How to Read GLP-1 Clinical Trials: 8 Checks + 2026 Trial Matrix
Published:
· Re-verified quarterly — next check August 2026The bottom line
To read GLP-1 clinical trials without getting tricked by the headline, run eight quick checks: who was studied, what drug and dose were used, what it was compared to, how long the trial ran, which endpoint was primary, what the actual effect size was, how missing data and dropouts were handled, and what the safety numbers showed.
The single biggest trick most people miss? Modern GLP-1 trials usually report two different weight-loss numbers from the exact same data. STEP 1 reported both −14.9% and −16.9% mean weight loss for semaglutide 2.4 mg. Both are correct. They answer different questions. When a claim quotes only the higher number, check which estimand it is from before repeating it.
Once you can spot that move, every other GLP-1 claim — Zepbound, Mounjaro, Foundayo, compounded versions, telehealth ads, social-media debates — gets a lot easier to read. We will show you exactly how to do it, with verified numbers from STEP, SURMOUNT, SURPASS, ATTAIN, and SELECT.
If the headline says this, ask this first
A 30-second filter before you trust any GLP-1 number you see online.
| If the headline says… | First question to ask |
|---|---|
| "Patients lost 22.5% of their body weight" | Which estimand? The trial-product/efficacy-style number is usually 1–2 percentage points higher than the treatment-policy number. |
| "20% lower risk of heart attack and stroke" | Is that relative risk or absolute risk? In SELECT, "20% lower" was the gap between 6.5% and 8.0% events — a 1.5-point absolute difference. |
| "Tirzepatide beats semaglutide" | Head-to-head trial or two separate trials being compared? Only one of those proves it. |
| "People regain all the weight when they stop" | A withdrawal trial in selected responders, or a real-world cohort? Those tell different stories. |
| "Average weight loss: 27 lbs" | At what dose, in what population, over how many weeks, with what dropout rate? |
| "Compounded semaglutide is the same as Wegovy" | The FDA sent warning letters to 30 telehealth firms in February 2026 for that exact wording. Compounded drugs are not FDA-approved. |
Who this guide is for
Curious patient
You saw a Wegovy, Zepbound, or Foundayo headline and want to know if the number applies to you.
Already on a GLP-1
Your real results do not match the trial average and you want to know why.
Skeptical reader
A telehealth ad or a friend's claim sounded too good. You want a way to check it.
Healthcare-adjacent
You are a coach, RD, nurse, or journalist who needs to interpret a press release without faking expertise.
Investor or analyst
You read NEJM abstracts but do not have the GLP-1-specific shorthand.
How to read GLP-1 clinical trials in 8 checks
Reading a GLP-1 clinical trial means working through the trial's population, intervention, comparator, endpoint, duration, effect size, statistical method, and safety data — instead of stopping at the headline percentage. GLP-1 trials in 2026 (STEP, SURMOUNT, SURPASS, SELECT, ATTAIN) typically report two valid weight-loss numbers from the same data: one for everyone who was randomized, and another estimated under the assumption that everyone took the drug as intended.
- 1Who was actually in the trial?
- 2What drug, dose, and route?
- 3What was the comparator?
- 4What was the primary endpoint?
- 5How long was the trial?
- 6Which estimand did the headline use?
- 7What about dropouts and adverse events?
- 8Does the trial population match you?
Check 1: Who was actually in the trial?
Answer capsule: Trial results depend heavily on the population studied. People without type 2 diabetes consistently lose more weight on every GLP-1 than people with diabetes do, on the same drug at the same dose. ATTAIN-1 studied orforglipron in 3,127 adults without diabetes and reported 11.1–12.4% mean weight loss; ATTAIN-2 studied the same drug in 1,613 adults with type 2 diabetes and reported 5.1–9.6%. The population determines whether a headline number applies to you.
The first thing to find in any GLP-1 paper is the participant table — usually Table 1 in the published study. Check:
- Diabetes status. Every major GLP-1 program runs separate trials for people with and without type 2 diabetes. The same drug consistently produces less weight loss in T2D patients.
- Baseline BMI. Most obesity trials require BMI ≥30, or ≥27 with at least one weight-related condition like high blood pressure, sleep apnea, or high cholesterol.
- Age, sex, race. STEP 1 was 74% female. SURMOUNT-1 was 67% female. Most major GLP-1 trials have been majority white. Real-world results across demographics may differ.
- Exclusions. Trials often exclude people with pancreatitis history, certain thyroid cancers, gallbladder disease, pregnancy, severe kidney disease, or recent bariatric surgery. Headlines almost never mention exclusions.
- Prior GLP-1 use. Some trials exclude experienced users; others enroll them. Experienced users may respond differently.
The quick test: If you cannot tell from the headline whether the participants were diabetic, what their baseline weight was, or how old they were, the headline is not telling you enough to apply the result to anyone.
Check 2: What drug, dose, and route?
Answer capsule: GLP-1 results vary by drug (semaglutide vs tirzepatide vs liraglutide vs orforglipron), dose (semaglutide is sold at 2.4 mg/week and was tested at 7.2 mg/week in STEP UP), and route (injection vs oral). Higher doses generally produce more weight loss. In the only head-to-head trial published to date (SURMOUNT-5), tirzepatide at maximum tolerated dose produced 20.2% mean weight loss vs 13.7% for semaglutide at maximum tolerated dose.
The molecules
| Molecule | Brands | Receptor target | Key program |
|---|---|---|---|
| Semaglutide | Ozempic, Wegovy, Rybelsus | GLP-1 only | STEP (obesity), SUSTAIN (diabetes) |
| Tirzepatide | Mounjaro, Zepbound | GLP-1 + GIP | SURMOUNT (obesity), SURPASS (diabetes) |
| Liraglutide | Saxenda, Victoza | GLP-1 only | SCALE (obesity), LEADER (CVD) |
| Dulaglutide | Trulicity | GLP-1 only | REWIND (CVD), SURPASS-CVOT comparator |
| Orforglipron | Foundayo | GLP-1 only (small molecule, oral) | ATTAIN (FDA-approved April 2026) |
Dose matters more than people think. STEP 1 used semaglutide 2.4 mg/week and produced 14.9% mean weight loss. STEP UP used semaglutide 7.2 mg/week — three times the dose — and produced 18.7%. Same drug, different dose, different result.
A note on Foundayo dosing. The FDA-approved Foundayo tablet maxes out at 17.2 mg once daily. The ATTAIN-1 trial reported results for investigational orforglipron capsules at 6 mg, 12 mg, and 36 mg, where the 36 mg capsule is equivalent to the 17.2 mg approved tablet because the tablet has different bioavailability. When you read "27 lbs lost on 36 mg" in news coverage, that is the trial dose; the equivalent prescription tablet is 17.2 mg.
The quick test: When you see "tirzepatide produces 22.5% weight loss," ask what dose. (Answer: 15 mg/week, the highest tested in SURMOUNT-1. The 5 mg dose produced 16%.)
Check 3: What was the comparator?
Answer capsule: A placebo-controlled trial answers "does the drug beat doing nothing structured?" An active-comparator trial answers "does the drug beat another existing drug?" A head-to-head trial like SURMOUNT-5 (tirzepatide vs semaglutide, both at maximum tolerated doses) is the only design that directly proves one drug works better than another.
Placebo-controlled trials
STEP 1, SURMOUNT-1, ATTAIN-1. "Placebo" in GLP-1 trials usually does not mean doing nothing — both groups get structured diet and activity counseling. That is why placebo arms typically lose 2–4% of body weight, not 0%.
Active-comparator trials
SURPASS-CVOT compared tirzepatide to dulaglutide. Stronger evidence for "is the new one better?" — but only against that specific comparator.
Head-to-head trials
SURMOUNT-5 (tirzepatide vs semaglutide, n=751, 72 weeks) is the gold standard for comparing two drugs directly. Without one, "Drug A is better than Drug B" is usually a cross-trial comparison — which controls fewer variables and proves less.
Non-inferiority designs
SURPASS-CVOT was designed as non-inferiority vs dulaglutide. It met non-inferiority for the primary 3-point MACE outcome (12.2% vs 13.1%, HR 0.92) but did not reach statistical superiority. "Non-inferior" means "not worse by more than the pre-set margin." It does not mean "better than."
The quick test: If a headline implies superiority but the comparator was placebo, that just means "better than not taking the drug" — not "better than other treatments."
Check 4: What was the primary endpoint?
Answer capsule: The primary endpoint is the specific outcome the trial was designed and powered to test. In GLP-1 obesity trials (STEP, SURMOUNT, ATTAIN), the co-primary endpoints are usually percent change in body weight and the proportion of participants reaching ≥5% weight loss. In cardiovascular outcomes trials (SELECT, SURPASS-CVOT), the primary endpoint is 3-point MACE. Secondary endpoints may be impressive but are not what the trial was built to demonstrate.
A trial gets designed around one or two primary endpoints. Anything else is secondary or exploratory. A drug can "miss" its primary endpoint but still produce a statistically significant improvement on some secondary outcome — and the press release will lead with that secondary outcome. That is not lying. But it is also not the result the trial was built to prove.
Co-primary endpoints in GLP-1 obesity trials
- Percent change in body weight from baseline to end of trial
- Proportion of participants achieving at least 5% body weight reduction
Secondary endpoints often include the proportion hitting ≥10%, ≥15%, ≥20%, and ≥25% weight loss. These are useful — they show distribution, not just averages — but they are not the primary win.
The quick test: If the headline emphasizes a secondary endpoint, find out whether the trial's primary endpoint also won. If the primary missed and only the secondary hit, that is a much weaker result.
Check 5: How long was the trial?
Answer capsule: Most major GLP-1 obesity trials run 68 to 104 weeks. STEP 1 ran 68 weeks; SURMOUNT-1, SURMOUNT-5, and ATTAIN-1 ran 72 weeks; STEP 5 ran 104 weeks. Weight loss in semaglutide trials typically reaches its nadir around week 60 and then plateaus. Trials shorter than 40 weeks are usually Phase 2 reads and may not predict long-term results.
| Duration | Trials |
|---|---|
| 68 weeks | STEP 1, STEP 3 |
| 72 weeks | SURMOUNT-1, SURMOUNT-2, SURMOUNT-5, ATTAIN-1, ATTAIN-2, STEP UP |
| 104 weeks | STEP 5 — the 2-year durability trial |
| Variable, event-driven | SELECT, SURPASS-CVOT — run until enough MACE events have occurred to test the hypothesis |
In STEP 1, semaglutide weight loss accelerated for the first 20 weeks, slowed through weeks 20–60, and reached its nadir around week 60. After that it was essentially maintained through week 68. In STEP 5, the same plateau-and-maintain pattern continued to week 104 with only modest additional regain.
The quick test: When was the weight measured? A 16-week result, a 40-week result, and a 72-week result are not interchangeable.
Check 6: Which estimand did the headline use?
This is the single most-misunderstood concept in GLP-1 reporting.
Modern GLP-1 trials report results under at least two statistical estimands — two precise definitions of the treatment effect. The treatment-policy estimand reflects what happened to everyone who was randomized, regardless of whether they stopped the drug or needed rescue therapy. The trial-product or efficacy-style estimand uses statistical modeling to estimate what the result would have been under the assumption that participants took the drug as intended. STEP 1 reported −14.9% under treatment-policy and −16.9% under trial-product. Both are correct. The trial-product number is often higher, and it is the one that usually makes the headline.
Treatment-policy estimand
(sometimes called treatment-regimen estimand)
Asks: "What is the average effect for everyone who was randomized to this drug — including people who quit, paused, or had to take something else?" Uses all the data, regardless of intercurrent events. This is the closer match to real-life outcomes.
Trial-product estimand
(sometimes called efficacy-style or hypothetical estimand)
Asks: "What is the average effect under the statistical assumption that everyone took the drug as intended?" A model-based estimate — not a simple count of completers. Usually higher than treatment-policy. The one headlines quote.
The key examples side by side
| Trial | Treatment-policy estimand | Trial-product estimand | Which one headlines quote |
|---|---|---|---|
| STEP 1 (sema 2.4 mg) | −14.9% | −16.9% | Trial-product (−16.9%) |
| SURMOUNT-1 (tirz 15 mg) | −20.9% (treatment-regimen) | −22.5% (efficacy) | Efficacy (−22.5%) |
| ATTAIN-1 (orforglipron 36 mg) | −11.1% (all randomized) | −12.4% (completers/efficacy) | Efficacy / "27 lbs" |
Which one should you use for your own expectations? For most people setting personal expectations, the treatment-policy number is closer to real life because it reflects real adherence and dropout patterns. But your individual outcome depends on dose, persistence, baseline weight, T2D status, and lifestyle co-intervention. No average is a prediction for one person.
The quick test: When a GLP-1 result is quoted, look at the trial's published methods section or supplementary appendix. It will say "treatment-policy estimand" or "trial-product/efficacy estimand" or both. When both numbers are reported, headlines almost always pick the larger one. Now you know.
Check 7: What about dropouts and adverse events?
Answer capsule: GLP-1 trials have meaningful dropout, mostly driven by gastrointestinal side effects. In STEP 1, adverse events occurred in 89.7% of the semaglutide arm vs 86.4% of placebo; serious adverse events in 9.8% vs 6.4%; discontinuation specifically due to gastrointestinal events in 4.5% vs 0.8%. Most GI events were mild to moderate and concentrated during dose escalation. Total discontinuation rates matter most for interpreting the trial-product vs treatment-policy gap.
Key adverse-event data from published trials
| Trial | Any AE (drug vs placebo) | Serious AEs | GI discontinuation |
|---|---|---|---|
| STEP 1 (sema 2.4 mg) | 89.7% vs 86.4% | 9.8% vs 6.4% | 4.5% vs 0.8% |
| STEP 4 (continued vs switched) | GI events 49.1% vs 26.1% in randomized phase | — | 2.4% vs 2.2% (similar) |
| SURMOUNT-2 (tirz, T2D) | GI ~51–57% vs 18% placebo | Mostly mild-moderate | Concentrated during titration |
| ATTAIN-1 (orforglipron) | Nausea, diarrhea, constipation, vomiting, abdominal pain | Consistent with GLP-1 class | Higher than placebo |
What to look for
- Common adverse events broken out by severity (mild, moderate, severe)
- Serious adverse events (SAEs) — usually defined as hospitalization, life-threatening, or causing significant disability
- Discontinuation due to adverse events — usually the most practical "tolerability" number
- Events of special interest — gallbladder events, pancreatitis, thyroid C-cell findings, retinopathy in diabetes trials
A note on the suicidality question. In January 2026, the FDA requested removal of the suicidal ideation and behavior warning from the labels of Saxenda (liraglutide), Wegovy (semaglutide 2.4 mg), and Zepbound (tirzepatide). The decision followed an FDA meta-analysis of 91 placebo-controlled trials with 107,910 participants, which found no increased risk. The FDA still recommends patients discuss any new or worsening depression, suicidal thoughts, or unusual mood/behavior changes with a healthcare professional.
The quick test: Find the discontinuation rate and the serious adverse event rate. If they are not in the abstract, find them in the supplementary appendix or the CONSORT diagram. Those are the two numbers most consumer summaries skip.
Check 8: Does the trial population match you?
Answer capsule: Trial averages describe trial participants. Your outcome depends on your specific characteristics. People with type 2 diabetes consistently lose less weight on every GLP-1 than people without diabetes. A JAMA Network Open cohort study of 125,474 US adults found that within one year of starting a GLP-1, 46.5% of patients with T2D and 64.8% without T2D had discontinued.
The quick test: Find Table 1 of the trial — the baseline characteristics table. Compare it to yourself. If you are materially different from the average participant, the trial average is a worse predictor for you.
Not sure which trial population you match?
A licensed clinician can help map your specific situation — diabetes status, BMI, cardiovascular history, prior GLP-1 use — to the right drug and dose for you. Our quiz walks through the key questions in 3 minutes.
Affiliate disclosure: we may earn a commission from provider links.
The 2026 GLP-1 Trial Truth Table
Verified against primary peer-reviewed publications where available, FDA documentation where not. Cells re-verified quarterly — next check August 2026.
Obesity trials — without type 2 diabetes
| Trial | Drug & dose | Population | n | Duration | Treatment-policy estimand | Trial-product estimand | Placebo result | Source |
|---|---|---|---|---|---|---|---|---|
| STEP 1 | Semaglutide 2.4 mg/wk SC | BMI ≥30 or ≥27 + comorbidity; no T2D | 1,961 | 68 wk | −14.9% | −16.9% | −2.4% | NEJM |
| STEP 5 | Semaglutide 2.4 mg/wk SC | BMI ≥30 or ≥27 + comorbidity; no T2D | 304 | 104 wk | −15.2% | Single primary estimand reported | −2.6% | Nature Med |
| STEP UP | Semaglutide 7.2 mg/wk SC | BMI ≥30; no T2D | ~1,400 | 72 wk | −18.7% (7.2 mg) vs −15.6% (sema 2.4 mg) | −20.7% (7.2 mg, efficacy) | −3.9% | PubMed |
| SURMOUNT-1 | Tirzepatide 15 mg/wk SC | BMI ≥30 or ≥27 + comorbidity; no T2D | 2,539 | 72 wk | −20.9% (treatment-regimen) | −22.5% (efficacy) | −3.1% / −2.4% | NEJM |
| SURMOUNT-5 | Tirzepatide MTD vs Semaglutide MTD | BMI ≥30 or ≥27 + comorbidity; no T2D | 751 | 72 wk | −20.2% (tirz) vs −13.7% (sema) | Single primary reported | Active comparator (no placebo) | NEJM 2025 |
| ATTAIN-1 | Orforglipron 36 mg/day investigational (≡ Foundayo 17.2 mg tablet) | BMI ≥30 or ≥27 + comorbidity; no T2D | 3,127 | 72 wk | −11.1% (all randomized) | −12.4% (completers/efficacy) | −2.1% | FDA approval |
Obesity + type 2 diabetes trials
| Trial | Drug & dose | Population | n | Duration | Headline weight loss | Comparator result | Source |
|---|---|---|---|---|---|---|---|
| SURMOUNT-2 | Tirzepatide 10/15 mg/wk SC | BMI ≥27, T2D, A1c 7–10% | 938 | 72 wk | −12.8% / −14.7% | −3.2% placebo | ACC |
| ATTAIN-2 | Orforglipron oral | Obesity/overweight + T2D | 1,613 | 72 wk | −5.1% to −9.6% by dose | −2.5% placebo | Foundayo label |
Note: ATTAIN-2 reports weight-loss data in adults with T2D enrolled because they had obesity or overweight. It is not a standalone FDA diabetes-treatment indication.
Withdrawal and maintenance trials
| Trial | Design | n | Duration | Result | Source |
|---|---|---|---|---|---|
| STEP 4 | 20-wk semaglutide run-in → randomize to continue or switch to placebo | 803 | Wk 20 to 68 | −7.9% (continued) vs +6.9% (placebo switch) | JAMA 2021 |
| SURMOUNT-3 | 12-wk lifestyle lead-in (had to lose ≥5%) → tirzepatide vs placebo | 579 | 72 wk after randomization | −18.4% vs +2.5% from randomization; −24.3% vs −4.5% from baseline (selected responders only) | Nature Med 2023 |
| SURMOUNT-4 | 36-wk tirzepatide lead-in → continue or switch to placebo | 670 | Wk 36 to 88 | −5.5% (continued) vs +14.0% (placebo switch); total wk 0–88: −25.3% vs −9.9% | PMC 2024 |
Cardiovascular outcome trials
| Trial | Drug | Population | n | Median follow-up | Primary endpoint (3-MACE) | HR | Source |
|---|---|---|---|---|---|---|---|
| SELECT | Semaglutide 2.4 mg/wk | Overweight/obese + established CVD; no T2D | 17,604 | ~40 mo | 6.5% vs 8.0% placebo | 0.80 (p<0.001) | NEJM 2023 |
| SURPASS-CVOT | Tirzepatide vs Dulaglutide | T2D + established ASCVD | ~13,000 | ~4 yr | 12.2% (tirz) vs 13.1% (dula) | 0.92 — non-inferior, not superior | ACC summary |
Note on completeness: ATTAIN-1 and ATTAIN-2 full peer-reviewed publications continue to roll out through 2026; the numbers above come from the FDA-approved Foundayo label and supporting sponsor materials and will be re-verified row by row when each full peer-reviewed paper is final.
Relative risk vs. absolute risk: the SELECT lesson
Answer capsule: SELECT reported a 20% lower relative risk of MACE. The absolute risk difference was 1.5 percentage points: 6.5% of semaglutide participants experienced MACE vs 8.0% on placebo. Both numbers are correct. Relative risk reductions sound larger; absolute risk reductions show you what actually changed. Always look for both.
SELECT enrolled 17,604 adults with overweight or obesity and established cardiovascular disease, but without type 2 diabetes. They were followed for a mean of about 40 months. The primary endpoint was 3-point MACE: cardiovascular death, non-fatal heart attack, or non-fatal stroke.
6.5%
Semaglutide group — MACE events
8.0%
Placebo group — MACE events
HR 0.80
20% relative reduction, 1.5 pp absolute
A hazard ratio of 0.80 is described as a "20% relative risk reduction." That is accurate — the rate of events in the semaglutide group was about 20% lower. But the absolute risk reduction was 1.5 percentage points. Of every 100 patients like the SELECT participants, treatment with semaglutide over ~40 months prevented roughly 1.5 first MACE events.
Marketing materials almost always lead with the relative reduction because 20% sounds bigger than 1.5 percentage points. Neither is wrong. Both should be on the page.
The quick test: If you see "20% lower risk" without two raw event rates next to it, ask for the event rates. They are always in the published paper.
For a full breakdown of every major GLP-1 cardiovascular trial, including 3-point vs 4-point vs 5-point MACE definitions and 14-trial NNT estimates, see our GLP-1 MACE cardiovascular trial decoder.
What withdrawal trials actually prove
Answer capsule: Withdrawal trials like STEP 4 and SURMOUNT-4 test what happens after initial weight-loss response when treatment continues vs stops. These trials show that GLP-1 effects depend on continued treatment — they do not prove that everyone who stops will regain 100% of their weight, only that meaningful regain occurs on average after discontinuation.
Misread 1: "Stopping means losing everything."
False. Withdrawal trials show meaningful regain on average, but not total regain, and not for everyone. Behavior, structure, persistence of lifestyle changes, and individual physiology all matter.
Misread 2: "The drug doesn't work because weight comes back."
Also false. The drug worked when people were on it. Obesity is classified as a chronic disease by the AMA, AHA, and Obesity Society — and chronic diseases generally need continued management.
STEP 4 (Rubino et al., JAMA 2021)
- · 803 adults completed a 20-week semaglutide run-in
- · Then randomized to continue semaglutide or switch to placebo for another 48 weeks
- · From week 20 to week 68: continued arm lost an additional 7.9%; placebo-switch arm regained 6.9%
SURMOUNT-4 (Aronne et al., JAMA 2024)
- · 670 adults completed a 36-week tirzepatide lead-in
- · Then randomized to continue tirzepatide or switch to placebo for another 52 weeks
- · From week 36 to week 88: continued arm lost an additional 5.5%; placebo-switch arm regained 14.0%
- · Total weight change week 0 to week 88: −25.3% (continued) vs −9.9% (placebo-switched)
Why "−24.3% weight loss" on tirzepatide is not a fair comparison
Answer capsule: SURMOUNT-3 reported total weight loss of 24.3% on tirzepatide, but the trial used a lifestyle lead-in design that only randomized people who had already lost at least 5% of body weight during the first 12 weeks. Roughly 17.5% of initial enrollees did not reach that threshold and were not randomized. The 24.3% figure measures total change from the very beginning of the lifestyle phase through 72 weeks of tirzepatide in selected responders.
How SURMOUNT-3 was designed
- All enrollees did 12 weeks of intensive lifestyle intervention (ILI) — counseling, diet plan, physical activity
- Only people who lost ≥5% body weight in those 12 weeks were randomized into the trial
- Roughly 17.5% of initial enrollees did NOT hit the 5% threshold and were excluded from randomization
- The randomized group then received tirzepatide or placebo for 72 more weeks
| Time window measured | Tirzepatide result | Placebo result |
|---|---|---|
| From randomization (week 12 of overall study) to week 72 of treatment | −18.4% | +2.5% |
| From original baseline (start of lifestyle phase) to week 72 total | −24.3% | −4.5% |
The −24.3% includes the lifestyle-only weight loss before tirzepatide was even started, and only applies to people who already proved they could lose ≥5% with structured lifestyle alone. The 18.4% randomized-to-week-72 number is the cleaner comparison to SURMOUNT-1's 20.9% treatment-regimen estimand.
The quick test: Any time you see a weight-loss number that looks unusually high compared to other trials of the same drug, check whether the trial used a lead-in design that selected responders before randomization.
FDA-approved vs. label-supported vs. investigational vs. compounded
Answer capsule: GLP-1 claims fall into four categories, each with different evidence weight. FDA-approved drugs have been reviewed for safety, efficacy, and quality. Label-supported uses are those the FDA has authorized in the prescribing information. Investigational claims describe drugs or doses still being studied. Compounded products are not FDA-approved finished drug products.
FDA-approved GLP-1 drugs as of May 2026
| Brand | Molecule | Approved use |
|---|---|---|
| Saxenda | Liraglutide | Weight management |
| Victoza | Liraglutide | Type 2 diabetes |
| Wegovy | Semaglutide 2.4 mg SC | Weight management; cardiovascular risk reduction in adults with CVD + obesity/overweight |
| Ozempic | Semaglutide SC (lower doses) | Type 2 diabetes; cardiovascular risk reduction in T2D |
| Rybelsus | Oral semaglutide | Type 2 diabetes; MACE risk reduction (Oct 2025 label) |
| Zepbound | Tirzepatide SC | Weight management; obstructive sleep apnea in adults with obesity |
| Mounjaro | Tirzepatide SC | Type 2 diabetes |
| Trulicity | Dulaglutide | Type 2 diabetes; cardiovascular risk reduction |
| Foundayo | Orforglipron (oral) | Chronic weight management — approved April 1, 2026 |
Investigational claims
Retatrutide (Eli Lilly's investigational triple agonist — GLP-1 + GIP + glucagon), CagriSema (Novo Nordisk's investigational cagrilintide + semaglutide combination), and the 36 mg orforglipron capsule used in ATTAIN-1 (the approved tablet maxes at 17.2 mg) are all investigational. Not FDA-approved.
Compounded products
Compounded GLP-1 products are made by compounding pharmacies and are not FDA-approved finished drug products. Per the FDA, some products sold as "semaglutide" use salt forms (semaglutide sodium or semaglutide acetate) which the agency considers different active ingredients than the semaglutide in approved drugs. On February 20, 2026, the FDA sent warning letters to 30 telehealth firms for advertising compounded GLP-1s in ways the agency found false or misleading.
Headline vs. reality: common claim patterns checked
Common claim patterns from GLP-1 marketing and journalism, each paired against what the primary source data actually shows.
| ❌ Common marketing or news claim | ✓ What the source data actually shows |
|---|---|
| "Patients lost 22.5% of their body weight on tirzepatide" | SURMOUNT-1, efficacy estimand, 15 mg dose. The all-randomized treatment-regimen number is 20.9%. The 5 mg dose produced 16.0%. (NEJM 2022) |
| "Average weight loss: 27 lbs on the highest dose" | ATTAIN-1, trial completers on the 36 mg investigational dose (equivalent to the approved 17.2 mg tablet). All-randomized average was ~25 lb. Placebo lost ~5 lb. (FDA approval, April 2026) |
| "Semaglutide cuts heart attacks and strokes by 20%" | SELECT primary 3-MACE outcome: 6.5% vs 8.0% placebo, HR 0.80. The 20% is a relative risk reduction; the absolute difference is 1.5 percentage points. (NEJM 2023) |
| "Tirzepatide is twice as effective as semaglutide" | SURMOUNT-5: 20.2% vs 13.7% at maximum tolerated doses — about 1.5× more weight loss, not 2×. Open-label, head-to-head, 72 weeks. (NEJM 2025) |
| "Stop the drug and you regain everything" | Withdrawal trials (STEP 4, SURMOUNT-4) show meaningful but not total regain. In STEP 4's placebo-switch group, weight increased 6.9% over 48 weeks after stopping — not back to baseline. |
| "Tirzepatide produces 24% weight loss" | SURMOUNT-3 total weight change from lifestyle start to week 72 in lifestyle responders only. From randomization to week 72 it is 18.4%. (Nature Medicine 2023) |
| "Tirzepatide is cardiovascular-superior to dulaglutide" | SURPASS-CVOT met non-inferiority for the primary 3-MACE outcome, HR 0.92. Did not reach statistical superiority. (ACC/NEJM 2025) |
| "Our compounded semaglutide has the same active ingredient as Wegovy" | The FDA has stated that some compounded products use salt forms (semaglutide sodium and semaglutide acetate) which the agency considers different active ingredients than those in approved drugs. The agency sent warning letters to 30 telehealth firms for this kind of wording on February 20, 2026. |
| "Foundayo is the only oral GLP-1 with no food or water restrictions" | Accurate. Foundayo's FDA-approved label allows once-daily dosing with or without food. Oral semaglutide (Rybelsus) requires a 30-minute fast and ≤4 oz of water. (FDA approval, April 2026) |
The pattern most commonly used in telehealth marketing: quoting completer/efficacy numbers without naming the estimand, paired with the highest dose tested. Once you can spot it, you can read these ads in seconds.
Ready to find a GLP-1 provider that uses honest language?
We vet telehealth platforms for pricing transparency, realistic expectation-setting, and clinical oversight. Here are providers our research team covers:
Affiliate disclosure: The RX Index may earn a commission from these links. Provider selection on this page reflects our independent research coverage, not payment for placement.
Trial results vs. real-world results: why the gap exists
Answer capsule: Real-world GLP-1 outcomes are generally smaller than randomized-trial outcomes for three reasons: lower persistence, incomplete dose titration (often cost-driven), and weaker lifestyle co-intervention. A 2025 JAMA Network Open cohort study of 125,474 US adults found 1-year discontinuation rates of 46.5% in patients with type 2 diabetes and 64.8% without.
1. Persistence
Most trial participants stay on drug because the protocol pays for it, schedules follow-up, and provides counseling. Rodriguez et al. (JAMA Network Open 2025) followed 125,474 US adults newly started on liraglutide, semaglutide, or tirzepatide and found that within one year, 46.5% of patients with T2D and 64.8% without had discontinued. People who stop early do not get the weight loss the trial reports.
2. Dose titration
Trials titrate participants to the maximum tolerated dose. In real life — especially in self-pay telehealth — patients sometimes stay on lower doses to control costs. Lower doses produce less weight loss; SURMOUNT-1's 5 mg arm produced 16% vs 22.5% on 15 mg.
3. Lifestyle co-intervention
The pivotal trials built structured diet and activity counseling into both drug and placebo arms. Real-world prescribing often does not. The drug alone produces less weight loss than the drug plus the trial's lifestyle program.
Questions to ask a provider
- What dose are they planning to titrate me to, and how fast?
- What happens if my insurance changes mid-treatment?
- What is their patient persistence at 6 and 12 months?
- What lifestyle support comes with the prescription?
- How do they handle plateaus and incomplete response?
The 10-point GLP-1 claim score
Answer capsule: A simple 10-point score helps you judge how interpretable any GLP-1 trial claim is, before you decide whether to act on it. Claims that name the trial, the population, the drug and dose, the comparator, the duration, the primary endpoint, both absolute and relative results, the estimand used, the adverse events and discontinuation rate, and the applicability limits score 10. A claim scoring under 5 is essentially a marketing headline.
| Criterion | Points |
|---|---|
| Names the trial and links to the source | 1 |
| States who was studied (population, exclusions) | 1 |
| States drug, dose, and route | 1 |
| States the comparator | 1 |
| States the duration and follow-up | 1 |
| Identifies the primary endpoint | 1 |
| Gives both absolute and relative result numbers | 1.5 |
| Names the estimand or analysis method | 1 |
| Reports adverse events and discontinuation rates | 1 |
| States applicability limits | 0.5 |
| Total | 10 |
How to interpret your score
A worked example: A typical compounded-GLP-1 telehealth ad usually scores 2–3/10. A well-written Foundayo launch summary covering trial name, population, drug dose, comparator, duration, primary endpoint, both estimands and event rates, adverse event summary, and applicability caveats scores ~8.5/10.
Where to find the actual trial papers
Primary peer-reviewed journals
- NEJM (nejm.org) — STEP 1, SURMOUNT-1, SURMOUNT-5, SURPASS-CVOT, SELECT
- The Lancet / Lancet D&E — STEP UP, SURMOUNT-2
- JAMA — STEP 4, SURMOUNT-4
- Nature Medicine — STEP 5, SURMOUNT-3
- Diabetes Care — SUSTAIN program, many sub-analyses
Free access sources
- PubMed Central (PMC) — many GLP-1 publications freely accessible
- ClinicalTrials.gov — full protocols, SAPs, and results summaries
- FDA Drugs@FDA — official labels and approval documents
Source-reliability ladder
- The peer-reviewed paper (NEJM, Lancet, JAMA, etc.)
- The trial's ClinicalTrials.gov record and statistical analysis plan
- The FDA approval documentation and label, where applicable
- Sponsor press releases (clearly labeled)
- Reputable secondary summaries (ACC, Healio, HCPLive)
- General news coverage
When sources disagree, the higher rung wins.
Frequently asked questions about reading GLP-1 trials
What is the single most important thing to check in a GLP-1 trial?
The estimand. Modern GLP-1 trials report at least two — the treatment-policy estimand (closer to real-life adherence) and the trial-product or efficacy-style estimand (a model-based estimate assuming participants took the drug as intended). The headline almost always quotes the higher of the two. Find both before repeating the number.
What does ETD mean in a GLP-1 trial?
ETD stands for estimated treatment difference — the difference between the drug arm and the comparator arm. STEP 1's ETD was −12.4 percentage points: semaglutide produced 12.4 percentage points more weight loss than placebo. Always read the ETD with its 95% confidence interval, which tells you how precise the estimate is.
Why do people on placebo lose weight in GLP-1 trials?
Two reasons. First, the major pivotal GLP-1 obesity trials included structured lifestyle programs — diet, activity, counseling — for both groups. Placebo participants got the same lifestyle support; they just did not get the drug. Second, regression-to-the-mean and trial-participation effects produce small weight losses in placebo arms. Typical placebo weight loss in 68–72 week GLP-1 obesity trials is 2–4%.
Are SURMOUNT-1 and SURMOUNT-5 the same trial?
No. SURMOUNT-1 (Jastreboff et al., NEJM 2022) was the original placebo-controlled Phase 3 trial of tirzepatide in obesity without diabetes — 2,539 adults over 72 weeks. SURMOUNT-5 (Aronne et al., NEJM 2025) was a smaller (n=751), open-label, Phase 3b head-to-head trial of tirzepatide vs semaglutide at maximum tolerated doses, also 72 weeks. Different designs, different conclusions.
Are compounded GLP-1 trial results the same as Wegovy or Zepbound?
No. Compounded GLP-1 products are not FDA-approved finished drug products and have not been evaluated in the trials that supported the approval of Wegovy, Ozempic, Zepbound, or Mounjaro. The FDA has stated that some compounded products sold as 'semaglutide' use salt forms such as semaglutide sodium or semaglutide acetate, which the agency considers different active ingredients. The FDA sent warning letters to 30 telehealth firms on February 20, 2026 for marketing language the agency found misleading.
What does a hazard ratio of 0.80 mean in a GLP-1 trial?
A hazard ratio (HR) of 0.80 means the rate of the event in the treatment group was about 20% lower than in the placebo group over the trial period. It is a relative number. To understand what that meant in real terms, look at the absolute event rates. In SELECT, 6.5% of semaglutide participants experienced MACE vs 8.0% on placebo — a 1.5 percentage-point absolute difference.
Why do people with type 2 diabetes lose less weight on GLP-1s?
Across every major GLP-1 program — STEP vs SUSTAIN, SURMOUNT vs SURPASS, ATTAIN-1 vs ATTAIN-2 — people with T2D consistently lose less weight on the same drug at the same dose than people without diabetes. The exact mechanism is not fully understood. Likely factors include longer disease duration, ongoing use of diabetes medications that can affect weight, metabolic adaptation, and possibly differential GLP-1 receptor signaling.
Is the trial abstract enough to interpret a GLP-1 result?
Not really. The abstract gives you the headline number, but it usually omits estimand details, dropout rates, secondary endpoint statistics, and important subgroup notes. Proper interpretation requires reviewing design, prespecified outcomes, statistical assumptions, conduct, uncertainty, clinical importance, and generalizability — none of which fit in an abstract.
Should I use the efficacy or treatment-policy estimand for my own expectations?
For most people setting personal expectations, the treatment-policy estimand is closer to real life because it reflects actual adherence and dropout patterns. But your individual outcome depends on dose, persistence, baseline weight, T2D status, and lifestyle co-intervention. No single trial average is the right expectation for any one person.
How current does a GLP-1 trial need to be?
For established drugs — semaglutide, tirzepatide, liraglutide — the foundational trials from 2018–2024 remain the reference. For newer drugs such as orforglipron (Foundayo, approved April 2026), retatrutide (still investigational), and CagriSema (still investigational), the evidence base is still consolidating, with full peer-reviewed publications continuing through 2026 and 2027.
What is the difference between STEP and SUSTAIN?
STEP is the semaglutide obesity program — no diabetes required. SUSTAIN is the semaglutide diabetes program. The drug is the same active ingredient, but the trials are designed around different primary endpoints (weight change in STEP, A1c reduction in SUSTAIN) and different patient populations.
Did the FDA recently change the suicidality warning on GLP-1s?
Yes. In January 2026, the FDA requested removal of the suicidal ideation and behavior warning from the labels of Saxenda (liraglutide), Wegovy (semaglutide 2.4 mg), and Zepbound (tirzepatide). The decision followed an FDA meta-analysis of 91 placebo-controlled trials with 107,910 participants, which found no increased risk. The FDA still recommends patients discuss any new or worsening depression, suicidal thoughts, or mood and behavior changes with a healthcare professional.
What we actually verified, and how
Last verified: May 18, 2026. Next scheduled re-verification: August 2026, or sooner if any trial above publishes updates, corrections, label changes, or new analyses.
Each row of the Trial Truth Table links to the primary source used for that row: the peer-reviewed publication where one is available (NEJM, Lancet, JAMA, Nature Medicine), the FDA approval documentation and label where appropriate (Foundayo / orforglipron), and reputable trial summaries (ACC) where used. ATTAIN-1, ATTAIN-2, and STEP UP are areas where peer-reviewed publication is still expanding through 2026; those rows will be re-verified as each full paper publishes.
What we did not do
- · We did not independently reanalyze any patient-level trial data
- · We did not claim medical review — no licensed clinician has signed off on this specific page
- · We did not make individualized medical recommendations
- · We did not treat Reddit, forums, or social media anecdotes as clinical evidence
- · We did not cite any compounded-GLP-1 company's internal data as if it were trial-equivalent
Primary sources used on this page
- STEP 1 — Wilding JPH et al. NEJM 2021;384:989–1002. Link
- STEP 4 — Rubino D et al. JAMA 2021;325(14):1414–1425. PubMed
- STEP 5 — Garvey WT et al. Nature Medicine 2022;28:2083–2091. Link
- STEP UP — Wadden TA et al. PubMed
- SURMOUNT-1 — Jastreboff AM et al. NEJM 2022;387:205–216. Link
- SURMOUNT-2 — Garvey WT et al. The Lancet 2023;402:613–626. ACC summary
- SURMOUNT-3 — Wadden TA et al. Nature Medicine 2023;29:2909–2918. Link
- SURMOUNT-4 — Aronne LJ et al. JAMA 2024. PMC
- SURMOUNT-5 — Aronne LJ et al. NEJM 2025. PubMed
- SURPASS-CVOT — Nicholls SJ et al. NEJM 2025. ACC summary
- SELECT — Lincoff AM et al. NEJM 2023;389:2221–2232. PubMed
- ATTAIN-1 and ATTAIN-2 — FDA approval, April 1, 2026. FDA press announcement; Foundayo label
- ICH E9(R1) addendum on estimands. PDF
- Rodriguez PJ et al. JAMA Network Open 2025;8(1):e2457349. Link
- FDA warning letters to telehealth companies, February 20, 2026. FDA
- FDA Drug Safety Communication — suicidality warning removal, January 2026. FDA
About this page
Researched and assembled by The RX Index Editorial Team. This is an evidence-literacy guide based on primary trial publications, FDA documentation, and peer-reviewed secondary sources. It is not medical advice and does not recommend any drug, dose, or provider. Always work with a licensed clinician for personal decisions.
Editorial position: The RX Index may earn affiliate commissions from some GLP-1 telehealth providers in our directory. We do not accept payment to alter editorial content or rankings. If you spot an error or have a verification request, contact the editorial team.
Affiliate disclosure: The RX Index earns a commission when you sign up with some of the providers mentioned on this page. It does not affect what you pay, and it never determines our rankings or which providers we cover. Read the full disclosure.
Published: