Failed peptide trials archive: when primary endpoints don't make it

The peptide field has a survivor-bias problem. The trials that succeed are loud — they get press releases, journal commentary, podcast circuits, and clinical-guideline footnotes. The trials that fail are quiet. The sponsor issues a brief statement, the stock falls, the principal investigator publishes a sober results paper a year later in a mid-tier journal, and the molecule disappears from clinical conversation while persisting unchanged in compounded stacks, gray-market price lists, and influencer protocols that were built on the earlier signal. The result is a public conversation about peptides in which the negative half of the evidence base is structurally underweighted.

This archive exists to reweight it. Each entry catalogs a peptide trial that did not meet its prespecified primary endpoint, identifies the sponsor and registry, names the design and sample size, reports the result without softening, describes what happened to the development program afterward, and extracts what the field learned. The framing is non-editorial: the page describes negative results because they are evidence, not because they reflect on the integrity of the molecules or the investigators. Some failures validate the mechanism while disqualifying the indication. Some failures qualify the mechanism itself. A meaningful subset reframe what "validation" means in the first place — the trial design's choice of primary endpoint encoded a hypothesis about what would matter clinically, and the hypothesis turned out to be wrong even though the molecule was working on something the secondary endpoints could measure.

The conventional reading of a negative trial in the regulatory framework treats prespecified primaries as load-bearing and secondaries as hypothesis-generating. This page applies that frame consistently. The archive is not a graveyard — it is the reference layer that the rest of the corpus implicitly assumes.

AOD-9604 — METAOD006, the failed obesity Phase IIb (Metabolic Pharmaceuticals, 2007)

Sponsor: Metabolic Pharmaceuticals Ltd. (Melbourne). Trial: METAOD006, ACTRN12605000067673. Indication: Obesity. Design: 24-week randomized double-blind placebo-controlled multicentre dose-ranging trial of oral AOD-9604 in six arms (placebo, 1 mg, 5 mg, 10 mg, 20 mg, 30 mg). N: 536 obese adults. Primary endpoint: weight loss at 12 weeks versus placebo. Result: no dose arm separated from placebo on the primary endpoint. The development program for obesity ended in 2007.

AOD-9604 is the cleanest case study in the corpus of a peptide whose preclinical mechanism story progressed through Phase IIa, encouraged a larger and longer Phase IIb, and then did not reproduce in the larger trial. The preceding 12-week 300-patient Phase IIa (METAOD005, Herd et al. 2005) had produced ~2.8 kg weight loss on 1 mg oral dosing versus ~0.8 kg on placebo — a separation that justified the IIb investment. The longer 24-week trial diluted that signal to non-significance against placebo across all five dose arms. The integrated safety summary was published a decade later as part of the program retrospective (Stier et al., J Endocrinol Metab 2013, 3(1-2):7-15), but no efficacy paper from the failed METAOD006 was published in a high-impact obesity journal. The public record consists of the company's ASX announcements, the safety publication, and the trial registry record.

What happened next is the part that makes this entry load-bearing for the modern reader. The originator company rebranded toward a cartilage / osteoarthritis preclinical track and out-licensed the asset; the obesity indication did not progress to Phase III in any jurisdiction. In 2014 the company issued a Generally Recognized As Safe self-determination for AOD-9604 as a food and supplement ingredient, citing the integrated safety dataset. GRAS is a safety determination, not an efficacy endorsement; the FDA's public GRAS notice inventory does not list a corresponding agency "no questions" letter for AOD-9604 as of 2026. Twelve years after the failed pivotal, AOD-9604 has become one of the most commonly stacked peptides in compounded weight-loss protocols and gray-market research-channel use — almost entirely on the marketing strength of the rodent lipolysis case and the Phase IIa signal, not the Phase IIb outcome that ended pharmaceutical development. The AOD-9604 peptide page walks the molecule's full story; the critic response on AOD-9604 fat-loss claims walks the marketing-versus-evidence reconstruction.

What the field learned: a Phase IIa signal in 12 weeks is not the same as a Phase IIb signal in 24 weeks, and the modal outcome when a small early-phase effect meets a longer endpoint window with broader recruitment is dilution. The mechanism story may be true — AOD-9604 plausibly increases adipose lipolysis via β3-adrenergic upregulation in rodents — and the clinical effect can still be too small to detect against the noise of free-living human weight regulation. A mechanism with no efficacy is just a mechanism.

Elamipretide — MMPOWER-3, the failed primary mitochondrial myopathy pivotal (Stealth BioTherapeutics, 2019/2020)

Sponsor: Stealth BioTherapeutics. Trial: MMPOWER-3, NCT03323749. Indication: Primary mitochondrial myopathy. Design: Phase 3 randomized double-blind placebo-controlled trial across 27 centers in seven countries. N: ~218 randomized. Primary endpoints: change from baseline in six-minute walk test distance and change in the Primary Mitochondrial Myopathy Symptom Assessment fatigue subscale at 24 weeks. Result: neither primary endpoint separated from placebo. The lead-indication development program for SS-31 in primary mitochondrial myopathy ended.

The MMPOWER-3 result is the most consequential single failed trial of any peptide on this site, because it ended what was supposed to be the molecule's pivotal indication. The earlier MMPOWER-2 dose-ranging Phase 2 had produced acute 6MWT signals that motivated the Phase 3 design (Karaa et al. 2018). MMPOWER-3 was the confirmatory trial, designed to support an FDA filing in primary mitochondrial myopathy. The Class I-rated Neurology paper (Karaa et al. 2023, doi:10.1212/WNL.0000000000207402) reports that both prespecified primaries — 6MWT distance and fatigue subscale — were statistically equivalent between elamipretide and placebo arms at 24 weeks.

What happened next is one of the more interesting program pivots in the modern peptide literature. A prespecified subgroup analysis of nuclear-DNA mitochondrial disease patients showed apparent response signals, which informed the subsequent NuPOWER trial design. In parallel, Stealth had been running a smaller program in Barth syndrome — a rare X-linked cardiolipin disorder where the mechanism specificity was unusually high and the indication tightly defined (Thompson 2021). The Barth program produced the data that supported FDA approval of Forzinity (elamipretide) for Barth syndrome in September 2025. The pivotal that ended the lead indication thus produced a path-of-development pivot toward the narrower indication that actually reached the market. The molecule did not die; the original indication did.

What the field learned: cardiolipin-targeting peptides can produce acute physiology signals (cardiac volume changes in Daubert et al. 2017, structural retinal signals in ReCLAIM-2 below) that do not reliably translate to prespecified functional endpoints in larger blinded trials of heterogeneous mitochondrial-disease populations. The exception is high-mechanism-specificity rare-disease indications where the molecular target maps tightly to the clinical phenotype. The 24-week treatment duration may also be short for a chronic-mitochondrial-disease intervention; whether the same trial run for 48 or 72 weeks would have produced different results is unaddressed.

Elamipretide — ReCLAIM-2, the failed dry-AMD primary endpoint with positive structural secondaries (Stealth, 2024)

Sponsor: Stealth BioTherapeutics. Trial: ReCLAIM-2, NCT03891875. Indication: Dry age-related macular degeneration with geographic atrophy. Design: Phase 2 randomized double-blind placebo-controlled trial. N: 176 patients aged ≥55 with at least one eye affected. Primary endpoints: mean change from baseline in low-luminance best-corrected visual acuity (LL BCVA) and change in square-root-converted geographic atrophy area on OCT, both at 48 weeks. Result: neither primary endpoint was met (LL BCVA −3.0 letters on elamipretide vs −4.4 letters on placebo, p = 0.49; GA area progression 0.312 mm vs 0.275 mm, p = 0.34). Secondary structural endpoints around ellipsoid zone integrity separated more cleanly: a 43% reduction in mean progression of macular percentage of total EZ attenuation (3.69% vs 6.47%, p = 0.0034) and 47% reduction in partial EZ attenuation (4.10% vs 7.68%, p = 0.0040) (Ehlers et al. 2024, doi:10.1016/j.xops.2024.100628).

ReCLAIM-2 is the cleanest published case in the peptide literature of a trial that missed its prespecified clinical-functional primaries but produced statistically significant secondary signals on a structural biomarker that the sponsor has cited as proof-of-mechanism for future development. The conventional regulatory framing treats prespecified primaries as load-bearing, which means this is a primary-endpoint miss. The Stealth framing treats the EZ-attenuation secondaries as mechanism validation that justifies a redesigned Phase 3 with EZ integrity moved into the primary position. Both readings are coherent given different priorities; only the first satisfies the FDA-grade reading the regulatory framework applies.

What happened next: Stealth has framed the trial as a mechanism-validation result and has signaled further AMD development. The molecule's parallel rare-disease path proceeded to the Barth syndrome approval as described above. The dry-AMD indication remains an open development question rather than a closed one.

What the field learned: secondary structural biomarkers can produce clean signals in trials that miss clinical-functional primaries, and the question of whether structural-biomarker change predicts downstream clinical benefit is the unresolved one. Reading the SS-31 cardiology, mitochondrial-myopathy, and AMD programs together is the most informative single comparison in the cardiolipin-stabilization literature — the molecule has produced acute-physiology or biomarker-level signals across three distinct indications that did not consistently translate to prespecified clinical-functional primary endpoints in larger blinded trials. The exception was Barth syndrome, where the indication was tightly defined and the mechanism specificity unusually high. See the SS-31 peptide page for the full pharmacology.

Tirzepatide — SURMOUNT-4, the maintenance-withdrawal trial that reframed chronic use (Eli Lilly, 2024)

Sponsor: Eli Lilly. Trial: SURMOUNT-4, NCT04660643. Indication: Obesity (weight-loss maintenance). Design: 36-week open-label tirzepatide lead-in followed by 52-week randomized double-blind placebo-controlled withdrawal. N: 670 randomized after lead-in (from 783 enrolled). Primary endpoint: mean percent weight change from randomization (week 36) to week 88. Result: continued tirzepatide produced −5.5% additional weight change versus +14.0% weight regain on placebo — a treatment difference of −19.4 percentage points (p < 0.001). This was a positive trial on its prespecified primary, not a failure (Aronne et al. 2024, doi:10.1001/jama.2023.24945).

SURMOUNT-4 belongs in this archive not because it failed but because its result reframed long-term use of incretin pharmacotherapy in a way that some readers register as a kind of mechanism failure. The 14-percentage-point weight regain over 52 weeks of placebo withdrawal — after 20.9% mean weight loss during the 36-week lead-in — established that GLP-1/GIP-mediated weight loss does not persist after stopping the drug. The clinical conclusion that flowed from this result is that tirzepatide is chronic pharmacotherapy, not a weight-loss-then-stop intervention. For practitioners and patients who understood incretin treatment as a "course" with a definable endpoint, this trial substituted a different and less convenient mental model.

The trial design is a randomized-withdrawal pattern that positively selects for both tolerability and efficacy — only participants who tolerated and responded to 36 weeks of tirzepatide were eligible for randomization. Real-world discontinuation populations include patients who stopped earlier for adverse events, cost, or inadequate response; the 14% regain figure should be read as the maintenance-population estimate, not a universal post-stop trajectory. A 2025 post hoc analysis extended the result to cardiometabolic parameters and reported that the lead-in cardiometabolic improvements regressed in proportion to weight regain in the placebo arm — clinically the more important framing for the chronic-care position.

What the field learned: the modal post-stop trajectory on GLP-1 class agonists is regain, not stabilization, and the magnitude of the regain captures the steepest portion of the curve within the first year. The trial does not establish what happens beyond 52 weeks — whether regain continues toward full pre-treatment weight, partially stabilizes, or plateaus at a new set point remains unaddressed by SURMOUNT-4 and is the most important open question in the GLP-1 discontinuation literature. The GLP-1 discontinuation playbook walks the operational consequences. See also the Tirzepatide peptide page.

Liraglutide — ELAD, the failed Alzheimer's Phase IIb cerebral glucose primary (Imperial College, 2025/2026)

Sponsor: Imperial College London (academic IIT, drug supplied by Novo Nordisk). Trial: ELAD (Evaluating Liraglutide in Alzheimer's Disease), NCT01843075. Indication: Mild-to-moderate Alzheimer's disease in non-diabetic adults. Design: Multicentre randomized double-blind placebo-controlled Phase 2b across 24 UK clinics. N: 204. Primary endpoint: change in cerebral glucose metabolic rate measured by 18F-FDG PET imaging at 52 weeks. Result: no significant difference (between-group difference −0.17; 95% CI −0.39 to 0.06; p = 0.14). Secondary endpoints showed reduced brain atrophy on MRI and 18% slower cognitive decline on ADAS-Exec, both reaching statistical significance (Edison et al., Nat Med 2026, 32:353–361).

ELAD is the modern-era reference trial for the GLP-1 / cognitive decline hypothesis applied to Alzheimer's disease, and it followed the same primary-missed / secondaries-positive pattern that ReCLAIM-2 produced in ophthalmology. The trial was led by Paul Edison at Imperial College, with Novo Nordisk supplying drug rather than running the trial — an investigator-initiated design that the result-driven reader will weigh differently from sponsor-led pivotals. The prespecified primary was a metabolic-imaging measure (18F-FDG PET) chosen on the hypothesis that GLP-1 receptor agonism in the CNS would modulate cerebral glucose handling in a measurable direction; that prediction was not borne out at the trial's sample size and dose.

The secondary endpoint story is the part that has driven the discourse since publication. Brain atrophy on volumetric MRI was reduced by nearly 50% in the liraglutide arm versus placebo, and cognitive decline on the ADAS-Exec executive-function subscale slowed by 18%. Neither was the trial's primary; both reached conventional statistical significance. The split between mechanism-implied imaging primary and structural-and-functional secondaries is exactly the kind of result that does not resolve cleanly under regulatory framing but that the field reads as suggestive.

What happened next: the program for liraglutide in Alzheimer's effectively ended at this trial, because the indication moved on to semaglutide, where the larger Phase 3 EVOKE / EVOKE+ pair completed in 2025 (see below). Liraglutide remains FDA-approved for type 2 diabetes (as Victoza) and obesity (as Saxenda); the Alzheimer's indication did not advance to Phase III for this molecule. See the Liraglutide peptide page for the full indication landscape.

What the field learned: cerebral glucose-metabolism imaging as a primary endpoint for a GLP-1 cognitive trial may be too far upstream of the clinical outcome the indication actually targets. The trial is also a useful case study in how a sponsor-supplied / academic-led design produces a different interpretive context from a sponsor-pivotal — the result was published with no pressure to spin the secondaries, and the discussion explicitly framed the trial as primary-negative. Whether the secondary signals on atrophy and ADAS-Exec would have replicated in a larger confirmatory design is now a question the molecule itself will not answer, because development has moved to semaglutide.

Semaglutide — EVOKE / EVOKE+, the failed early-Alzheimer's Phase 3 pair (Novo Nordisk, 2025)

Sponsor: Novo Nordisk. Trials: EVOKE (NCT04777396) and EVOKE+ (NCT04777409). Indication: Early-stage symptomatic Alzheimer's disease. Design: Two parallel Phase 3 randomized double-blind placebo-controlled trials of once-daily oral semaglutide 14 mg versus placebo, 104-week main treatment plus 52-week extension. N: approximately 1855 in EVOKE and 1953 in EVOKE+. Primary endpoint: change in Clinical Dementia Rating-Sum of Boxes (CDR-SB) from baseline. Result: neither trial demonstrated superiority of semaglutide over placebo on the primary endpoint, despite some improvements in disease-associated biomarkers. Novo Nordisk announced discontinuation of the one-year extension periods on November 24, 2025.

EVOKE / EVOKE+ replaced the failed ELAD liraglutide hypothesis with a larger, longer, dose-optimized semaglutide test of the same underlying mechanism question — does GLP-1 receptor agonism slow cognitive decline in established Alzheimer's disease? — and answered it negatively on the prespecified clinical primary. The biomarker-positive / clinical-negative split is again the pattern: the molecule appeared to move some disease-associated biomarkers in the direction the mechanism predicted, but those biomarker changes did not translate into measurable slowing of cognitive decline on the regulatory-grade endpoint at the dose and duration tested.

What happened next: the GLP-1-receptor-agonist Alzheimer's indication is essentially closed at the molecule-monotherapy level following these results. Novo Nordisk has signaled potential interest in combination therapy approaches; the next-generation hypotheses center on combining incretins with anti-amyloid antibodies (lecanemab, donanemab) on the theory that the metabolic-and-vascular contribution to AD progression is real but additive to the proteinopathy rather than substitutive. None of the combination hypotheses are confirmed as of 2026.

What the field learned: the GLP-1 / cognitive decline hypothesis at the dose and duration tested in symptomatic early Alzheimer's does not produce clinically meaningful slowing. Whether that conclusion generalizes to preclinical / preventive populations (the alz.org-aligned framing) or to higher-dose / longer-duration designs is not addressed. The pattern of a positive observational signal (the alz.org GLP-1-and-dementia association in diabetes cohorts) producing a negative confirmatory trial in a related-but-different population is a familiar one in modern Alzheimer's drug development. See the Semaglutide peptide page.

Anamorelin — ROMANA-1 and ROMANA-2, the lean-mass-yes / hand-grip-no co-primary split (Helsinn, 2015/2016)

Sponsor: Helsinn Healthcare. Trials: ROMANA-1 (NCT01387269) and ROMANA-2 (NCT01387282). Indication: Cachexia in advanced non-small-cell lung cancer. Design: Two identical Phase 3 randomized double-blind placebo-controlled trials at 93 sites across 19 countries. N: 484 (ROMANA-1) + 495 (ROMANA-2) = 979 total. Co-primary endpoints: median change in lean body mass and handgrip strength over 12 weeks. Result: lean body mass increased significantly (ROMANA-1: +0.99 vs −0.47 kg, p < 0.0001; ROMANA-2: +0.65 vs −0.98 kg, p < 0.0001). Handgrip strength did not differentiate (ROMANA-1: p = 0.15; ROMANA-2: p = 0.65) (Temel et al., Lancet Oncol 2016, 17:519–531).

ROMANA-1 and ROMANA-2 are the cleanest published example in the peptide literature of a trial where one co-primary endpoint was unambiguously met and the other was unambiguously missed. The lean-mass effect was statistically and clinically real — anamorelin's ghrelin-receptor agonism is a credible muscle-anabolic mechanism, and the trials demonstrated it works at scale in cancer cachexia. The hand-grip strength endpoint did not move. Because both endpoints were prespecified as co-primaries, the trials are conventionally read as having missed.

The mass-versus-function divergence is the load-bearing lesson. Cancer cachexia destroys muscle for catabolic-cytokine reasons that pharmacologic muscle anabolism can partially counteract on the mass axis but does not necessarily restore on the strength axis within a 12-week window. The strength measure may also be intrinsically less sensitive than mass measures over short trial durations — muscle protein turnover responds faster than neuromuscular function recovery, particularly in patients whose underlying disease burden continues to advance.

What happened next is geographically split. The FDA did not approve anamorelin for cancer cachexia in the United States; the EMA did not approve it in Europe. Japan's PMDA approved anamorelin (marketed as Adlumiz, ONO-7643) on December 11, 2020, for cachexia in non-small-cell lung cancer, gastric cancer, pancreatic cancer, and colorectal cancer, based on smaller domestic Phase 2 and Phase 3 trials that focused on lean body mass and body-weight endpoints rather than the hand-grip co-primary that ROMANA had used. The same molecule is on the market under one regulator's framework and not on the market under two others, with the difference depending on whether the regulator weighted the mass endpoint or the function endpoint as decisional (Wakabayashi et al., J Cachexia Sarcopenia Muscle 2021, 12:14–21).

What the field learned: peptide trial design must select endpoints that are both responsive to the drug's mechanism and matter clinically, and "both" is the hard part. Mass measures are responsive but may not predict function. Function measures matter clinically but may not respond within typical trial durations. The mass-versus-function divergence in ROMANA reframed the cachexia endpoint discussion across the field — subsequent trials in this space have tended to use multidomain composite endpoints or to power separately for mass and function rather than treat them as co-primaries that both must be met. The sarcopenia and peptides dossier walks the parallel question in age-related muscle loss; the same mass-versus-strength divergence shows up in the Nass MK-677 trial and elsewhere across the GH-axis class.

Tesamorelin — non-HIV muscle wasting and older-adult body composition (Stanley group, 2013–2019)

Sponsor: Investigator-initiated (NIH, MGH) plus Theratechnologies as drug supplier in some studies. Indication: Body composition and hepatic fat in non-HIV / older-adult populations. Design: A series of smaller randomized trials extending the HIV-lipodystrophy indication into broader metabolic populations. Result: mixed. The on-label HIV-lipodystrophy and HIV-NAFLD trials produced clean primary-endpoint wins (Falutz et al. 2007, Falutz et al. 2010, Stanley et al. 2014, Stanley et al. 2019). The non-HIV extension trials have been smaller, shorter, and more equivocal in their results.

This entry is honestly framed as a class-level pattern rather than a single failed pivotal. The Phase 3 pivotal evidence for tesamorelin sits in HIV-associated lipodystrophy, where the trials were large, blinded, and definitively positive. The off-label biohacker interest in tesamorelin focuses on non-HIV visceral-fat reduction, metabolic-syndrome-adjacent body composition, NAFLD outside the HIV context, and sarcopenic-obesity-adjacent older-adult body recomposition. The trial evidence in those non-HIV populations is substantially thinner. The Tesamorelin peptide page treats this distinction explicitly.

What the field learned: a regulatory-approved indication does not generalize to nearby off-label use cases without trial evidence specific to those populations, and the gap between the HIV-lipodystrophy efficacy data and the non-HIV recreational-use case is the central honest caveat that any tesamorelin discussion must communicate. The mechanism — GHRH-axis-mediated lipolysis and IGF-1 elevation — is plausibly active across populations, but plausible activity is not measured efficacy. A formal large-N non-HIV pivotal would resolve the question; none has been run.

Bremelanotide — failed male sexual-dysfunction development (Palatin Technologies, 2007/2008)

Sponsor: Palatin Technologies. Trials: Multiple Phase 2 trials of intranasal bremelanotide in male erectile dysfunction, ended in 2007–2008. Indication: Erectile dysfunction in men, then hypoactive sexual desire in men. Design: Various Phase 2 designs of intranasal route. N: several hundred across the program. Result: the intranasal program was halted in 2007 by Palatin following FDA concerns about dose-related blood-pressure increases. The molecule did not advance to a Phase 3 for any male sexual-dysfunction indication. The development program pivoted to subcutaneous bremelanotide in female hypoactive sexual desire disorder, which produced the positive RECONNECT trials (Kingsberg et al. 2019) and the June 2019 FDA approval of Vyleesi.

The bremelanotide / Vyleesi story is the modern reference case in the peptide field for how a development program reframes from a failed lead indication to an approved adjacent indication by changing route, dose, target population, and clinical-endpoint framing. The intranasal Phase 2 evidence in male erectile dysfunction was suggestive enough to motivate further development; the cardiovascular safety signal that emerged at higher intranasal doses and on FDA review halted the program. Palatin pivoted to subcutaneous dosing (which produced a cleaner pharmacokinetic profile with less blood-pressure variability), a female HSDD indication (where the regulatory standard for "clinically meaningful" was differently defined than for male ED), and on-demand cadence (rather than chronic daily dosing). The result was an FDA approval in 2019, eleven years after the male-indication program ended.

What the field learned: melanocortin agonism produces dose-related cardiovascular effects that constrained the development envelope at the intranasal route, and the dose / route / indication that ultimately reached the market was a substantially narrower envelope than the original program targeted. Off-label recreational use of bremelanotide in men today operates outside the indication and population the approval supports, on the implicit assumption that the subcutaneous PK profile produces less cardiovascular risk than the failed intranasal route. The PT-141 / bremelanotide peptide page treats the indication-by-indication evidence breakdown.

Selank — the Western RCT replication that does not exist (1990s–present)

Sponsor: V.V. Zakusov Research Institute of Pharmacology (Moscow). Indication: Generalized anxiety disorder and neurasthenia. Status: approved in Russia for GAD and neurasthenia in 2009 based on the active-comparator trial against medazepam (Zozulya et al. 2008, PMID 18454096). Western replication: none.

This entry is honestly framed as "absence of replication" rather than as a discrete failed trial. The Russian literature on Selank contains several active-comparator and open-label trials that report comparable anxiolytic magnitude to benzodiazepines plus a cleaner tolerability profile. The Western peer-reviewed literature has not produced an independent placebo-controlled replication of those results. No large Phase III RCT of Selank has been completed in a Western trial network for any indication, and no European or North American regulator has approved the molecule.

The honest framing across the corpus is that absence of replication is information. Selank may work essentially as the Russian literature reports; it may also produce smaller or different effects in Western populations under Western trial conditions; the data to discriminate between those possibilities does not exist. The single peer-reviewed comparative trial (Zozulya et al. 2008) was conducted at the institute that developed the molecule, was published in abridged form, and did not specify explicit randomization or blinding language. A Western replication would resolve the question; none has been undertaken. The Selank peptide page walks the evidence depth in detail, and the Russian-evidence critic response treats the methodology questions specifically.

What the field learned: a national regulatory approval grounded in a single-institute clinical trial does not transfer automatically to other jurisdictions, and the modal pattern for Russian-origin neuropeptides (Selank, Semax) in 2026 is local approval, international research-channel use, and unresolved Western-evidence questions. The absence of large independent replication is the central reason these molecules sit in research-use rather than rx-only space in most jurisdictions outside Russia.

Cerebrolysin — CASTA, the negative confirmatory stroke trial that the Cochrane review then synthesized (EVER Neuro Pharma, 2012 / 2020 / 2023)

Sponsor: EVER Neuro Pharma. Trial: CASTA (Cerebrolysin Acute Stroke Treatment in Asia), and the subsequent Cochrane Stroke Group reviews. Indication: Acute ischemic stroke. Design: CASTA — double-blind placebo-controlled randomized multicenter trial in 1,070 acute hemispheric ischemic stroke patients across Asia. Primary endpoint: confirmatory combined global directional test across modified Rankin Scale, Barthel Index, and NIHSS at day 90. Result: CASTA's confirmatory primary was not significantly different from placebo (Heiss et al., Stroke 2012, 43:630–636). A post hoc subgroup analysis in severely affected patients (NIHSS >12) suggested a favorable mortality trend.

CASTA is the largest single Cerebrolysin trial and the most consequential negative result in the molecule's stroke program. The Cochrane Stroke Group then pooled the broader Cerebrolysin acute-ischemic-stroke trial body in successive reviews — Ziganshina et al. 2020 and Ziganshina et al. 2023 (7 trials, 1,773 participants in the 2023 update) — and concluded with moderate-certainty evidence that Cerebrolysin "probably has little or no beneficial effect on preventing all-cause death" in acute ischemic stroke (risk ratio 0.96, 95% CI 0.65–1.41). The 2023 update also flagged an apparent increase in non-fatal serious adverse events, more prominent in the 30 mL / 10-day subgroup.

What happened next: smaller positive trials in motor-recovery rehabilitation contexts (CARS-1, CARS-2) have produced results favorable to Cerebrolysin on functional-recovery primaries (Muresanu et al. 2016) and have been pooled by manufacturer-affiliated meta-analyses (Bornstein et al. 2018) into a favorable interpretation of the broader trial body. The European Stroke Organisation and European Academy of Neurology joint guidelines on post-stroke cognitive impairment (Quinn et al. 2021) issued a weak recommendation against routine use of Cerebrolysin for post-stroke cognitive impairment given the balance of probable modest benefit against risk of serious adverse events.

What the field learned: a trial body that is large in patient numbers but concentrated in a relatively small set of investigator networks, with primary endpoints set as combined global directional tests that invite post-hoc reframing when the headline component is negative, will produce reviewer conclusions that diverge by reviewer affiliation. The manufacturer-aligned meta-analyses and the Cochrane synthesis examine the same trial body and reach opposite conclusions. The conventional 2026 Western-neurology reading favors the Cochrane interpretation. See the Cerebrolysin peptide page for the full evidence-base walk-through.

Fosgonimeton — LIFT-AD, the failed HGF/MET Alzheimer's pivotal (Athira Pharma, 2024)

Sponsor: Athira Pharma. Trial: LIFT-AD, NCT04488419. Indication: Mild-to-moderate Alzheimer's disease. Design: Phase 2/3 randomized double-blind placebo-controlled trial of once-daily subcutaneous fosgonimeton 40 mg versus placebo in treatment-free adults. N: 315. Primary endpoint: Global Statistical Test combining ADAS-Cog11 (cognition) and ADCS-ADL23 (function) at 26 weeks. Result: primary endpoint not met (treatment difference favoring fosgonimeton on ADAS-Cog11 was −0.70, p = 0.35). Key secondaries (ADAS-Cog11 and ADCS-ADL23 individually) also did not reach significance (Athira Pharma topline release, September 3, 2024).

Fosgonimeton was Athira's lead HGF/MET-system modulator and the indirect descendant of the academic dihexa research program. The trial reported numerically favorable plasma biomarker changes (NfL, GFAP, p-tau217, p-tau181, Aβ42/Aβ40) over placebo, and the company highlighted larger numerical treatment effects in more advanced disease and in APOE ε4 carriers as hypothesis-generating subgroup signals. The prespecified clinical primary did not separate from placebo.

What happened next: Athira announced in September 2024 that it would discontinue further development of fosgonimeton and focus on its oral HGF/MET modulator ATH-1105 in amyotrophic lateral sclerosis. The Alzheimer's indication for the dihexa-adjacent compound class is, as of 2026, without an active clinical-development program. The dihexa academic research program produced its own integrity story when the 2025 retraction landscape for the foundational potency claim was published (JPET retraction landscape 2025); the two stories are related but distinct — the LIFT-AD failure does not depend on the retraction questions, and the retraction does not by itself disprove the mechanism.

What the field learned: HGF/MET-system modulation in established symptomatic Alzheimer's disease at the dose and duration tested does not produce clinically meaningful slowing on regulatory-grade composite endpoints. The mechanism remains an active research question in other CNS indications (ALS in the Athira pivot); whether HGF/MET modulators have therapeutic potential at earlier-disease, longer-duration, or different-population designs is unaddressed. See the Dihexa peptide page for the academic-research context.

BPC-157 — the absence of any completed Phase II RCT in any indication

Sponsor: None at the Phase II level. Indications attempted: ulcerative colitis (PL 14736 / Pliva). Status: the PL 14736 program reached Phase II in inflammatory bowel disease; detailed efficacy results were never published in a peer-reviewed journal. As of 2026 there is no completed, published, peer-reviewed Phase II RCT of BPC-157 for any indication.

This entry exists because the absence of a failed trial in BPC-157 is informationally distinct from a successful trial in BPC-157. The molecule has the deepest mechanistic and animal-model literature of any peptide on this site, almost exclusively from Sikiric's group at Zagreb plus collaborators, and the smallest formal human evidence base relative to the breadth of its claimed effects. A negative trial would generate testable information; the current state of the literature generates no such information at the Phase II human-trial level.

The PL 14736 program reached Phase II for ulcerative colitis under Pliva's development envelope in the 2000s. Detailed clinical-trial results from that program were not published in a peer-reviewed journal. Whether the program ended because the trials failed, because the corporate strategy shifted, or for other reasons is not characterized in the public record. The musculoskeletal repair claims that drive most current biohacker use have never been tested in any completed published human RCT. See the BPC-157 peptide page for the full evidence walk and the no-human-RCTs critic response for the longer treatment.

What the field learned: the absence of completed trials is not equivalent to negative trials, but it is also not equivalent to validation. A peptide with strong mechanistic plausibility and no Phase II human trial sits in a different evidentiary position from a peptide with a failed Phase II human trial, and both sit at a substantial distance from a peptide with a positive Phase II human trial. The honest reading of BPC-157 in 2026 is that the molecule is mechanistically interesting, rodent-validated, and clinically untested — and that recreational use rests on extrapolation rather than on the controlled-trial confirmation that would resolve the open questions.

What this archive is for

The archive's purpose is to make negative results legible. A reader navigating the modern peptide field encounters a steady stream of positive signal — the Phase IIa success, the rodent mechanism, the practitioner testimonial, the conference poster — and almost no exposure to the parallel stream of negative signal that runs underneath. The trials catalogued here did not validate the molecules they tested at the indications they tested, and that fact matters even when the molecules continue to be used in adjacent indications, at adjacent doses, by adjacent routes.

Three patterns recur across the entries. First, the small-trial-to-large-trial dilution pattern — a Phase IIa signal that did not survive Phase IIb, as in AOD-9604, or a Phase II signal that did not survive Phase III, as in MMPOWER-3. Second, the primary-missed / secondaries-positive split — where structural or biomarker secondaries reach significance while clinical-functional primaries do not, as in ReCLAIM-2, ELAD, and (with one direction) LIFT-AD. Third, the mass-versus-function divergence — where a mass or biomarker endpoint moves while a strength or functional endpoint does not, as in ROMANA-1 / ROMANA-2 and in the Nass MK-677 sarcopenia-adjacent trial that the sarcopenia and peptides dossier covers. Each pattern has methodological implications for how to read future trial readouts in the same space.

The brand-aligned reading of failed trials is not schadenfreude — the molecules and their developers are doing serious work in hard indications — but it is also not protective. A field that takes its own evidence base seriously names the negative trials and walks the consequences. The page exists to do that consistently across the corpus.

Sources cited

• Karaa et al. 2018 — MMPOWER-2 dose-ranging Phase 2 of elamipretide in primary mitochondrial myopathy
• Karaa et al. 2023 — MMPOWER-3 Phase 3 trial of elamipretide in primary mitochondrial myopathy
• Ehlers et al. 2024 — ReCLAIM-2 Phase 2 of elamipretide in dry AMD
• Daubert et al. 2017 — first elamipretide HFrEF trial
• Thompson 2021 — elamipretide in Barth syndrome
• Aronne et al. 2024 — SURMOUNT-4 maintenance-withdrawal Phase 3 of tirzepatide
• Falutz et al. 2010 — Tesamorelin pooled Phase 3 in HIV-associated lipodystrophy
• Stanley et al. 2014 — Tesamorelin liver fat reduction in HIV
• Stanley et al. 2019 — Tesamorelin Phase 2 in HIV NAFLD
• Kingsberg et al. 2019 — RECONNECT bremelanotide Phase 3 in female HSDD
• Zozulya et al. 2008 — Selank vs medazepam in GAD and neurasthenia
• JPET dihexa retraction landscape 2025 — dihexa potency-claim retraction analysis