weight. Someone buying tirzepatide online to lose thirty pounds without a diabetes diagnosis is outside every entry criterion this trial had — a distinction worked through in the diabetes-against-weight-loss article.
One more caveat belongs on the record. The Lancet paper carries a published correction, filed as a Department of Error notice in the following issue. It has no abstract and is paywalled, so what it amended could not be established, and the figures above are quoted as the abstract and the registry state them.
What was in the pen
Every figure here belongs to branded, FDA-approved tirzepatide supplied free, escalated under protocol from 2.5 mg in 2.5 mg steps to a fixed assigned dose, injected weekly for 40 weeks under double-blind conditions, in adults with type 2 diabetes taking no other glucose-lowering drug. Sellers listed on the compounded tirzepatide board dispense compounded preparations, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they reach a patient.
The figure to be careful with is 9.5. It is the largest weight number this trial produced, at the highest dose, over forty weeks, in a diabetes population, and it is roughly half of what the obesity trials of the same molecule report. The trial’s own primary endpoint moved 0.02 points between the first and second doses. Whatever case exists for climbing the ladder, SURPASS-1 makes it on kilograms and not on blood sugar.