Evidence review
What Happens When a Peptide Trial Fails
Four negative results, and what each one killed. A trial can fail an indication, a molecule, a surrogate, or nothing at all — and the difference matters.
Failure is a specific event
A trial fails when it does not show what it was designed to show. That is narrower than it sounds, and confusing it with "the drug does not work" causes most of the misreading in this area.
Four negative results follow. Each one killed something different.
A working drug, in a disease it does not treat
The evoke and evoke+ trials tested oral semaglutide up to 14 mg once daily in people aged 55 to 85 with amyloid-confirmed Alzheimer's disease and either mild cognitive impairment or mild dementia. They ran across 566 sites in 40 countries, screened 9,981 people and randomised 3,8081.
The primary endpoint was change in the Clinical Dementia Rating–Sum of Boxes score at week 104. The result: an estimated difference of −0.08 (95% CI −0.35 to 0.20, p = 0.57) in evoke, and 0.10 (95% CI −0.17 to 0.38, p = 0.46) in evoke+. Both trials were discontinued due to negative clinical outcome1.
Now note what did not happen. Semaglutide remains an approved medicine with substantial randomised evidence in weight and cardiovascular outcomes — the semaglutide page carries that record. What died was a hypothesis about the brain, tested at a scale that could settle it.
This is the cleanest available demonstration that "the trial failed" and "the drug does not work" are different statements. It took 3,808 randomised participants to distinguish them.
The surrogate moved and nothing else did — twice
MK-677 raised serum IGF-1 by 60.1% at six weeks and 72.9% at twelve months in a 12-month trial of 563 patients with mild to moderate Alzheimer's disease. Across the CIBIC-plus, ADAS-Cog, ADCS-ADL and CDR-sob, the authors reported no significant differences between treatment groups. Their conclusion states that despite evidence of target engagement, the compound was ineffective at slowing progression2.
In a separate six-month trial of 161 hip-fracture patients aged 65 and over, MK-0677 raised IGF-1 by 84% against 17% on placebo, and produced no significant differences in functional performance measures or in the overall Sickness Impact Profile score. Three of four lower-extremity measures favoured the drug numerically without reaching significance3.
What failed here is a mechanism claim: that raising IGF-1 produces the clinical benefit people wanted from it. Two trials, two indications, two negatives, twenty years ago. The compound is still discussed as though those trials had not happened. Our page on growth-hormone secretagogues covers the class.
The indication died and the molecule did not
Bimagrumab is a monoclonal antibody against type II activin receptors. Its phase 2b trial in sporadic inclusion body myositis, RESILIENT, enrolled 251 participants across three dose arms and placebo, with infusions every four weeks for at least 48 weeks4.
The primary outcome was six-minute walking distance at week 52. It did not differ from placebo at any dose: 17.6 m for 10 mg/kg (99% CI −19.6 to 54.8, p = 0.22), 18.6 m for 3 mg/kg (p = 0.19), and −1.3 m for 1 mg/kg (p = 0.93)4.
That killed the myositis programme. It did not kill the molecule.
The same antibody was subsequently tested for body composition in adults with type 2 diabetes and obesity5, and then in a 507-participant phase 2 trial in obesity alongside semaglutide, where least-squares mean weight change at week 48 was −9.3 kg with bimagrumab 30 mg/kg, −14.2 kg with semaglutide 2.4 mg and −17.8 kg with the high-dose combination, against −3.3 kg on placebo6.
A failure in one indication is not a verdict on a mechanism. It is a verdict on a mechanism *in that population, on that endpoint, at that dose, over that duration*.
"Terminated" is a status, not a result
A registry record marked terminated tells you a trial stopped early. It does not tell you why, and in most cases it does not tell you anything about efficacy at all.
Trials stop for enrollment problems, funding, business decisions, sponsor reorganisation, and — sometimes — safety or futility. A search on the term "CJC-1295" in August 2026 returned exactly one registered study, NCT00267527, a phase 2 record in HIV patients with visceral obesity, marked terminated.
Reading that as a negative efficacy result would be an error. Reading it as evidence of efficacy would be a bigger one. What it establishes is that the one registered trial did not finish.
Coming in lower is not failing
One more distinction, because it gets collapsed constantly.
Survodutide's phase 2 reported −14.9% at its highest dose at week 46, with its primary analysis based on the dose assigned at randomisation and COVID-19-related discontinuations censored7. Its phase 3, SYNCHRONIZE-1, reported −12.2% and −13.0% at week 76 under a treatment-regimen estimand that incorporates early discontinuation and prohibited-medication use8.
That trial met its primary endpoints. It did not fail. It produced a smaller number under a stricter analysis of a different population over a longer period — which is what phase 3 is for. We separate those in triple versus dual agonists.
Why the negatives are the most useful pages in this field
Positive results get press releases, conference slides and second lives on social media. Negative results get a journal article and silence.
That asymmetry means the picture most people carry is systematically more optimistic than the literature. Reading the negatives is the cheapest available correction.
It is also worth noticing which negatives get published at all. Both evoke trials were reported in full in a major journal, with a linked commentary and a plain-language summary1. By contrast, in the thymosin β4 set, a phase 3 record with a listed enrollment of 700 was marked completed and carried no posted results when retrieved in August 2026 — covered in healing peptides.
How to check whether a trial failed
Find the NCT number, then look at three things.
The registry record's status and whether results are posted. Completed with no results posted is its own kind of answer.
The published paper's primary endpoint — not its abstract's most quotable sentence. A trial that missed its primary and hit a secondary has failed, whatever the secondary says.
And the p-value on the primary, next to the confidence interval. In evoke, −0.08 with an interval from −0.35 to 0.20 is a result that excludes any meaningful benefit, which is more informative than "no significant difference" makes it sound.
Our research index keeps the negatives alongside the positives, and why most peptide trials are small and short covers why so many compounds never generate either.
Frequently asked questions
Did semaglutide fail?
It failed in Alzheimer's disease. The evoke and evoke+ phase 3 trials randomised 3,808 people and reported an estimated difference on the primary endpoint of −0.08 (p = 0.57) and 0.10 (p = 0.46); both were discontinued due to negative clinical outcome. That result concerns one disease. Semaglutide remains an approved medicine with substantial randomised evidence in weight and cardiovascular outcomes.
Does a failed trial mean the mechanism is wrong?
Not necessarily. Bimagrumab missed its primary endpoint in a 251-participant trial in sporadic inclusion body myositis — six-minute walking distance differences of 17.6 m (p = 0.22), 18.6 m (p = 0.19) and −1.3 m (p = 0.93). The same antibody later produced substantial weight and body-composition effects in obesity trials. A failure is specific to a population, endpoint, dose and duration.
What does a terminated trial tell me?
That it stopped early, and usually nothing more. Trials terminate for enrollment problems, funding, business decisions and sponsor changes as well as for safety or futility. The one registered CJC-1295 study returned by a search in August 2026 is marked terminated, which establishes that it did not finish — not that the compound failed or succeeded.
Is a phase 3 number lower than the phase 2 a failure?
No. Survodutide's phase 2 reported −14.9% at week 46 under an analysis that censored COVID-19-related discontinuations; its phase 3 reported −12.2% and −13.0% at week 76 under a treatment-regimen estimand that absorbs early discontinuation. The phase 3 met its primary endpoints. A smaller number under a stricter analysis in a different population over a longer period is what phase 3 is designed to produce.
Why are negative results so hard to find?
They are published without amplification, and some are not published at all. In the thymosin β4 set, a phase 3 record with a listed enrollment of 700 was marked completed and carried no posted results when retrieved in August 2026. Checking the registry record for a posted result is the fastest way to spot the gap.
References
- Cummings JL, Atri A, Sano M, et al. (2026). Efficacy and safety of oral semaglutide 14 mg (flexible dose) in early-stage symptomatic Alzheimer's disease (evoke and evoke+): two phase 3, randomised, placebo-controlled trials. The Lancet. https://pubmed.ncbi.nlm.nih.gov/41865758/
- Sevigny JJ, Ryan JM, van Dyck CH, et al. (2008). Growth hormone secretagogue MK-677: no clinical effect on AD progression in a randomized trial. Neurology. https://pubmed.ncbi.nlm.nih.gov/19015485/
- Bach MA, Rockwood K, Zetterberg C, et al. (2004). The effects of MK-0677, an oral growth hormone secretagogue, in patients with hip fracture. Journal of the American Geriatrics Society. https://pubmed.ncbi.nlm.nih.gov/15066065/
- Hanna MG, Badrising UA, Benveniste O, et al. (2019). Safety and efficacy of intravenous bimagrumab in inclusion body myositis (RESILIENT): a randomised, double-blind, placebo-controlled phase 2b trial. The Lancet Neurology. https://pubmed.ncbi.nlm.nih.gov/31397289/
- Heymsfield SB, Coleman LA, Miller R, et al. (2021). Effect of Bimagrumab vs Placebo on Body Fat Mass Among Adults With Type 2 Diabetes and Obesity: A Phase 2 Randomized Clinical Trial. JAMA Network Open. https://pubmed.ncbi.nlm.nih.gov/33439265/
- Heymsfield SB, Aronne LJ, Montgomery P, et al. (2026). Bimagrumab plus semaglutide alone or in combination for the treatment of obesity: a randomized phase 2 trial. Nature Medicine. https://pubmed.ncbi.nlm.nih.gov/41772149/
- le Roux CW, Steen O, Lucas KJ, et al. (2024). Glucagon and GLP-1 receptor dual agonist survodutide for obesity: a randomised, double-blind, placebo-controlled, dose-finding phase 2 trial. The Lancet Diabetes & Endocrinology. https://pubmed.ncbi.nlm.nih.gov/38330987/
- le Roux CW, Wharton S, Startseva E, et al. (2026). Survodutide Once Weekly for the Treatment of Adults with Obesity. New England Journal of Medicine. https://pubmed.ncbi.nlm.nih.gov/42253238/
Medical disclaimer: This content is for general educational purposes only and is not medical advice, diagnosis, or treatment. Always consult a licensed healthcare professional before starting, stopping, or changing any treatment.
Continue reading
Amylin Analogs: The Next Class After Incretins
An amylin analog has been FDA-approved since 2005. What changed is not the hormone — it is how long the molecule lasts. The trials, in order.
ReadCertificates of Analysis: What They Prove
A COA is a claim about a lot, not a guarantee about a vial. What each test answers, what a purity figure cannot say, and the gap the document never closes.
ReadCompounded Is Not the Trial Drug
A trial tests one defined article: a molecule, a form, a concentration, a container. Compounding changes some of those, and the evidence travels only that far.
Read503A vs 503B: What Is Actually in the Vial
One scheme is inspected against manufacturing standards; the other is not. What each may legally start from, and what the inspection record shows in practice.
ReadGrowth-Hormone Secretagogues: What the Evidence Actually Shows
These compounds reliably raise IGF-1. That is the surrogate. When trials measured function, fracture recovery or disease progression, the results changed.
ReadHealing Peptides and the Gap Between Anecdote and Trial
BPC-157 and TB-500 have thirty years of animal work and a very thin human record. Here is exactly what is registered, what is published, and what neither shows.
ReadHow to Read a ClinicalTrials.gov Record
A registration is a filing, not a review. Fifteen fields, two real peptide records, and the specific places where a record quietly tells you it has stalled.
ReadIncretins Explained: GLP-1, GIP and Glucagon
Three hormones, three receptors, and a class of drugs built on them. What each one does, and why the receptor diagram predicts less than it looks like.
ReadMuscle During Rapid Weight Loss: What Has Been Measured
Roughly a quarter of the weight lost is lean mass — and that was true on placebo too. What the DXA substudies actually show, and how small they are.
ReadOpen-Label vs Blinded: What You Can Conclude
Masking is not a quality score — it decides which explanations a trial can rule out. Five real registry records, and the honest evidence on blinding itself.
ReadOral vs Injectable Peptides: What Changes
One molecule, two routes, one label — and roughly 73 times the milligrams per week. What the gut does to a peptide, and what it costs to get past it.
ReadHalf-Life and Dosing Frequency
From 1.5 minutes to a week, on labels for the same hormone family. Half-life is the number that decides whether a peptide is dosed before meals or once a month.
ReadReconstitution: The Arithmetic, Not the Advice
Concentration is mass divided by volume — except the naive division is wrong, and an FDA label's own numbers prove it. The maths, and what it cannot tell you.
ReadWhat a Phase 2 Result Does and Does Not Tell You
Phase 2 is a dose-finding experiment, not a verdict. Two peptide programmes where the phase 3 exists show exactly which parts of the number survive.
ReadRegistered but Never Reported: The Trials That Vanish
Completed trials that post no results are common and measurable. How to check a compound's registry file, and what the silence does and doesn't tell you.
ReadWhy n=30 Is Not the Same as n=3,000
Enrolment size is not a quality badge — it is precision, and it is the ability to see rare harm. Worked on real peptide trials, with the arithmetic shown.
ReadSurrogate Endpoints vs Outcomes That Matter
Most peptide claims rest on a marker moving, not on anything happening to a person. Two paired trials show what the difference costs to establish.
ReadWhy a Triple Agonist Is Not Simply Better Than a Dual
Adding a receptor adds a mechanism, a side-effect profile, and a reason a trial might read differently. The numbers do not rank the way the labels suggest.
ReadWhat a Peptide Is, and What the Word Hides
Three amino acids or sixty-three thousand daltons — both get called peptides. The definitional confusions that cause the most misreading, settled by labels.
ReadWhat “Research Chemical” Actually Means
The phrase is a sales category, not a regulatory one. What FDA has actually published about seventeen of these peptides, and what the label is doing instead.
ReadHow to Tell a Peptide Claim Is Outrunning Its Evidence
Twelve tells, each with a real case from the registry or the literature. Overstated peptide claims tend to fail on one of them, and usually the same one.
ReadWho Funded the Trial, and Why It Matters
Sponsorship shifts conclusions more than it shifts methods. Where to find the funding line, what the Cochrane evidence measured, and what it did not.
ReadWhy Most Peptide Trials Are Small and Short
A registry census, August 2026: one compound has 759 registered studies, another has zero. The reasons are economic, not about whether the drug works.
Read