Semaglutide: what the STEP trials show and do not

STEP 1 to 8, STEP-HFpEF and SELECT with the real figures: 14.9% weight loss, HR 0.80 for CV events, two-thirds regained after stopping, lean mass by DXA.

Peptalis: information about peptide research. Peptalis does not sell products.

Peptalis does not sell products. This site only provides information about research.

ARTICLE · METABOLISM RESEARCH · EVIDENCE REVIEW

Nine phase 3 trials, one dose. What the STEP programme and SELECT established about semaglutide 2.4 mg, where the evidence stops, and what that means for reading the dual and triple agonists that followed.

MVMichel van der Veen — Founder of Peptalis Sources checked via PubMed · 17 referencesReviewed October 2026 · 13 MIN READ

Semaglutide is the compound on which the current class of weight-lowering incretin drugs, medicines that mimic gut hormones, was calibrated. How the molecule works, why it lasts a week and what GLP-1 does to appetite, stomach and pancreas is covered in the GLP-1 article in this series. This article is about the evidence itself: the STEP programme, a set of phase 3 trials that tested one dose, 2.4 mg subcutaneously once a week, for one standard duration, 68 weeks, in a series of populations, and the cardiovascular outcome trial SELECT that followed. That design also sets the limits of what the trials can show.

Illustration, schematic: fat under the skin and fat around the organs. In the DXA substudy of STEP 1, the fat around the organs fell the most in relative terms.

What the programme set out to test

Semaglutide first came to the clinic for type 2 diabetes, at 0.5 and 1.0 mg a week, in the SUSTAIN programme. The trial from that programme that matters here is SUSTAIN-6: 3297 patients with type 2 diabetes, most of them with established cardiovascular or kidney disease, randomised to semaglutide or placebo for 104 weeks. The composite of cardiovascular death, non-fatal infarction and non-fatal stroke occurred in 6.6 percent on semaglutide against 8.9 percent on placebo, a hazard ratio (the risk relative to placebo) of 0.74 with an interval of 0.58 to 0.95. Retinopathy complications (damage to the retina) rose, with a hazard ratio of 1.76 and an interval of 1.11 to 2.78 (Marso 2016). That trial was designed to rule out cardiovascular harm, not to prove benefit; the benefit it found set the expectation for everything that came after.

STEP, Semaglutide Treatment Effect in People with obesity, took the same molecule to a higher dose and a different population. The design is the same across the programme. The dose is 2.4 mg once a week, reached after 16 weeks of escalation. The duration is 68 weeks. Every participant, on drug or placebo, receives a lifestyle intervention: counselling on a 500 kcal deficit and 150 minutes of activity a week. The coprimary endpoints are the percentage change in body weight and the proportion losing at least 5 percent. And the primary analysis uses the treatment-policy estimand: everyone is counted in the group they were randomised to, whether or not they stayed on the drug.

That design answers one question very well: what does adding 2.4 mg of semaglutide to a modest lifestyle programme do, over a year and a quarter, in people recruited for a trial? Every other question has to be read against it.

What the trials show: weight

The anchor trial and the diabetes trial

STEP 1 is the anchor. 1961 adults with a BMI of 30 or higher, or 27 with a weight-related condition, without diabetes, were randomised 2:1 to semaglutide or placebo. At 68 weeks weight had fallen by 14.9 percent against 2.4 percent, a difference of 12.4 percentage points with an interval of 11.5 to 13.4; in kilograms, 15.3 against 2.6. At least 5 percent loss was reached by 86.4 against 31.5 percent, at least 10 percent by 69.1 against 12.0, and at least 15 percent by 50.5 against 4.9. Nausea and diarrhoea were the most common adverse events, mostly mild to moderate and transient; 4.5 percent stopped because of gastrointestinal complaints against 0.8 percent on placebo (Wilding 2021).

STEP 2 asked the same question in type 2 diabetes. 1210 adults with a BMI of 27 or higher and an HbA1c, a measure of average blood sugar over recent months, between 7 and 10 percent were randomised to 2.4 mg, to 1.0 mg, the diabetes dose, or to placebo. Weight fell by 9.6 percent on 2.4 mg against 3.4 percent on placebo, a difference of 6.2 percentage points with an interval of 5.2 to 7.3; 68.8 percent reached 5 percent loss against 28.5. Adverse events occurred in 87.6, 81.8 and 76.9 percent of the three groups and gastrointestinal events in 63.5, 57.5 and 34.3 percent (Davies 2021). The gap with STEP 1 is one of the most consistent findings in the field: people with type 2 diabetes lose roughly two-thirds of what people without diabetes lose on the same dose, and the programme does not explain why.

Lifestyle intensity, two years, and a different population

STEP 3 changed the background. 611 adults without diabetes at 41 sites in the United States received an eight-week low-calorie diet and 30 counselling sessions over 68 weeks, with semaglutide or placebo on top. Weight fell by 16.0 percent against 5.7 percent, a difference of 10.3 percentage points with an interval of 8.6 to 12.0. Gastrointestinal adverse events occurred in 82.8 against 63.2 percent, and 3.4 percent stopped for that reason against none on placebo (Wadden 2021). The more intensive lifestyle support mainly raised the result in the placebo group; on top of the drug it added little.

STEP 5 doubled the duration. 304 adults, 152 per arm, were followed for 104 weeks. Weight fell by 15.2 percent against 2.6 percent, a difference of 12.6 percentage points with an interval of 9.8 to 15.3; 77.1 percent reached 5 percent loss against 34.4. Gastrointestinal events occurred in 82.2 against 53.9 percent, and 92.8 percent of participants attended the end-of-trial visit. The population was 77.6 percent female and 93.1 percent white, with a mean BMI of 38.5 (Garvey 2022). Two years is the longest placebo-controlled weight follow-up in the programme.

STEP 6 tested the drug in Japan and South Korea, where obesity is defined at a lower BMI and body composition differs. 401 adults with a BMI of 27 or higher with two weight-related conditions, or 35 with one, with or without type 2 diabetes, received 2.4 mg, 1.7 mg or placebo. Weight fell by 13.2 and 9.6 percent against 2.1 percent, a difference of 11.1 points at the higher dose. In a subset scanned by CT, abdominal visceral fat area fell by 40.0 and 22.2 percent against 6.9 percent. Discontinuation for adverse events was 3 percent in both active groups against 1 percent (Kadowaki 2022). The effect size is close to STEP 1, in a population with a lower starting BMI.

Against the predecessor, and by mouth

STEP 8 is the only trial in the programme with an active comparator. 338 adults without diabetes were randomised, open-label (participants knew which drug they received), to weekly semaglutide 2.4 mg or daily liraglutide 3.0 mg, each with a small blinded placebo arm. Weight fell by 15.8 percent against 6.4 percent, a difference of 9.4 points with an interval of 6.8 to 12.0; pooled placebo lost 1.9 percent. At least 20 percent loss was reached by 38.5 against 6.0 percent. Discontinuation for any reason ran at 13.5 percent on semaglutide and 27.6 percent on liraglutide, with gastrointestinal events equally frequent at 84.1 and 82.7 percent (Rubino 2022). Same receptor, longer-acting molecule, more than twice the effect: this trial shows most clearly that pharmacokinetics, how long the molecule stays active, also determines what a GLP-1 agonist delivers.

OASIS 1 is a side step rather than part of STEP. 667 adults without diabetes took oral semaglutide 50 mg once a day, or placebo, for 68 weeks. Weight fell by 15.1 percent against 2.4 percent, a difference of 12.7 points with an interval of 11.3 to 14.2; gastrointestinal events occurred in 80 against 46 percent (Knop 2023). The numbers sit on top of STEP 1. By mouth the trial needed 50 mg a day, against 2.4 mg a week under the skin; the result hardly differs.

What the trials show: beyond weight

Blood pressure, waist circumference, lipids, HbA1c and C-reactive protein, a marker of inflammation, moved in the direction expected of weight loss across the programme, as secondary endpoints. Two trials measured outcomes that matter in themselves.

STEP-HFpEF took 529 patients with heart failure with preserved ejection fraction, a form in which the heart still pumps out a near-normal share of its blood per beat, and a BMI of 30 or higher. After 52 weeks the Kansas City Cardiomyopathy Questionnaire score, which runs from 0 to 100 and measures symptoms and physical limitation, had risen by 16.6 points against 8.7, a difference of 7.8 with an interval of 4.8 to 10.9. Six-minute walk distance improved by 21.5 metres against 1.2. Weight fell by 13.3 percent against 2.6, and C-reactive protein by 43.5 percent against 7.3. Serious adverse events were less frequent on semaglutide, 13.3 against 26.7 percent (Kosiborod 2023). This is a symptom and function trial, not a mortality trial, and it cannot say how much of the effect runs through weight and how much through something else.

SELECT is the outcome trial. 17,604 patients aged 45 or older with established cardiovascular disease, a BMI of 27 or higher and no diabetes were randomised to semaglutide 2.4 mg or placebo and followed for a mean of 39.8 months, with a mean exposure of 34.2 months. The composite of cardiovascular death, non-fatal infarction and non-fatal stroke occurred in 6.5 percent against 8.0 percent, a hazard ratio of 0.80 with an interval of 0.72 to 0.90. Permanent discontinuation for adverse events ran at 16.6 percent against 8.2 (Lincoff 2023). A relative reduction of 20 percent in hard events, in a secondary-prevention population without diabetes, is the strongest single result in the semaglutide literature. It applies to that population: people with prior infarction, stroke or peripheral arterial disease, mostly male, mean age over 60. The STEP 1 population was younger, mostly female and without cardiovascular disease.

Where the evidence stops

What happens when the drug stops

Two trials answer this, and they agree. STEP 4 gave 902 participants 20 weeks of semaglutide, by which point they had lost a mean of 10.6 percent, and then randomised the 803 who had reached 2.4 mg to continue or to switch to placebo for 48 weeks. Those who continued lost a further 7.9 percent; those who switched regained 6.9 percent, a difference of 14.8 points with an interval of 13.5 to 16.0. Waist circumference and systolic blood pressure separated in the same way (Rubino 2021).

The STEP 1 extension followed 327 participants for a year after the trial ended, with drug and lifestyle support both withdrawn. The semaglutide group had lost 17.3 percent at 68 weeks and regained 11.6 percentage points by week 120, a net loss of 5.6 percent from baseline. The placebo group ended at 0.1 percent. The improvements in cardiometabolic markers reverted towards baseline for most variables (Wilding 2022). Two-thirds of the weight came back within a year. The data show that the effect is tied to the presence of the drug; the programme contains no evidence on what a tapered dose, a lower maintenance dose or a structured transition to lifestyle alone would do.

What the weight is made of

The STEP trials measured weight, not composition. One substudy of STEP 1 used DXA (dual-energy X-ray absorptiometry), a body scan that measures fat and lean mass, in 140 participants with a BMI of 40 or lower, 95 on semaglutide and 45 on placebo. On semaglutide, total fat mass fell by 19.3 percent and visceral fat mass by 27.4 percent; lean body mass fell by 9.7 percent. Because fat fell faster than lean mass, the proportion of lean mass rose by about 3 percentage points. In absolute terms the trial publication reports a fat loss of 8.36 kg against 1.37 on placebo and a lean loss of 5.26 kg against 1.83 (Wilding 2021). Of the weight lost, roughly two-fifths was lean mass.

The earlier SUSTAIN 8 substudy, 178 people with type 2 diabetes on metformin, compared semaglutide 1.0 mg with canagliflozin by DXA for 52 weeks and found fat mass down 3.4 and 2.6 kg and lean mass down 2.3 and 1.5 kg, with no placebo arm to anchor either (McCrimmon 2020). Dubin and colleagues reviewed 28 trials of GLP-1-based agents with DXA data and found the lean fraction of weight lost ran between 20 and 40 percent, with most trials above 25 percent, which is the figure usually quoted for dietary weight loss. They also state the limitation plainly: DXA lean mass is not muscle, it includes organs, bone and water, and muscle is only about 55 percent of it (Dubin 2024). Neeland and colleagues reach a more reassuring reading from the few MRI studies available, in which muscle volume fell about as much as would be expected for the weight lost and muscle fat infiltration improved, which they call adaptive (Neeland 2024). Neither review can cite a trial of semaglutide 2.4 mg with muscle strength or physical function as a prespecified primary endpoint, because there is none. The question of whether these drugs cost muscle that matters is open, and the STEP programme was not designed to close it.

Populations, duration, comparators

The populations are narrower than the headline suggests. The trials excluded people with a history of pancreatitis, with type 1 diabetes, with recent cardiovascular events outside SELECT, and in STEP 1 and 3 with diabetes of any kind. The participants were mostly women, mostly white outside STEP 6, with a mean age in the mid-forties and a mean BMI around 38. There is no trial in people over 75, in people with a BMI below 27, or in people whose aim is to lose a few kilograms rather than a sixth of their body weight.

The duration is 68 weeks, with one trial at 104 and SELECT at a mean of 40 months of follow-up. Beyond that there is no randomised evidence on weight. The comparators are placebo plus modest lifestyle advice, and once intensive behavioural therapy, for 68 weeks. No trial compares semaglutide with a sustained, well-resourced lifestyle programme over several years, and no trial compares it with bariatric surgery. The within-programme comparison that exists, STEP 3 against STEP 1, suggests intensive support shifts the placebo arm far more than the drug arm, but it is a comparison across trials, not within one.

Adverse events and who leaves

The adverse-event profile is consistent: gastrointestinal complaints in 60 to 85 percent on drug against 30 to 63 percent on placebo, mostly during escalation, mostly mild to moderate. Discontinuation for adverse events runs at 3 to 7 percent in the 68-week trials. In SELECT, with older and sicker participants over a longer period, 16.6 percent stopped the drug for adverse events against 8.2 percent on placebo. What the trials cannot show is what happens in the people they did not enrol, and how the people who stopped, 13.5 and 27.6 percent in the two arms of STEP 8, would have done had they stayed.

How this sets the measure for what followed

The STEP programme fixed the reference point against which the next molecules are read. Tirzepatide, which adds GIP activity, produced 15.0, 19.5 and 20.9 percent weight loss at 5, 10 and 15 mg against 3.1 percent on placebo over 72 weeks in SURMOUNT-1, in 2539 adults without diabetes (Jastreboff 2022). Retatrutide, which adds glucagon activity as well, produced 24.2 percent at 12 mg against 2.1 percent over 48 weeks in its phase 2 trial of 338 adults (Jastreboff 2023). Both are discussed in the articles on combining GLP-1 with GIP, on glucagon, and on retatrutide in this series. The semaglutide evidence shows how to read them. The newer molecules were tested with the same design, the same estimand and largely the same populations, so the weight figures are comparable. They inherit the same limits: short duration, on-treatment evidence only, composition measured in substudies, and for retatrutide no outcome trial yet. Semaglutide has SELECT and more years of follow-up; on that point it still has more evidence than its successors.

What stands above are outcomes of clinical trials with the pharmaceutical drugs semaglutide, tirzepatide and retatrutide, in the populations and regimens described.

About the authorMVMichel van der VeenFounder of Peptalis

Founder of Peptalis. Writes the platform's knowledge layer: compound profiles and evidence reviews. Works from primary literature (PubChem for chemistry, PubMed for studies) and states where evidence is absent.

Sources checked via PubMed · 17 references

Compound profilesEvidence reviewsResearch phasesIN THIS ARTICLE
  • What the programme set out to test
  • What the trials show: weight
  • What the trials show: beyond weight
  • Where the evidence stops
  • How this sets the measure for what followed
SCIENTIFIC REFERENCESView all references →RELATED COMPOUNDRetatrutide →RELATED ARTICLES

SCIENTIFIC REFERENCES

These references are provided for informational and research purposes only. They do not constitute medical advice.

  1. 01Human RCTOnce-Weekly Semaglutide in Adults with Overweight or Obesity.Wilding JPH, et al. N Engl J Med. 2021.View on PubMed →DOI →
  2. 02Human RCTSemaglutide 2·4 mg once a week in adults with overweight or obesity, and type 2 diabetes (STEP 2): a randomised, double-blind, double-dummy, placebo-controlled, phase 3 trial.Davies M, et al. Lancet. 2021.View on PubMed →DOI →
  3. 03Human RCTEffect of Subcutaneous Semaglutide vs Placebo as an Adjunct to Intensive Behavioral Therapy on Body Weight in Adults With Overweight or Obesity: The STEP 3 Randomized Clinical Trial.Wadden TA, et al. JAMA. 2021.View on PubMed →DOI →
  4. 04Human RCTEffect of Continued Weekly Subcutaneous Semaglutide vs Placebo on Weight Loss Maintenance in Adults With Overweight or Obesity: The STEP 4 Randomized Clinical Trial.Rubino D, et al. JAMA. 2021.View on PubMed →DOI →
  5. 05Human RCTTwo-year effects of semaglutide in adults with overweight or obesity: the STEP 5 trial.Garvey WT, et al. Nat Med. 2022.View on PubMed →DOI →
  6. 06Human RCTSemaglutide once a week in adults with overweight or obesity, with or without type 2 diabetes in an east Asian population (STEP 6): a randomised, double-blind, double-dummy, placebo-controlled, phase 3a trial.Kadowaki T, et al. Lancet Diabetes Endocrinol. 2022.View on PubMed →DOI →
  7. 07Human RCTEffect of Weekly Subcutaneous Semaglutide vs Daily Liraglutide on Body Weight in Adults With Overweight or Obesity Without Diabetes: The STEP 8 Randomized Clinical Trial.Rubino DM, et al. JAMA. 2022.View on PubMed →DOI →
  8. 08Human RCTSemaglutide in Patients with Heart Failure with Preserved Ejection Fraction and Obesity.Kosiborod MN, et al. N Engl J Med. 2023.View on PubMed →DOI →
  9. 09Human RCTSemaglutide and Cardiovascular Outcomes in Obesity without Diabetes.Lincoff AM, et al. N Engl J Med. 2023.View on PubMed →DOI →
  10. 10Human studyWeight regain and cardiometabolic effects after withdrawal of semaglutide: The STEP 1 trial extension.Wilding JPH, et al. Diabetes Obes Metab. 2022.View on PubMed →DOI →
  11. 11Human RCTSemaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes.Marso SP, et al. N Engl J Med. 2016.View on PubMed →DOI →
  12. 12Human RCTEffects of once-weekly semaglutide vs once-daily canagliflozin on body composition in type 2 diabetes: a substudy of the SUSTAIN 8 randomised controlled clinical trial.McCrimmon RJ, et al. Diabetologia. 2020.View on PubMed →DOI →
  13. 13ReviewGlucagon-like peptide-1 receptor agonist-based agents and weight loss composition: Filling the gaps.Dubin RL, et al. Diabetes Obes Metab. 2024.View on PubMed →DOI →
  14. 14ReviewChanges in lean body mass with glucagon-like peptide-1-based therapies and mitigation strategies.Neeland IJ, et al. Diabetes Obes Metab. 2024.View on PubMed →DOI →
  15. 15Human RCTOral semaglutide 50 mg taken once per day in adults with overweight or obesity (OASIS 1): a randomised, double-blind, placebo-controlled, phase 3 trial.Knop FK, et al. Lancet. 2023.View on PubMed →DOI →
  16. 16Human RCTTirzepatide Once Weekly for the Treatment of Obesity.Jastreboff AM, et al. N Engl J Med. 2022.View on PubMed →DOI →
  17. 17Human RCTTriple-Hormone-Receptor Agonist Retatrutide for Obesity: A Phase 2 Trial.Jastreboff AM, et al. N Engl J Med. 2023.View on PubMed →DOI →

Peptalis is a knowledge platform. All content on this page is provided for scientific and educational reference purposes only and does not constitute medical advice. Compounds discussed are research compounds and are not intended for human or veterinary use.