BPC-157 vs NSAIDs: Repair or Suppress?

How a torn tendon repairs itself

A tendon is rope: bundles of collagen fibres running in parallel, anchored into bone at each end. Tear it and the body runs a sequence.

First it swells. Swelling is not just the part that hurts — it is the opening instruction. The chemicals that make an injury hot and sore are the same chemicals that call in the cells that clear the wreckage and start laying down new rope.

Then it rebuilds, and rebuilding runs on blood. Blood brings the oxygen and the repair cells. A tendon has very little blood running through it, far less than muscle, which is why a torn tendon takes months to heal and a torn muscle takes weeks.

Last it re-anchors. Where tendon meets bone there is a gristly pad that grips one to the other. That pad is the weakest point of the whole repair and the place a healed tendon tears again.

Three things decide how well you end up: whether the opening instruction is allowed to fire, how much blood reaches the damage, and whether that pad rebuilds properly. Everything below is what each of these two does to those three things.

What the painkiller does to it

It reduces the pain, and it does that reliably. Take ibuprofen 400 mg and roughly two people in every five get at least half their pain gone who would not have got it from a dummy pill. That figure comes from about 460 separate trials covering roughly 50,000 people, which makes it one of the better-established numbers in medicine.

It produces that relief by shutting down the swelling. That is the same swelling that is the opening instruction, so the relief and the interference are not two separate effects — they are one effect, seen from two ends.

What that does to the tissue has been measured in rats. Both the newer and the older painkillers left a repaired tendon needing less force to tear off the bone again, and less stiff. The gristly pad at the bone was patchy and half-formed, where untreated animals had rebuilt it by four weeks. Giving the drug later does not avoid this: the damage showed up when it was given in the first five days, again at days 6 to 14, and again at days 11 to 20 after a repair.

The steroid injection is the stronger version of the same thing. It relieves for weeks to months, it changes where you end up a year later not at all, and it rebuilds nothing. What it buys is time before a decision about surgery.

Taken for months rather than days, the painkillers carry their own bill: serious stomach bleeds and ulcers run about four times higher on ibuprofen and about four times higher again on naproxen, heart failure risk roughly doubles across the class, and for every thousand people taking one for a year there are three extra heart attacks or strokes, one of them fatal.

What BPC-157 does to it

It grows new blood vessels into damaged tissue. More vessels means more blood reaching the part of the repair that was starved of it — which is the second of the three things above, and the one a tendon is worst at.

In rats, a tendon cut clean off the heel bone and stitched back came away needing more force to tear again, stiffer, with the collagen fibres lying straight rather than scarred and more vessels grown into the join. The dose was 10 micrograms per kilo of body weight a day. About 150 animal studies exist and they point the same way. In two of them it undid damage a steroid injection had already done to a healing tendon.

In people it has been given in five studies, about 130 people in total. Every one of them was looking for side effects. Not one measured whether anything healed.

At the amount people actually inject — 250 to 500 micrograms a day for four to eight weeks — no trial has ever been run. What exists at that amount is 37 people who wrote down in public what happened to them: 18 said it helped, 5 said partly, 10 said nothing happened, 4 said something got worse. Nobody tested what was in their vials, nobody was randomly assigned, and people who improve write about it more often than people who do not.

What follows from all of that

BPC-157 grows new blood vessels into damaged tissue, which is the step the drug's mechanism works against. In rats, at 10 µg/kg a day, an Achilles cut off the heel bone came back taking more load before it failed, stiffer, with the collagen running straight instead of scarred and more vessels grown in — across roughly 150 animal papers and at least six injury models. In two of those studies it reversed the healing damage a corticosteroid injection had already done. In rats given diclofenac, it blocked the drug's entire gut, liver and brain cascade. In people, five early-phase studies covering about 130 people measured safety and never measured healing; at the dose and duration people actually run — 250 to 500 mcg a day for four to eight weeks — no trial has ever been run at all, and the counted first-person record at that exposure is 37 accounts stating an outcome: 18 helped, 5 partly, 10 nothing, 4 worse.

Held to one standard, the ledger is this. Neither has a finished controlled trial showing it rebuilds a torn tissue in a person — the peptide has none, and no anti-inflammatory has ever been shown to heal one either. In the animal work where both have been measured, they move the repair in opposite directions: the drug lowers the strength of the healed join, the peptide raises it. And the counted human record at the exposure people actually run belongs only to the peptide, because nobody has run a trial there.

If the only question is what will hurt less in the next four hours, the answer is the drug. The price is a join that fails at a lower load.

Now the two things that make this a real question rather than a rout.

For back pain specifically, the drug's own trials are unimpressive

The 50,000-participant number comes from post-surgical pain. Back pain is a different question and it has its own trials, and they are far less flattering.

Acute low back pain. Cochrane pooled 32 trials, 5,356 participants. Against placebo, anti-inflammatories reduced pain by 7.29 points on a 0–100 scale (95% CI −10.98 to −3.61), moderate-quality evidence. Disability improved by 2.02 points on a 0–24 scale, high-quality evidence. The reviewers' own summary of that magnitude: "small and probably not clinically relevant." No difference between the newer type of anti-inflammatory and older ones. Almost half the studies were industry-funded.

Chronic low back pain. Cochrane, 13 trials. Against placebo, 3.30 points on the same 0–100 scale (95% CI −5.33 to −1.27), low-quality evidence. Disability, 0.85 points on a 0–24 scale. And this line, which deserves to be read twice: when the reviewers restricted the analysis to trials at low risk of bias, the difference between the drug and placebo got smaller.

Set those numbers beside the one intervention with a genuinely large effect for the same condition: structured exercise, pooled across 249 randomised trials, produced a 15.2-point pain reduction, which cleared the reviewers' pre-set threshold for a clinically important difference.

So the honest ranking for a back, on human trial evidence, is: exercise 15.2 points, anti-inflammatories 7.3 points for acute and 3.3 for chronic, BPC-157 untested. The drug beats the peptide because the peptide has nothing. It loses badly to walking and lifting.

What each one actually does to the tissue

An anti-inflammatory blocks an enzyme called the newer type of anti-inflammatory. the newer type of anti-inflammatory makes prostaglandins. Prostaglandins are what make an injury swell, throb and hurt. Block the enzyme, lose the prostaglandins, lose the pain. That is the whole drug.

The complication is that prostaglandins are not only a pain signal. They are also the opening instruction of tissue repair — the message that recruits repair cells to a damaged site and starts the rebuilding. One molecule, two jobs. The drug cannot separate them.

BPC-157 runs the opposite way. It is a short chain of fifteen amino acids based on a sequence found in human stomach juice, and in animals it raises VEGF, the signal that grows new blood vessels into damaged tissue, and drives connective-tissue cells to migrate into and repair an injury. It has no pain-blocking action of its own worth speaking of.

One suppresses the signal. The other is studied for accelerating what the signal is calling for. That is the actual axis of this comparison, and neither side of it is automatically the right answer — it depends entirely on what you are trying to buy.

The case against months of anti-inflammatories is specific, numbered, and not about healing

Most arguments against these drugs are vague and moralising. The real case is none of those things. It is four separate, measured harms, three of which are not in serious dispute.

The stomach

The largest analysis of individual patient data — 280 trials against placebo covering 124,513 participants, plus 474 head-to-head trials covering 229,296 — put numbers on it by drug. Upper stomach and gut complications, meaning a perforation, obstruction or bleed:

Note what that table does to the popular story. Naproxen, the one usually recommended as the heart-safe choice, has the worst gut number in the set. There is no drug in that column that is free.

The head-to-head randomised trial confirms the direction. PRECISION randomised 24,081 patients with arthritis and raised heart risk to celecoxib, ibuprofen or naproxen for a mean of 20 months. Stomach and gut problems were significantly lower with celecoxib than with either naproxen (p=0.01) or ibuprofen (p=0.002).

The heart

From the same 124,513-participant analysis:

Heart failure risk was roughly doubled by all of them. In absolute terms: for every 1,000 people taking a coxib or diclofenac for a year, three extra heart attacks and strokes, one of them fatal.

That is the honest trade-off between the two tables. Naproxen is the kindest to your heart and the harshest on your stomach. Celecoxib is the reverse. There is no free option, and PRECISION found celecoxib was not worse than the other two on heart outcomes at moderate doses, which softened but did not erase the coxib concern.

The kidneys

Forty studies covering 1,757,118 participants. Long-term use raised the odds of developing or worsening chronic kidney disease: pooled odds ratio 1.24, pooled hazard ratio 1.50. In people who already had kidney disease, the hazard ratio was 1.67.

In PRECISION, kidney events were significantly lower with celecoxib than with ibuprofen (p=0.004).

The report pile at the regulator, and exactly what it is not

FDA's public reported-harm database holds reports sent in voluntarily by patients, doctors and manufacturers. Queried on 4 August 2026, it holds 283,544 reports naming ibuprofen and 165,063 naming naproxen.

Say what those numbers are before reading them. They are counts of reports, not counts of people harmed. There is no denominator anywhere in the system — nobody knows how many people took either drug and reported nothing, and that missing number is the one an incidence rate would need. A report is not a finding that the drug caused the event. The same event can arrive more than once. Reporting rises when a drug is in the news. Nothing in this table is a rate, and it cannot be turned into one.

Statement

Official disclosure

openFDA adverse event reports naming ibuprofen and naproxen (FAERS, queried 2026-08-04)

“283,544 total reports name ibuprofen and 165,063 name naproxen. Most frequent reported terms for ibuprofen: DRUG INEFFECTIVE 27,383; PAIN 19,490; FATIGUE 18,754; NAUSEA 18,293. Serious event counts for ibuprofen / naproxen: acute kidney injury 7,141 / 3,145; renal failure 4,216 / 2,562; gastrointestinal haemorrhage 2,689 / 1,486; myocardial infarction 1,781 / 1,770; gastric ulcer 1,394 / 867; melaena 1,208 / 478; cardiac failure 634 / 305; gastric haemorrhage 622 / 489; duodenal ulcer 614 / 359. FAERS reports are voluntary and uncontrolled; they carry no denominator and cannot produce an incidence rate.”

Aug 4, 2026 · api.fda.gov

2026-08-04 19:45s42 · #ee2b10ba8a56

Two things are worth taking from it. The kidney rows sit above the gut rows for both drugs, which is the opposite of the ordering most people carry in their heads, and it lines up with the direction of the 1,757,118-participant kidney analysis rather than contradicting it. And the single most reported term for both drugs is not a harm at all: "drug ineffective" — 27,383 times for ibuprofen and 22,584 for naproxen. Set that beside the Cochrane back-pain numbers at the top of this page and it is the same finding arriving from the other end of the evidence pyramid.

The controlled trials above give you rates. This gives you the shape of what lands when something goes wrong. Use it for the second and never for the first.

Tendon and soft-tissue healing — the one that is genuinely contested

This is the argument most often made and least often made carefully. In animals it is consistent and specific. In a rotator cuff repair model, both an older anti-inflammatory (indomethacin) and a the newer type of anti-inflammatory one (celecoxib) produced significantly weaker tendon-to-bone healing than untreated controls, out to eight weeks.

The timing pattern is the strongest part of the case. Give ibuprofen during the early repair window and tendon healing is measurably worse; give the same drug later and the effect fades. That is exactly the pattern you would predict if the drug is interrupting the initial repair instruction rather than merely masking pain.

Bone shows it too. Pooled animal data show these drugs measurably reduce the mechanical strength of healing fractures.

The mechanism is not a coincidence. Knock out the newer type of anti-inflammatory in mice and fractures heal badly. Restore the downstream prostaglandin signal and the healing is rescued. The enzyme the drug exists to block is the same enzyme repair runs on.

It reaches human repair cells. In a human study, infusing an anti-inflammatory locally during exercise abolished the normal rise in muscle satellite cells — the stem cells that rebuild muscle after loading.

And now the counter-evidence, which is real. A study that pools other studies of human fracture outcomes that adjusted for confounders found no significant increase in fracture non-union among people taking these drugs.

There is even a species-level reason it might not translate. Human skeletal stem cells were directly compared with rodent ones and turned out to be substantially less dependent on the newer type of anti-inflammatory for their function.

So the fair verdict, stated as a rate rather than a mood: animal healing impairment is consistently reproduced; human cellular repair signals are measurably blunted; human clinical healing outcomes show no proven harm. Anyone claiming these drugs are proven to wreck human healing is overstating. Anyone claiming the concern is invented is ignoring the timing data and the knockout mice.

Why the argument sharpens when the injury is one the drug's own trials do badly on

A herniated disc is the one case where the healing question is not theoretical, because the mechanism that removes a herniation is an inflammatory mechanism.

Displaced disc material meets blood supply and immune cells for the first time. Immune cells digest it, new blood vessels grow in to supply the operation, and about 70% of herniations reabsorb on their own within roughly six months.

Suppress that inflammation continuously and you are working against the clearance. This is not purely inference: corticosteroids inhibited the bulge shrinking on its own in animal studies, and a clinical series that deliberately withheld anti-inflammatory drugs reported the bulge shrinking on its own in every patient.

That evidence is a case series, not a randomised trial, and it should be weighted as such. But it points the same direction as the tendon timing data, and it is the reason "a few days for a flare" and "eight months of daily tablets" are different decisions rather than different doses of the same decision.

The slower wear process behind many of these discs is a separate question with its own numbers, and it changes what a long course of anything is being asked to do:

What BPC-157 actually has, listed without inflation

In animals, the tendon results are the strongest thing it has. A transected rat Achilles tendon treated with BPC-157 came back with greater load-to-failure and better collagen organisation than controls.

The cellular behaviour matches. Tendon cells showed better outgrowth, survival and migration — the three behaviours that define active repair rather than scarring over.

And the mechanism is the mirror image of the drug's. BPC-157 raises VEGF and drives new blood vessel growth into healing muscle and tendon — the same vascular pathway that the newer type of anti-inflammatory blockade suppresses.

It protects against the drug's own gut damage, in rats. Given alongside diclofenac, BPC-157 counteracted the stomach and gut, liver and brain lesions the drug caused.

The same protection held when adjuvant arthritis and drug-induced gut lesions were put on the animal together.

What none of that is: a human result. Every item above is a rodent. There is no completed randomised controlled trial of BPC-157 in a human injury, no dose established for a human tendon, and no study of any kind against a human disc.

Seven people on long-term anti-inflammatories, counted

Seven first-person accounts from X, counted. Four of the seven reported a harm. One was warned off by a doctor before anything went wrong. One reported no problems over more than ten years. One is worried and getting tested, with no result back yet.

Read this panel knowing exactly what is wrong with it: people post about the ulcer that sent them to hospital, not about the uneventful decade. The harm fraction here is not the population rate — the population rate is in the tables above, and it is a 4× relative increase on a small absolute base, not four in seven. This panel tells you what the harm looks like when it lands, not how often it lands.

Reported a harm — 4 of 7.

Kidney filtration flagged on bloodwork after years of six tablets a day:

A stomach bleed that lasted a year, and twelve years of avoiding the whole drug class afterwards:

Five tablets a day for five years for back pain, and a stomach that can no longer tolerate the drug at all:

An ulcer, black stool, and an emergency room:

Warned off, no harm reported yet — 1 of 7. Nearly two years of naproxen with a stomach-protecting drug alongside it, then a new doctor capping it at six months:

No problems — 1 of 7. Over ten years of use, including long continuous stretches, and no reported issue:

Worried, no result yet — 1 of 7. Years of heavy use, now getting kidney and liver function tested:

Ten people on BPC-157, counted

Ten first-person accounts from X, all self-reported, none verified, none from a trial. Six said it helped. One said it helped and raised the possibility of placebo himself. Two said it did not work. One reported no outcome either way. Nobody reported a serious harm, and the only physical complaint in the set was at an injection site.

Two of the six positive accounts carry either a supplier link or a follow-me pitch attached to the post. That is not proof they are false, but it is a selection problem you should hold in mind while reading them.

Said it helped — 6 of 10.

A dried-out lower disc plus a year-old biceps tendon problem, six weeks of BPC-157 and TB-500 — tendon and shoulder resolved, disc improved but explicitly still needing physiotherapy:

Elbow tendon pain of months' standing, injected locally at the site:

A two-and-a-half-year knee injury the poster says nothing else had fixed:

Shoulder and joint pain on the BPC-157 plus TB-500 combination — note this post also links a supplier:

A rotator cuff for which surgery had been recommended, two eight-week cycles, surgery avoided — note this post ends in a follow-me pitch:

Two months, rotator cuff quiet during training, and heartburn gone as a side effect:

Helped, but he flagged placebo himself — 1 of 10. This is the most intellectually honest account in the set, and it is worth more than the enthusiastic ones:

Said it did not work — 2 of 10.

Ten years of on-and-off use and a flat verdict, including new tendon pain appearing while on it:

Eight weeks for a shoulder, judged not a success — with an interesting detail: stomach problems he had stopped noticing came back two weeks after stopping, which is the one place his account supports the gut-protective animal data even while rejecting the injury claim:

No outcome reported — 1 of 10. An injection-site problem from using too large a needle.

What ten accounts can and cannot establish. They cannot establish that it works — there is no control group, no blinding, no independent measurement, and a heavy bias toward people who bought the thing and want it to have worked. What they do establish, honestly: the reported experience is not uniformly positive, two of ten users describe no benefit at all, one of the enthusiasts volunteered placebo as an explanation unprompted, and nobody in this set reported a serious reported harm. Hold that against the 50,000-participant table at the top of this page and the size of the gap is the point.

What is legal, which is a separate question from what works

BPC-157 is not an approved drug anywhere. Its regulatory position has moved recently, so the specific dates matter more than the summary. FDA placed it in Category 2 of the 503A bulk substances list in September 2023, barring pharmacy compounding. It was off all three categories by the 14 May 2026 revision after the nominations were withdrawn. In July 2026 FDA's own briefing document recommended against adding it; on 23 July 2026 the Pharmacy Compounding Advisory Committee voted 8–6 with one abstention to recommend adding it anyway, against FDA staff. It is not on the 503B outsourcing-facility list either.

It is prohibited in sport at all times. WADA's 2026 Prohibited List names BPC-157 explicitly under S0, non-approved substances — in and out of competition, not only on competition day. The US Department of Defense lists it on its Prohibited Dietary Supplement Ingredients List and states plainly that it is not a dietary ingredient but an unapproved drug, at any route.

What it is sold as: research chemical, research use only. That is a shipping label, not a legal category granting you permission to use it.

Ibuprofen and naproxen are over-the-counter medicines with approved labels, approved dose ceilings, and pharmacist advice attached. That difference is not a footnote — it is the reason one of them has 50,000 participants of evidence behind it and the other does not.

The decision table

The one-line version of each side

For the drug: it demonstrably relieves pain in humans at scale, it is legal, it is cheap, and the healing concern is unproven in human outcomes. Its costs are measured, drug-specific, and rise with duration rather than dose alone.

For the peptide: it has a coherent repair mechanism, consistent animal tendon results, and animal evidence that it protects the gut lining the drug damages. It has no human trial for any injury, no established dose, no legal approval, and a user-report base with two in ten reporting nothing at all.

Those two paragraphs are not equivalent, and nothing above has tried to make them equivalent. One is a medicine with a known how big the difference was and a known bill. The other is an interesting hypothesis being taken by a lot of people ahead of its evidence.

What is compared above is the state of the evidence. It is not medical advice and not a dosing recommendation. Do not stop a prescribed medicine on the basis of a web page. BPC-157 is sold for research use only, is not approved for human use in any jurisdiction, and is prohibited in sport at all times.