General information only, and not medical advice. Nothing here is a recommendation for or against any treatment for any individual, and none of it substitutes for a conversation with your own doctor, who knows your history.

The short version

  • Peptides include some of the best medicines ever made, whereas most of what is sold as a “research peptide” has never been shown to do anything in a human, and both of those are true at once.
  • BPC-157, the one everyone asks about: no published human trial after roughly 30 years, and nobody has identified what it binds to.
  • TB-500: the parent molecule was trialled on skin wounds and did not beat placebo, and nothing at all supports injecting it into a muscle tear.
  • CJC-1295, ipamorelin, MK-677: all raise growth hormone, none has human evidence for muscle, fat loss or recovery. MK-677 also raises blood sugar.
  • MOTS-c: never tested in a human at all.
  • Melanotan-2: documented serious harm, including kidney failure and melanoma case reports.
  • Tesamorelin and PT-141 are FDA-approved medicines, though neither is registered in Australia. GHK-Cu is fine on your skin.
  • Of 44 research-chemical products analysed, only 18 contained the amount of active ingredient stated on the label.
  • “Natural” is not a safety category, and under the law these are drugs.
  • The July FDA vote covered what pharmacies may legally compound, and it was not an approval.
  • Your friend’s recovery story cannot settle it: pain improves on its own, people change several things at once, and belief measurably reduces pain.
  • For a sore tendon, progressive loading is the thing with human trials behind it.
  • One question does most of the work: did anyone give this to people and compare them against people who got nothing?

Someone at your gym is injecting BPC-157 for a sore shoulder while someone in your feed sells MOTS-c as “exercise in a syringe”. This guide asks what has actually been tested in people, then gives you a way to assess the next compound without waiting for another guide.

Peptides include some of the best medicines ever made, whereas most products sold as “research peptides” have never been shown to do anything in a human.

First, these are drugs

A peptide is a short amino-acid chain, which describes its chemistry without telling you whether it works or is safe. Insulin and semaglutide (Ozempic) are peptides, as is a vial of powder bought from a website that takes Bitcoin.

The law defines a drug as anything intended to treat or prevent disease, or affect the body’s structure or function. Injecting something to accelerate tendon healing is therefore drug use. Market language obscures this: “short chain of amino acids” sounds like nutrition, “research chemical” like a laboratory, and “peptide” like something already in your body.

Calling these products natural adds no safety information. Insulin is natural and can kill at the wrong dose; botulinum toxin is natural too. Every vial in this market is synthetic and manufactured, often after modifying the original molecule. Natural is not a safety category.

What the evidence actually shows

The compounds below are roughly ordered by how often they come up.

BPC-157, sold for tendon, ligament and gut healing. After about 30 years of claims, no human randomised trial has been published for any use. The only randomised human study concerned ulcerative colitis, never appeared as a full paper, and delivered the drug by enema, which is nobody’s protocol. Its difference between treatment and placebo is small enough to fit “no effect at all”. Nobody has identified what BPC-157 binds to, the protein it is supposedly a fragment of has never been found in a human, and most publications come from the group that discovered it. There is also a safety gap: the proposed pathways that grow blood vessels and move cells overlap with pathways used by tumours, while nobody has studied whether this matters. No long-term safety data exist in anybody.

TB-500, usually stacked with BPC-157 for muscle and tendon repair. The parent molecule has some human evidence when applied topically to skin wounds, though less than the usual account suggests. Both trials were powered for safety rather than efficacy, and complete wound closure was no better with the drug than placebo. No evidence supports injecting it into a muscle tear. The trials also used the full parent peptide, whereas the market sells a smaller fragment that cannot be assumed to behave like its parent.

CJC-1295 and ipamorelin, sold to raise growth hormone. Both raise it, but neither has human evidence for muscle, fat loss or recovery. CJC-1295 was abandoned after a trial participant died of a heart attack; ipamorelin was abandoned after failing the medical endpoint it was being tested for.

The premise behind this group has already been tested through the more direct route of giving people growth hormone. Two systematic reviews in the Annals of Internal Medicine found small body-composition changes and more side effects in healthy older adults, including joint pain, carpal tunnel and a drift towards diabetes. Across 27 studies of athletes and fit young people, measured lean mass increased while strength and exercise capacity did not, with hints that exercise capacity became slightly worse. The direct route failed to deliver the outcome these compounds chase indirectly. Raising growth hormone establishes a mechanism, not a result.

The evidence is also old. The latest randomised trial of any kind on MK-677 was in 2018 and measured a blood marker rather than strength. The newest evidence about a noticeable outcome is roughly 18 years old, while the marketing has continued to grow. Static evidence beside expanding claims is a useful pattern to recognise.

MK-677, an oral version of the same idea. The best study ran for a year in healthy older adults. IGF-1 rose substantially and lean mass rose about a kilogram, which is the number the marketing quotes, but fat rose as well. Strength and physical function did not improve, while fasting blood sugar rose and insulin sensitivity worsened. A separate trial in adults was halted after more heart failure occurred in the treated group.

MOTS-c, sold as exercise in a syringe. It has never been tested in a human at all.

5-amino-1MQ, sold for a “broken metabolism”. Mice only, and the impressive-sounding fat loss figure refers to one dissected fat pad rather than the whole animal. The version sold to people is a capsule, and nobody has published whether it even survives digestion.

Melanotan-2, sold for tanning. This one has documented serious harm, including muscle breakdown, kidney failure and case reports of melanoma. If you take one thing from this list, make it this one.

GHK-Cu, a copper peptide sold for skin and hair. The topical wound-healing evidence is weaker than usually presented (a larger confirmatory trial failed in 1994), and the cosmetic evidence is weak. It is benign on your skin, and there is no case for injecting it.

Tesamorelin and PT-141 are the two in this conversation that hold an actual marketing approval, tesamorelin for a specific pattern of fat redistribution in HIV, PT-141 for low sexual desire in women. Both are FDA approvals: neither is registered on the Australian ARTG, so in Australia they are unapproved goods reached through the Special Access Scheme or compounding. Where they are approved they have known doses, known side effects and known monitoring, which is the point. They are also not what most people are buying.

What is actually in the vial

Separate from whether a compound works, there is the question of whether you are receiving it.

When researchers chemically analysed 44 products sold through research-chemical channels, only 18 contained the amount of active ingredient stated on the label. Four contained no active compound at all, and 11 contained something that was not on the label.

A European analysis of falsified peptide products found purity ranging from 5 per cent to 75 per cent, along with arsenic and lead at up to 10 times the permitted limit for an injectable, in the more toxic inorganic form. That study looked at a specific illegal supply chain rather than everything on the market, so treat it as an illustration of what is possible rather than a survey of your particular vendor. It remains a reasonable thing to be frightened of, given that you are putting the contents under your skin.

A review outside the podcast world

While I was writing this, the American Journal of Sports Medicine published a scoping review of BPC-157, TB-500, CJC-1295, MK-677, ipamorelin and GHK-Cu across bone, fracture, muscle, tendon, ligament, meniscus and cartilage. It comes from mainstream, peer-reviewed orthopaedics and independently reaches the same conclusion.

Animal work makes up 67% of the publications it found, mostly in rats, while the human literature is “limited to a handful of investigations, most lacking robust controls or rigorous study designs”. MK-677 has the strongest evidence of the six, with five properly randomised trials, all in older people or hip-fracture patients rather than athletes. Those trials produced heart failure in one treatment arm and no functional benefit anywhere. “Best studied” does not amount to a recommendation. The review says the claimed benefits “remain unsubstantiated by current human trials” and advises against using these compounds as a replacement or adjunct for standard orthopaedic care.

The paper does offer one concession absent from the podcasts: “modest improvements at best” in metabolic bone health and degenerative knee pain. This appears to be a small human signal for the injury and recovery uses rejected elsewhere in this post, but both parts rest on studies already discounted here.

Degenerative knee pain comes from one study. Seventeen people received a BPC-157 injection into the knee for unrelated problems including osteoarthritis, meniscal tears and ligament sprains; four also received TB-500. Six months to a year later, investigators telephoned them to ask whether they felt better, and 14 of 16 said yes. The retrospective study had no control group and measured nothing. The review grades it as level 4 evidence, one rung above expert opinion on its scale, while its patient-counselling table still says BPC-157 cannot be recommended. It contains every reason a recovery story cannot establish efficacy.

Metabolic bone health means MK-677 blood tests rather than better bones. A bone-formation marker rose 29% while a bone-breakdown marker rose 23%, showing turnover rather than net gain. One trial found greater density at the femoral neck but none at the spine or the rest of the hip, and nobody broke fewer bones. The review says these changes “did not yield reductions in fracture risk in any trial”; glucose intolerance developed alongside them, and the MK-677 trials showed “an absence of clinically meaningful outcomes”.

The concession dissolves on inspection into movement in a blood marker and one uncontrolled phone survey. I first flagged it while the full paper remained behind a paywall because it cut against my argument. Having now read it, the studies underneath it do not change the conclusion.

Tewari et al., Peptide Supplements and Their Therapeutic Applications in Sports Medicine, American Journal of Sports Medicine, 11 August 2026. Abstract on PubMed.

Why your mate’s recovery story cannot settle it

The reasoning here applies well beyond peptides.

Symptoms fluctuate, and treatment starts near the worst point. Tendon and joint pain flare and settle. People seek a fix near the peak, when improvement is most likely regardless of what happens next. This regression to the mean can make an inert treatment appear effective.

You changed more than one thing. Almost nobody starts an injection while holding everything else constant. They rest, modify training, sleep better, start physio or take an anti-inflammatory. The injection gets credit because it is the most memorable and expensive change.

Belief changes pain. In a 2002 trial, patients with arthritic knees were randomly assigned to real keyhole surgery or a sham operation. The surgeon made skin incisions, splashed saline around for the sound, then closed without touching the joint. After two years, sham surgery produced the same pain relief and ability to walk and climb stairs as the real procedure. The patients were not lying; their pain improved for reasons unrelated to what the scalpel did. If a fake knee operation buys two years of measured improvement, a better elbow after an injection says little about the injection.

Only successes get reported. The person who improves posts about it; the person who feels nothing stops quietly and disappears from the evidence you see.

A mechanism is not a result. “It increases blood flow to the tendon, and tendons need blood flow to heal” sounds persuasive, especially to clever people. Medicine is full of plausible mechanisms that turned out backwards. A coherent story justifies running a study without substituting for one.

Find the comparison group. Retatrutide is an investigational obesity drug in phase 3 trials and is not approved anywhere. One trial reported about a 76 per cent improvement in knee-pain scores, which is the headline number; placebo improved about 40 per cent. The attributable effect is the gap between them. A single impressive percentage is incomplete until you know what happened to the people who received nothing.

Medicine has made exactly this mistake, repeatedly

The reason this standard exists is that the profession kept fooling itself in exactly this way, at scale.

Fen-phen. Two appetite suppressants prescribed together for weight loss. The combination was never approved and never submitted, and the evidence behind it was a single small study. It worked, people lost weight, and millions of prescriptions were written by trained doctors whose patients came back lighter and happier. Then a cardiologist reported a cluster of women with damaged heart valves. The damage was silent, so there was no symptom to notice and no test anyone was running. It came off the market in 1997.

Hormone replacement therapy. Observational data suggested it protected women against heart disease, the mechanism made sense, and it was prescribed on that basis for years. When it was finally tested in a large randomised trial, the cardiovascular benefit was not there, and parts of the trial found harm. The women in the observational studies had been healthier to begin with in ways nobody had adequately accounted for.

Beta-carotene. Antioxidants should reduce cancer risk, the reasoning was sound, and the epidemiology was encouraging. Two large trials gave supplements to smokers, and lung cancer went up rather than down. Both were stopped early.

Suppressing abnormal heartbeats. Extra beats after a heart attack predict death, so suppressing them should save lives. The CAST trial found that the drugs which suppressed them killed more people than placebo.

Stents for stable angina. You can see the blocked artery on the image and you can see it open afterwards, which is about as satisfying as a mechanism gets. Trials comparing it against medical therapy did not find the reduction in heart attacks that everyone expected, and a sham-controlled trial found much less symptom benefit than the procedure’s reputation implied.

Each treatment had a plausible mechanism and expert support, reinforced by satisfied patients who felt better. Only a comparison group exposed the error. That is the entire argument for wanting one, and it is not an appeal to authority because the authorities were wrong in every case.

The lesson runs in the other direction too. Beta blockers were considered obviously dangerous in heart failure, because suppressing the heart’s contractility in someone whose heart already pumps poorly sounds reckless. They are now standard treatment and they save lives. Plausibility talked people out of something that worked.

You already reason better than this everywhere else

Almost nobody applies these standards to anything else they spend money on, and they are correct not to.

You would not buy a car because one friend liked his. You would want to know the fault rate for that model, because you understand that your friend is one person, that he chose the car himself, and that he would rather not conclude he wasted $40,000. His report is contaminated by the same thing that makes it feel trustworthy, which is that he has skin in it.

You do not judge a builder by the testimonials on his own website. You go looking for the complaints, because you know which side of the ledger a seller publishes. A peptide vendor’s site is a wall of before-and-after photos, and it is the same ledger.

You discount a real estate agent’s valuation of a house he is selling. Not because he is lying, but because you know which way his interest points. The person telling you a compound is safe and effective is usually the person shipping it.

You do not accept “nobody has complained yet” about structural work on your house. You know that nobody has been up in the roof to look. That is exactly the position with long-term peptide safety: there is no registry, no follow-up, and no one reporting. Silence from a system where nobody is checking is not a safety record.

You know a bigger engine does not automatically mean a faster car, because weight and gearing and aerodynamics all get a vote. That is the same instinct you need for “it increases blood flow, and tendons need blood flow”. A plausible mechanism earns a test, and the test is what tells you the answer.

You are right in every one of those cases, and the reasoning transfers exactly. The reason it usually does not get transferred is that a car costs money up front and visibly, whereas the cost of a peptide that does nothing is deferred, invisible, and easy to attribute to something else.

What about the FDA news?

In July 2026, an FDA advisory committee recommended adding six peptides, including BPC-157, to the list of ingredients available to compounding pharmacies. Reports describing this as FDA approval confused permission to obtain a raw ingredient with evidence that the resulting product is safe or effective. The regulation governing the list says explicitly that it establishes neither.

The committee’s scientific staff had recommended against all seven substances under review. Its vote is advisory rather than binding, and any change requires rulemaking that will extend into 2027. “The FDA just approved it” will still become a sales line.

If you have a sore tendon

This is what most people are actually trying to solve, so it deserves a direct answer.

Progressive heavy slow resistance training has human trials behind it. It reduces pain, remodels the tissue, increases blood flow around the tendon several fold, and lays down and reorganises collagen in a way no injection does. Corticosteroid and platelet-rich plasma injections do not beat saline placebo when proper loading is in place; for tennis elbow, steroid produces a worse result at one year.

The evidence is strongest for Achilles and kneecap tendons and weaker at the elbow. Loading also has a real failure rate, so “just load it” does not answer someone who still hurts after 12 months of good rehabilitation. This market sells directly to that person, for whom medicine does not yet have a good answer. Treating the gap as a lack of discipline does not fill it.

The question to carry with you

The list will have changed in a year, so memorising it matters less than asking one question before anything goes into your body:

Did anyone give this to people and compare them against people who got nothing?

No statistics are needed. If the answer is no, all you have seen is people who took it and felt better, which is exactly what an inert treatment could produce. If the answer is yes, find out how the comparison group did; that is the number advertising leaves out.

Two further questions take seconds:

Who is telling me, and what do they get if I believe them? This tests the direction of pressure rather than honesty. You already make the distinction with car salesmen.

If this were doing nothing, would I be able to tell? For a sore shoulder likely to settle anyway, the answer is usually no. Know that before spending six months and several hundred dollars to find out.

Every compound starts at no demonstrated effect until somebody shows otherwise. The burden belongs to the claimant, who is usually also the seller.

Appendix: details worth having if you go looking

Several specifics in this area get repeated in a slightly wrong form, including by people who are broadly making the right argument. If you chase these claims yourself, here is what the underlying records actually say.

The CJC-1295 death was not a healthy volunteer. You will often hear that a healthy subject in a trial died of a heart attack. The 2006 ConjuChem study was a multicentre randomised double-blind phase 2 in about 192 participants with HIV-related visceral obesity. One participant died of a myocardial infarction roughly two hours after his 11th weekly injection. The trial physician attributed it to pre-existing asymptomatic coronary artery disease and judged it unrelated to the drug. That attribution was never independently confirmed, because no results from the programme were ever published. The death happened and the programme did stop, so the substance of the concern holds, but the version with a healthy volunteer in it is more alarming than the record supports.

The TB-500 cardiac trial never ran at all. The common claim is that thymosin beta-4 failed a phase 2 trial for cardiac repair after heart attack despite good results in mice. The registry entry is NCT01311518, sponsored by RegeneRx: status withdrawn, reason given as “Trial never initiated”, actual enrolment zero. It was stopped by an FDA clinical hold relating to manufacturing compliance at a contract manufacturer, which the company stated was not directed at the drug’s safety or development plan. A manufacturing problem is not an efficacy result. The damaging finding is separate: the striking mouse result failed independent replication in 2012.

The thymosin beta-4 skin-wound trials did not beat placebo on wound closure, and the number usually quoted has no comparison group in it. You will see the figure that about 25% of patients achieved complete healing on the 0.03% dose. That is the American Journal of Sports Medicine review’s own summary of these trials, and it is one treatment arm of three with nothing to compare it against. The comparison group exists. Both RegeneRx phase 2 trials registered safety and tolerability as their primary endpoint and wound healing as a secondary one, and their posted results give complete closure at day 84 as 12 of 55 on drug against 4 of 17 on placebo in venous stasis ulcers (NCT00832091), and 8 of 54 against 3 of 18 in pressure ulcers (NCT00382174). Both drug figures pool all three doses, because neither posting reports closure by dose. Those are 21.8% against 23.5%, and 14.8% against 16.7%. Resist reading the direction of that: the gap is under two percentage points each time, and with 17 and 18 patients on placebo, one more healed patient would move the placebo figure by about six points. This is a drug doing nothing detectable, not a drug doing harm. The other figure in circulation, that it accelerated healing by almost a month, comes from a review co-written by the compound’s own investigators and is restricted, in their words, to “those patients that did heal”. As for how a peer-reviewed review came to print the flattering half, it says so itself in its limitations: the authors searched one literature database and nothing else, so the registry postings were never in scope. Peer review means qualified people read the paper. It does not mean somebody chased every number back to where it came from.

MK-677 did increase fat, and the “no functional benefit” result is weaker than it sounds. In the one-year study in healthy adults aged 60 to 81, fat-free mass rose 1.1 kg against a 0.5 kg loss on placebo. Limb fat also rose, by 1.1 kg against 0.24 kg (p=0.001), and body weight rose 2.7 kg against 0.8 kg, though total and visceral fat did not differ significantly. Fasting glucose rose 0.3 mmol/L and insulin sensitivity fell. On the absence of strength and function gains, the authors state plainly that “study power (duration and participant number) was insufficient to evaluate functional end points”, so that null, in this trial, is a gap in the evidence rather than a demonstration of no effect.

Tesamorelin’s cognition trial was positive, not inconclusive. A 2012 randomised trial in 152 adults aged 55 to 87 found improved executive function, and to a lesser extent short-term verbal memory, over 20 weeks. It is sometimes described as showing nothing. That is a separate question from whether the effect matters clinically or generalises, but the trial did not come back empty.

The BPC-157 colitis abstract is Ruenzi, not Renzi. If you go looking for the single randomised human trial, it is indexed under Ruenzi (sometimes Rünzi) et al., Gastroenterology 2005, and it exists only as a meeting abstract.

The contamination figures come from different studies. The “only 18 of 44 products matched their label” analysis is a 2017 examination of products sold as SARMs through research-chemical channels. The 5 to 75 per cent purity range with arsenic and lead contamination is a separate 2018 analysis of 10 falsified peptide products from suspected illegal internet pharmacies in Belgium. Both are alarming and neither is a current survey of what any particular vendor ships.

Credits, and where to go for the long version

Almost none of the underlying work here is mine. Four episodes did the heavy lifting.

Barbell Medicine, and particular thanks to Jordan Feigenbaum. Two episodes form the backbone of this post: The Peptide Market Audit, which goes compound by compound, and The FDA Voted for 6 Peptides, which is the clearest account of the July vote I have found. He reads the primary documents rather than the coverage, distinguishes “this failed” from “nobody looked”, and states what would change his mind.

Austin Baraki, who is on the Peptide Market Audit and the crossover below rather than the solo FDA episode, is where much of the reasoning material here comes from.

Docs Who Lift, Spencer and Karl Nadolsky, hosted the crossover episode with both of the above. It is the best single starting point if you only listen to one thing, and their run through biological plausibility misfiring in both directions is where the CAST and beta blocker examples come from.

Peter Attia, episode #403, gets to much the same place by a different route, and describes the grey market as the salvage yard for compounds the pharmaceutical industry tested and walked away from.

Where I have corrected any of them, it is in the appendix. Any errors remaining are mine.