🐺 The Wolverine Stack
BPC-157 + TB-500 Under Audit

Evidence verdict

Beyond published studies: the total-evidence model

A journal-only review is incomplete for an old, easily copied experimental peptide. Trials require capital, manufacturing controls, regulatory strategy, and a sponsor with something valuable to protect. EOS therefore treats published research as one evidence lane, not the boundary of discovery.

The point is not to flatten these sources into one score. It is to ask whether independent lanes converge. Perception can reveal a signal; objective corroboration can strengthen an observation; controlled outcomes are still needed to establish causality.

Patents, funding, and the missing-results problem

The common statement that “a peptide cannot be patented” is not correct. A naturally occurring sequence may face novelty and patent-eligibility limits, but analogues, salts, formulations, combinations, manufacturing methods, delivery systems, and methods of use can receive protection.

BPC's own record demonstrates this. A 1994 patent covered a larger gastric-derived BPC material. A 2001 US patent claimed short organoprotective peptide sequences and listed the 15-amino-acid sequence now called BPC-157. A Diagen patent granted in Europe in 2017 covers stable pentadecapeptide salts, their preparation, pharmaceutical use, and therapy. Later Chinese manufacturing patents and newer oral-delivery and combination applications also exist.^16,17,18

That does not mean a sponsor has a strong modern monopoly over generic BPC-157. Older sequence rights may have expired, small peptides are comparatively easy to copy, and a sponsor must still finance manufacturing, toxicology, and clinical development. Weak exclusivity can reduce investment. But the evidence gap cannot be explained simply as “unpatentable.”

The human program that disappeared from view

A controlled oral BPC-157 phase-I study was registered in 2015. NCT02637284 planned 42 healthy volunteers, randomization, placebo controls, quadruple masking, pharmacokinetics, and six months of adverse-event follow-up. The registry has not been updated since 2015, now carries Unknown status, and posts no results.¹⁹ We cannot infer whether it completed, failed, or produced favorable findings. Its existence is important; its silence is not a result.

The development record is stalled rather than active: the only controlled BPC-157 human registration located for this audit (NCT02637284) has carried Unknown status with no posted results since 2015, and two 2026 registry entries that appeared to show new BPC-157 and TB-500 development were identified during verification as example/mock records and excluded from this audit.

Thymosin-beta-4 attracted real capital, and still stalled repeatedly

Full-length thymosin beta-4 had a substantial commercial development program under RGN-137, RGN-259, and RGN-352. NIH RePORTER confirms a $3 million 2010 RegeneRx award for nonclinical post-heart-attack development.²⁰ ClinicalTrials.gov also records programs withdrawn before enrollment because of manufacturing problems, terminated because of expired study material or slow recruitment, and stopped for business decisions.^{21,22,23,24,25,26} Some terminated studies posted results; others did not.

This broader history supports a balanced interpretation: thin evidence can reflect weak exclusivity, manufacturing failure, recruitment difficulty, endpoint strategy, financing, and partner decisions. None of those explanations establishes efficacy. Conversely, the ability to sell unapproved products without pivotal trials can reward marketing more than verification.

What the name hides

“Wolverine stack” is a market name, not a standardized scientific formulation. It usually means BPC-157 plus something sold as TB-500, but the label does not establish sequence, purity, amount, sterility, route, or whether the second material is full-length thymosin beta-4 or a shorter fragment.

That matters because evidence belongs to the molecule actually studied. Findings from pharmaceutical-grade full-length thymosin beta-4 cannot be transferred automatically to Ac-LKKTETQ, to one of its metabolites, or to an unverified commercial vial. A named blend therefore has two separate evidence problems:

1. Component validity: Did each characterized molecule improve the relevant outcome?
2. Blend validity: Has the actual combination been shown to add benefit without adding harm?

The Wolverine stack has not cleared either question in controlled human musculoskeletal research.

BPC-157: animal signal, weak human evidence

The strongest case for BPC-157 is preclinical. Rat experiments have reported improved gait-related function, load-to-failure, tissue organization, and histology after Achilles-tendon transection, tendon-to-bone injury, medial-collateral-ligament transection, quadriceps transection, and muscle crush.^1,2,3,4,5,6

These are relevant signals, but the record has important limits: small experimental groups, severe surgical injury models, incomplete reporting by modern risk-of-bias standards, little large-laboratory replication, and substantial concentration within the University of Zagreb research network. Uniformly positive findings also raise publication-bias and selective-reporting questions.

What has actually been published in people?

• Knee pain: A retrospective Florida clinic series contacted 16 patients. Twelve had received BPC-157 alone and four BPC-157 plus material reported as thymosin beta-4. Eleven of 12 in the BPC-only group reported significant pain improvement. There was no control group, diagnostic stratification, validated function measure, standardized follow-up, blinding, or structural-healing assessment.⁷
• Interstitial cystitis: An uncontrolled 12-woman bladder-pain pilot reported symptom improvement. It was not a musculoskeletal study and cannot establish musculoskeletal efficacy.²⁷
• Short-term IV safety: A two-person, three-day pilot reported no short-term adverse effects in the selected measurements. Both participants had previously received IV BPC-157. This cannot establish general, long-term, or rare-event safety and did not test healing.²⁸

A 2025 systematic review found 36 eligible papers: 35 preclinical studies and one clinical knee report. It classified the evidence as level IV–V.⁸ (The review summarizes the same knee series as 7 of 12 patients reporting relief beyond six months, versus 11 of 12 reporting significant improvement in the primary report; EOS uses the primary source's figures.) A 2026 ClinicalTrials.gov entry suggesting an active BPC-157 hamstring trial was identified during verification as an example/mock record and is excluded from this audit.

Publication-safe conclusion: BPC-157 has a repeated animal signal. It has not yet demonstrated tendon, ligament, muscle, cartilage, or bone healing in a published randomized human trial.

TB-500 is not full-length thymosin beta-4

Full-length human thymosin beta-4 is a 43-amino-acid endogenous actin-regulating peptide. In contrast, primary mass-spectrometry studies identified material sold as TB-500 as Ac-LKKTETQ: an acetylated seven-amino-acid segment corresponding to residues 17–23 of thymosin beta-4.^9,10

A 2024 metabolism study found that intact Ac-LKKTETQ did not produce the reported fibroblast scratch-wound activity in that experiment; one metabolite, Ac-LKKTE, did.¹¹ A separate analysis of internet products found that TB500/TB1000 contents were not consistently aligned with their descriptions.²⁹

Human studies of full-length thymosin beta-4 include short healthy-volunteer exposure and topical eye or skin formulations. Those studies provide limited, formulation-specific information. They do not establish that injectable commercial TB-500 repairs musculoskeletal tissue or is safe.^12,13,14,15

Direct musculoskeletal evidence is especially thin. No controlled human tendon, ligament, or skeletal-muscle repair trial was identified. In one full-length-thymosin-beta-4 mouse study of muscular dystrophy, a histologic regeneration marker improved, but strength, cardiac function, and fibrosis did not.³⁰ A tendon-related study used cells on an engineered scaffold; it was not an animal or human tendon-repair trial.³¹

A 2026 registry entry defining its investigational TB-500 as the thymosin-beta-4 17–23 fragment was identified during verification as a fictional example record and is excluded from this audit.

What independent chemistry, recalls, and safety systems add

Formal efficacy research asks whether a defined molecule changes an outcome. Real-world users face a prior question: is the vial what the label says it is?

Confiscated products establish identity in individual samples

Anti-doping laboratories have identified the BPC-157 sequence in confiscated vials by mass spectrometry.³² A separate analysis by a Belgian anti-doping laboratory (Ghent University's Doping Control Laboratory, DoCoLab) identified “TB-500” as Ac-LKKTETQ, the acetylated seven-amino-acid fragment, not full-length thymosin beta-4.⁹ These analyses cannot describe every supplier, but they demonstrate why label identity must not be assumed.

A large product-quality dataset changes the safety context

A 2026 preprint analyzed a publicly available independent-testing dataset containing 6,441 samples across 14 direct-to-consumer peptide categories, including BPC-157, TB-500, and thymosin beta-4. Depending on the authors' quality framework, 41.6%–71.1% of the pooled samples failed basic quality criteria and 15% had measurable endotoxin.³³ The accessible abstract does not provide compound-specific Wolverine failure rates, so the pooled percentages must not be presented as BPC-157 or TB-500 prevalence estimates. The study was a preprint, not peer reviewed as of this audit.

Recalls document failures; they do not estimate prevalence

A dated openFDA query returned nine unique Class-II recall records matching BPC-157 or thymosin beta-4, including one combined injectable. All nine cited lack of sterility assurance; initiation dates ranged from 2018 through 2025.^34,35 These records prove that product-level manufacturing failures occurred. They cannot show how common nonsterility is across an undefined market.

Spontaneous reports are signals, not rates

An openFDA/FAERS string-match snapshot returned 17 unique reports across BPC-157 and thymosin-beta-4 searches. In most BPC-string reports the product was coded as concomitant rather than suspect. Seven unique reports across both searches coded a matching product as suspect. Events included injection-site reactions, suspected contamination, hypersensitivity-type events, dyspnea, urticaria/facial swelling, labeling or dispensing errors, and ineffectiveness.^36,37 Extensive co-medication, illness, and uncertain product identity prevent causal attribution. FAERS cannot provide incidence or comparative risk.

What this lane establishes: formulation identity, sterility, purity, abundance, and endotoxin are part of the intervention, not side issues.
What it does not establish: that every product is defective or that the peptide caused every reported event.

Evidence for the combination

The familiar explanation (that BPC-157 works locally while TB-500 recruits repair cells systemically) is a market narrative assembled from separate mechanistic findings. The audit did not identify a controlled human or animal injury study comparing BPC-157, chemically defined TB-500, their combination, and placebo.

The four combination recipients in the 2021 knee series are not proof of synergy. Three of four reported significant pain improvement, compared with 11 of 12 receiving BPC-157 alone.⁷ Those tiny uncontrolled proportions are neither a valid comparison nor evidence that the combination outperforms one component.

Claims that the components “complement,” “multiply,” or “create whole-body healing” should therefore be described as hypotheses, not demonstrated outcomes.

What gray literature and practitioner material add

An expanded search covered OpenAlex, Europe PMC, Crossref, Semantic Scholar, Croatian university repositories, preprint servers, conference and professional material, and practitioner presentations. Dozens of BPC-157 dissertation and repository records were located, heavily concentrated in the University of Zagreb network. They were overwhelmingly animal, bee, review, or in-vitro projects, not hidden human treatment datasets.

Practitioner videos and clinic pages repeatedly used the phrase “case study,” but the accessible material did not provide an auditable human BPC-157/TB-500 series with a denominator, standardized baseline and follow-up, concurrent-care accounting, losses to follow-up, null outcomes, and systematic adverse-event collection. Claims without those elements remain professional opinion or marketing evidence, not clinical outcome data.

That does not make the material worthless. It helps identify the injuries being treated, outcomes practitioners believe matter, concurrent rehabilitation patterns, claimed responder types, and adverse effects future cohorts should measure. EOS will upgrade a practitioner source when its methods and underlying cases become inspectable; professional status alone is not enough.

Access barriers: what remains unseen

A retrieval failure is not a negative finding. This audit encountered material access barriers across the very places where people are most likely to exchange longitudinal experiences, photographs, imaging claims, laboratory results, clinician feedback, and batch-specific adverse reports. The inaccessible record is therefore an important source of uncertainty, not empty space.

Where the audit was incomplete

• Reddit: Standard anonymous access failed. PullPush and Arctic Shift supplied only partial archived records; some responses were empty or minimal. Deleted posts, private communities, media attachments, edits, and complete comment trees may be absent.
• YouTube: Transcripts were retrievable, but a complete reproducible export of comments across the selected videos was not available. One preserved page snapshot is not equivalent to an exhaustive comment corpus.
• Closed social spaces: Private Facebook groups, Discord servers, Telegram channels, paid communities, patient groups, and clinic portals were not accessible through public search.
• Short-form and dynamic platforms: TikTok and other algorithmic feeds were not systematically auditable. Search visibility, removals, account status, geography, and personalization can change what is exposed.
• Repositories and documents: Some university pages presented bot checks, some full texts were unavailable even when metadata was retrievable, and one current preprint page blocked automated access although its abstract and metadata were available elsewhere.
• Non-public clinical material: Practitioner imaging, pathology, clinic outcome logs, and patient records cannot be evaluated unless their owners release anonymized source material with consent and provenance.

Why the missingness may be non-random

People with detailed injury timelines or medical documentation may prefer private recovery groups over public forums. Sellers and moderators may remove failures or adverse reports; platforms may remove sourcing discussions; dissatisfied users may leave without follow-up; successful users may stop posting; and people with privacy-sensitive imaging or diagnoses may avoid public disclosure. These forces could hide favorable evidence, unfavorable evidence, or both. They can also distort the visible balance in different directions.

It is therefore inaccurate to say “no documented experiences exist.” The defensible finding is narrower: no accessible case in the retrieved corpus supplied independently inspectable before-and-after records sufficient to demonstrate structural healing. A potentially valuable reservoir of inaccessible experience and documentation may exist, but its size, quality, direction, independence, and authenticity are unknown.

What access failure establishes: the searchable public corpus is incomplete and may be systematically selected.
What it does not establish: that hidden documentation exists, that inaccessible reports are favorable, or that their conclusions would survive verification.

What the sampled public reports say

EOS reviewed a bounded, purposive public-source sample collected on 21 July 2026. The analytical subset contained 33 first-person accounts: seven YouTube creators, 14 identifiable posters across three public forum threads, and 12 archived Reddit posts selected because they contained an outcome or possible adverse event. Creator commentary, questions without outcomes, duplicate reposts, and supplier-only claims were not counted as independent experiences.

This was not a random sample. It cannot estimate how many people use the stack, the percentage who improve, the percentage who experience harm, or whether one outcome is more common than another.

Sample auditability: the underlying transcripts, threads, and archived posts were not published alongside this brief, so an independent reader cannot reproduce the exact 7/14/12 account counts. The counts are internally consistent and were coded against explicit inclusion rules (first-person outcome or possible adverse event), but they should be treated as a bounded internal sample rather than an independently auditable corpus.

Positive reports existed, but attribution was weak

Some people described less pain, greater range of motion, earlier return to training, or faster-than-expected recovery after tendon, muscle, joint, or surgical injuries. Several mentioned MRI, surgery follow-up, or a surprised clinician, but the underlying records were not accessible. Rehabilitation, immobilization, surgery, training changes, growth hormone, collagen, sleep, nutrition, massage, and natural recovery frequently occurred at the same time.

Null, mixed, and negative accounts also appeared

The sample included people reporting no noticeable improvement, improvement only after rehabilitation began, help with acute soft-tissue problems but not chronic structural pain, or less benefit than expected. One creator stopped after more than two months and later attributed a larger improvement to better sleep and arch support. Another clinician-user explicitly said he could not separate the effect of time, rehabilitation, and the peptides.

Possible adverse experiences were part of the signal

Reports included injection-site irritation, abdominal tenderness, increased soreness or stiffness, anhedonia, insomnia, hives with lip swelling, cystic acne, muscle tension, and abdominal pressure. These events are not verified causal reactions. Product identity, contamination, concurrent compounds, underlying illness, and timing frequently prevented attribution.

What the public corpus supports: There is meaningful attention and a heterogeneous set of hypothesis-generating experiences.
What it does not support: unsupported population-size or consistency claims, a response rate, a safety rate, or proof that the blend accelerates healing.

Safety, product identity, and sport

• Human safety is not established: BPC-157 has only tiny, uncontrolled published human reports. Full-length thymosin beta-4 has more human exposure data, but those results do not transfer automatically to TB-500 fragments or unverified commercial products.
• Product risk is part of the intervention: Sequence, amount, sterility, aggregation, impurities, and storage can differ from the label. A clean pharmaceutical trial cannot validate a gray-market vial.
• Angiogenesis is not a demonstrated cancer outcome: Repair-related angiogenic pathways create theoretical questions. It would be inaccurate to claim that administered BPC-157 or TB-500 has been shown to cause cancer in humans; it would also be inaccurate to claim long-term cancer safety.
• Sport: The 2026 WADA Prohibited List identifies thymosin-beta-4 and derivatives such as TB-500 as prohibited at all times. Athletes must use current anti-doping guidance rather than vendor descriptions.³⁸

Evidence Scorecard

The honest conclusion

The Wolverine stack is not an obscure experiment. It is one of the most widely used unapproved recovery protocols in the peptide community, and this audit found a large, heterogeneous public-experience record behind that adoption (people reporting real functional gains, people reporting nothing, people reporting setbacks and adverse events). That scale is itself evidence: evidence of attention, of perceived benefit, and of a question the community has effectively been running at population scale for years. It is not, by itself, evidence of efficacy. Both things are true at once, and a credible audit has to hold them together.

The controlled record is thin, and it must be described accurately: BPC-157 has repeated positive animal models but no published randomized human healing trial; the only published human musculoskeletal report is one uncontrolled knee-pain series; TB-500 is a chemically distinct fragment, not full-length thymosin beta-4; and no controlled study has tested the blend itself or demonstrated synergy. Those are gaps in controlled evidence, not a finding that nothing is happening. The single most important correction this audit makes to the marketing narrative is narrower than "it doesn't work": it is that nobody has yet run the study that would let anyone say, with a straight face, how well this works, for which injuries, at what dose, with which product, and at what risk.

The absence of that study is not the community's fault, and it is not a verdict on the people using the stack. It is a structural feature of the market: an unapproved peptide that cannot justify pivotal-trial capital while unregulated sales reward marketing instead of verification. The people sharing their results in forums, YouTube, and recovery groups are doing the work the research system has not done, but self-reports, even in large numbers, cannot substitute for a trial that controls for rehabilitation, time, placebo, and product identity. The community's reports are the reason this question deserves a real study, not the reason to pretend one already exists.

What would move this from "widely used, not yet tested" to "tested" is specific and achievable: a controlled trial of a characterized product against placebo with standardized rehabilitation, structural endpoints where possible, and honest adverse-event collection. Until that exists, the evidence-honest position is not dismissal and not endorsement. It is respect for what the community has documented, precision about what that documentation can and cannot prove, and an insistence that the next step is a better study, not a better argument.

Widely used. Not yet tested. The people are ahead of the research. Look closer. Claim less.

Methods and limitations

Scientific and development-record searches were performed through 21 July 2026 across PubMed, Europe PMC, ClinicalTrials.gov, NIH RePORTER, Google Patents and patent-family records, SEC filings, Crossref/OpenAlex, Croatian OAI university repositories, preprint servers, openFDA recall and FAERS APIs, FDA materials, WADA materials, and citation chaining. Studies were separated by molecule, sequence, terminal modification, formulation, route, species, and endpoint. Pain was not treated as proof of structural repair. Patents were not treated as efficacy evidence; registrations without posted results were not treated as positive or negative outcomes; spontaneous reports were not treated as causal events or rates.

Public-experience searches covered accessible YouTube videos/transcripts, Professional Muscle, Rokslide, and Reddit records accessible through public archive APIs when direct Reddit pages were blocked. Sources were coded for first-person use, condition, reported outcome, timeline, objective corroboration claimed, concurrent interventions, adverse experiences, and commercial conflict. Duplicate content was counted once.

Public-source sampling remains incomplete. Standard Reddit access failed; partial public archives did not preserve every post, comment tree, edit, media attachment, or deleted record. YouTube transcripts were retrievable, but a complete reproducible comment export was not. Private communities, clinic-held documentation, short-form dynamic feeds, deleted material, non-English discussion, and inaccessible comments may differ materially. Because access is likely related to privacy, moderation, commercial interests, platform rules, and user follow-up behavior, this missingness may be non-random. The sample intentionally cannot support prevalence estimates or a claim that no stronger documented experiences exist elsewhere.

References