Pentosan polysulfate (PPS)

Pentosan polysulfate is technically not a peptide — it is a semi-synthetic sulfated polysaccharide made by sulfating beech-wood xylan, heparin-like in structure and behavior. It sits in this section because the peptide community treats it as one of theirs. It is also the least exotic compound here: a licensed human medicine in the US and Europe, a licensed veterinary joint injectable in half a dozen countries, and the subject of a pivotal phase 3 trial. What it is not licensed for anywhere is human osteoarthritis.

Mechanism

The mechanism case is better than most things on this site, and it has four legs of unequal strength.

The strongest is catabolism inhibition, and it is unusually well resolved at the molecular level. PPS blocks the aggrecanases ADAMTS-4 and ADAMTS-5 — the enzymes that execute early cartilage breakdown — at concentrations in the tens of nanomolar, binding their non-catalytic ancillary domains rather than their active sites. In human osteoarthritic chondrocytes it shuts down ADAMTS-4 activity without changing how much enzyme the cells make. It simultaneously raises cartilage and synovium levels of TIMP-3, the body's own inhibitor of those enzymes, by blocking the receptor that clears it and by increasing how much of the protein gets translated — an effect that is selective, leaving seven matrix metalloproteinases and the other three TIMPs alone. And it physically bridges enzyme and inhibitor into a single three-part complex, raising their mutual affinity more than a hundredfold. That bridging is electrostatic and depends on chain length: molecules of eleven or more sugar units work about a hundred times better than eight-unit ones.

Away from purified enzymes, whole cartilage behaves as the enzymology predicts. In bovine explants, PPS slows aggrecan breakdown dose-dependently and preserves tissue glycosaminoglycan, with the effect largest when catabolism has been deliberately provoked — and without suppressing chondrocyte metabolism to get there. That study has no author tie to the drug's development, which is part of why it is the version worth quoting.

The second leg is anabolic and anti-inflammatory. PPS raises the amount and the molecular weight of hyaluronan in inflamed tissue, the two properties osteoarthritic synovial fluid loses together. It suppresses the interleukin-1-driven transcription of inducible nitric oxide synthase, c-Jun and HIF-1α in chondrocytes — the route to MMP-13 and nitric-oxide-mediated matrix loss. In culture it pushes human bone-marrow stem cells and dedifferentiated chondrocytes back toward cartilage-producing behavior with type II collagen and aggrecan, more strongly at higher molecular weight. In rabbits it prevented steroid-induced loss of cartilage proteoglycan and hyaluronan.

The third leg is heparinoid systemic biology. PPS stimulates the vessel lining to release clot-dissolving tissue plasminogen activator, lowers its inhibitor, and mobilizes fat-clearing enzymes; the proposed route to the joint is improved blood flow through the small vessels of subchondral bone, whose health cartilage depends on. This is the leg with the least direct evidence in actual joints, and the trial designed to test it — the academic MaRVeL study, which enrols only people with dyslipidaemia — has not reported.

The fourth leg is about pain rather than tissue. Osteocytes in arthritic subchondral bone make and secrete nerve growth factor, the target of the anti-NGF analgesics, and PPS reduces both its cytokine-driven secretion and its messenger-RNA induction in human cells taken at knee replacement. If PPS relieves symptoms without changing structure, this is the most likely reason.

Cutting across all of it: the scientist who assembled the disease-modification case also developed PPS for arthritis and held commercial interests in it, and his name is on much of the mechanism literature. The aggrecanase work at Imperial College, the explant work at La Trobe, and the canine chondrocyte work in Hokkaido are the closest the mechanism gets to neutral ground.

Route and pharmacokinetics

This is the part a reader taking an injectable course most needs, and it is the part most write-ups skip.

Oral PPS is, systemically, almost not a drug. In a three-way crossover in eighteen healthy volunteers, fifty milligrams intravenously raised clotting times, anti-Xa activity and lipase activity exactly as an injected heparinoid should; fifteen hundred milligrams by mouth — thirty times the dose — moved none of them, and the point estimate for oral bioavailability came out around zero with narrow confidence intervals. A radiolabelled mass-balance study filled in where the dose goes: about eighty-four percent leaves in the faeces as intact PPS, and roughly six percent appears in urine, largely as smaller and desulfated fragments.

That is not a criticism of the oral product. Interstitial cystitis is treated at the bladder surface, and urinary excretion is how an oral dose gets there. But it means the licensed oral drug and an injected course are close to different medicines, and that inferences run poorly between them in both directions — which is why the absence of a warfarin interaction with oral PPS, demonstrated in a dedicated crossover study, cannot be carried across to injection.

Injection delivers the systemic drug. Where the two routes have been compared head to head at matched doses in animals, injection wins on delivery: in mucopolysaccharidosis rats, once-weekly subcutaneous dosing matched or beat daily oral dosing on every shared endpoint and reached tissues the oral route did not, with the optimum at two milligrams per kilogram — the same per-kilogram dose the human injection trials use. In a dog model the same comparison found effects "most evident" after subcutaneous dosing, and only the injected route lowered inflammatory markers in cerebrospinal fluid. The one measurement of how much injected drug reaches a joint is equine: after a single intramuscular dose, synovial-fluid concentrations landed in the range that acts on synoviocytes and cartilage in culture.

Human injection courses in trials have run at three milligrams per kilogram intramuscularly weekly for four weeks, or two milligrams per kilogram subcutaneously once or twice weekly for six weeks. For an eighty-kilogram adult that is roughly two hundred and forty milligrams a shot and about one and a half grams per course — a figure worth holding onto for the safety section.

The licensed indication

Pentosan polysulfate has been an approved human medicine since 1996, sold as Elmiron for interstitial cystitis and bladder pain syndrome, and approved in Europe for the same indication more recently. That matters here for two reasons that have nothing to do with bladders.

The first is exposure. Tens of thousands of people have taken this molecule daily for years — the defining ocular case series found a median of fifteen years of use — which is why PPS has a mature human safety file at all, and why its rarest adverse effects have been characterized well enough to be dose-mapped. Almost nothing else in this section has been given to humans at scale for long enough to know what it does over a decade.

The second is that the efficacy evidence behind the license is routinely invoked as though it settled something, and it is genuinely split. The two trials the US approval rests on were positive by modest margins: overall improvement above twenty-five percent in twenty-eight percent of patients versus thirteen percent on placebo in one, and significant global improvement in thirty-two percent versus sixteen percent in the other. A 1997 meta-analysis of four trials found absolute benefits over placebo of thirteen to seventeen percentage points for pain, urgency and frequency, with numbers needed to treat of six to seven and half. Against that: the first multicentre placebo-controlled trial, in 1987, found no difference from placebo on symptoms, urodynamics, cystoscopy or mast-cell counts. A federally funded factorial trial in 2003 found a global response of thirty-four percent with PPS against eighteen percent without, short of significance, and concluded the drug did not help most patients. A dose-ranging trial in three hundred and eighty patients found tripling the dose changed nothing. And the largest and most modern trial — three hundred and sixty-eight patients, placebo-controlled, stopped early at an interim analysis — found responder rates of 40.7 percent on placebo, 39.8 percent on one hundred milligrams daily and 42.6 percent on the labelled three hundred.

The two most recent syntheses disagree along funding lines. A 2019 systematic review prepared for the European approval dossier, with an author from the manufacturer, found significant benefit for overall response, pain and urgency. A Cochrane network meta-analysis the following year, covering eighty-one trials and 4,674 patients, found no evidence that PPS improved cure, pain, frequency or nocturia, at low to very low certainty throughout.

So what the licensed indication establishes is real but specific: decades of human pharmacovigilance, a well-characterized adverse-effect profile, and a regulatory history that makes injectable PPS an off-label use of a known medicine rather than a research chemical. What it does not establish is that PPS works — for the bladder or for anything else. And because oral dosing is essentially unabsorbed, the bladder trials are not evidence about systemic PPS at all.

The human joint evidence

A randomized, double-blind, placebo-controlled pilot in one hundred fourteen people with knee osteoarthritis tested intramuscular PPS at three milligrams per kilogram weekly for four weeks against a Ringer's-solution injection. Three of seven direct clinical assessments separated: the duration of early-morning stiffness through week sixteen, pain at rest from week eight onward, and patient global assessment at every visit after the fourth injection. The aggregated function score separated at weeks eight and twelve.

Two details set the size of that. Pain at rest was measured in the thirty-six patients whose baseline score reached two centimetres, and the difference was half a centimetre on a ten-centimetre scale. Pain on walking did not differ at any timepoint on the absolute scale, separating only on the proportion of patients improving by two centimetres or more. The paper is careful about that everywhere except its abstract, which asserts that pain on walking improved and names no measure; the discussion restates it as a frequency of response, and the closing conclusion leaves it out. Attrition was heavy and uneven: twenty-four of fifty-four on PPS and forty of sixty on control left the study, so the late visits rest on about thirty people and about twenty. Randomization also gave the treated arm the shorter symptom duration.

Two decades later it remains the only completed symptom-endpoint RCT in osteoarthritis — never independently replicated, and part-sponsored by the company that prepared both the drug and the control vials.

There is a second placebo-controlled injection trial with a joint endpoint, in a different disease. Twenty people with persistent arthralgia after Ross River virus infection were randomized two-to-one to subcutaneous PPS at two milligrams per kilogram twice weekly for six weeks. Its primary endpoint was safety; every efficacy measure was secondary, and the protocol planned no formal test of treatment effect. Grip strength in the dominant hand improved almost seven kilograms more than placebo at day fifteen — one hand, one timepoint, and placebo grip had dipped below its own baseline at that visit before rising past it later. RAPID3 pain and total scores separated at day fifteen; RAPID3 function and global estimate did not, and the trial's dedicated pain scale showed no difference at any timepoint. Serum COMP and urinary CTX-II fell further than placebo, two of six markers reaching significance out of forty-two measured, one of which moved the other way. Twenty participants, a sponsor-funded trial, and a viral arthritis rather than osteoarthritis — the same regimen the OA program uses, tested in a disease a cartilage tier cannot read.

A Japanese team ran the only other injectable trial in osteoarthritis: twenty women with mild knee OA, six weekly subcutaneous injections at two milligrams per kilogram, open-label, no control group. Effusions settled quickly, pain scores roughly halved and stayed improved for a year, and a blood marker of collagen breakdown fell about twenty percent. With no placebo arm, none of that is attributable to the drug — OA symptom trials routinely show placebo responses of this size — but the tolerability data are real, including measurably prolonged clotting times in the hour after each injection.

The modern program belongs to Paradigm Biopharma. Its phase 2 biomarker RCT in sixty-one people set as its primary endpoint a change at day fifty-six in one or more of eleven synovial-fluid biomarkers, with no correction for multiplicity. Synovial-fluid ARGS-aggrecan — a cartilage-degradation marker — rose in both arms and rose less on PPS, an adjusted twenty-three percent against seventy-nine on placebo, while serum CTX-I rose further on PPS than on placebo. At the later visit the ARGS gap held and serum C2C fell; the other six markers moved without reaching significance. The trial measured symptoms too: the pooled PPS arms did not separate from placebo on WOMAC pain, the twice-weekly arm did, and the once-weekly arm did not. The authors call the statistics hypothesis-generating rather than confirming, and sponsor employees are among them. A single-arm oral pilot in thirty-eight people reported pain reduction, uninterpretable for the same no-control reason.

The first academic trial is now registered and its protocol published: MaRVeL, a ninety-two-person placebo-controlled study at Royal North Shore Hospital in Sydney, testing two five-week cycles of oral PPS in people with knee OA and dyslipidaemia, with sixteen-week knee pain as the primary outcome. It is the first knee trial of this drug that Paradigm does not run, and the first to test the vascular hypothesis directly by selecting for it.

Symptoms, biomarkers, structure

PPS has been called a disease-modifying osteoarthritis drug since the 1990s. Here is what has actually been measured, tallied across every trial in this entry.

No controlled human trial has yet reported joint structure. The 2005 RCT measured symptoms. The phase 2 measured synovial-fluid biomarkers, and it did scan: MRI at baseline and at the later visit, listed as exploratory and described in the paper as reporting separately, which it has not yet done. The running phase 3's primary endpoint is WOMAC pain. The only published human structural data are uncontrolled: the twenty-patient open trial took radiographs at one year and found them unchanged in both directions, and a single sponsor-linked case report described MRI resolution of a bone marrow edema lesion.

The controlled structural evidence that does exist is veterinary and mixed. In eighteen horses with surgically induced joint damage, intramuscular PPS reduced cartilage surface fraying and raised a marker of cartilage synthesis versus saline. In an early canine cruciate-transection study, intramuscular PPS lowered histological damage scores and cartilage metalloproteinase activity, though only the arm combining PPS with an injected growth factor approached normal tissue. Against those: a second equine trial of an intravenous PPS combination improved macroscopic joint scores but not histology, histochemistry or cartilage biochemistry, and the authors said so; a year-long double-blind dog trial of oral PPS found no difference from placebo in radiographic progression; and the canine cruciate-surgery trial found none either.

So the disease-modification question is not answered no — it is unasked in humans with a control group, forty years into the drug's use in joints. A trial designed to measure structure is what would answer it, and none has been run.

The veterinary record

PPS has been a licensed veterinary injectable (Cartrophen Vet) for decades, which is usually presented as its deepest credential. The record deserves a closer look than its reputation gets.

The trial that set the standard dose — forty dogs, double-blind, four arms — reported better scores than placebo for lameness, pain on joint handling and willingness to exercise, with the middle dose of three milligrams per kilogram working best and the highest dose working least. Those were subjective clinical scores, and the drug's developer was a co-author. The most rigorous canine trial, in forty dogs after cruciate surgery, was null on its primary analyses — no difference in lameness, force-plate measures or radiographic progression — with benefits confined to subgroup analyses. The oral-PPS dog trial was null on function and structure alike over a full year. An independent 2025 review of the randomized canine evidence concluded exactly this: neither placebo-controlled PPS trial hit its primary endpoint. The equine practice data are veterinarian-perception surveys. The one recent dog trial read as positive — sponsor-run, fourteen treated dogs against six controls — reports pain, gait and cartilage-volume changes at twenty-six weeks that never reached significance between the groups; the paper says as much and calls its own interpretation descriptive.

What decades of veterinary use do establish is tolerability at scale: a nine-year review of UK adverse-reaction reports estimated real adverse-reaction incidence around seven in ten thousand doses even after assuming ninety percent underreporting, mostly brief vomiting or a day or two of lethargy. What veterinary use cannot establish is that PPS works — in dogs, on primary endpoints, it mostly has not been shown to, and dogs were supposed to be the strong case.

The veterinary literature does contribute two things the human literature has no equivalent of. One is a dose-response for the anticoagulant effect: in horses given nothing, three, six and ten milligrams per kilogram, partial thromboplastin time rose in proportion to dose and stayed above baseline for up to twenty-four hours at the clinical dose. The other is a set of concrete bleeding events, which belong in the safety section.

What to watch

Two registered trials matter, and one has gone quiet. PARA_OA_002 — the phase 2/3 adaptive trial with six hundred two participants — has produced positive company press releases but no peer-reviewed symptom results, and its registry record has not been updated since April 2024, with estimated completion dates now long past. Its observational extension, which followed a hundred and thirty-three participants to answer how long a response lasts, completed in December 2024 and has posted no results either. The decisive trial is the pivotal phase 3, NCT06917404: as of the registry's May 2026 update it is listed as recruiting, with an estimated four hundred sixty-six participants and primary completion projected for July 2027. The sponsor has guided to an earlier top-line; the registry's own dates are the verifiable ones. Biomarkers like ARGS-aggrecan are surrogate outcomes; the phase 3's WOMAC-pain data are the ones that count. The MaRVeL trial's readout will be the first from a group that does not sell the drug, though a second PPS company supplies it.

Safety: the eye

The signature issue is ocular, and the last two years have changed both its shape and its standing.

A distinctive pigmentary maculopathy was first described in 2018 in long-term users of oral PPS, and is now carried in US labeling. The defining case series found a median of fifteen years of daily oral use and a median cumulative dose of about one point six kilograms. A nationwide Korean cohort of over one hundred thirty thousand PPS users showed the risk is dose-dependent — hazard ratios of roughly one and a half beyond fifty grams cumulative, with maculopathy incidence rising across the bands from four to six to nine percent. The highest band is four hundred and forty-five people, three in every thousand of that cohort, so the gradient is firmer than the top figure. The same study found high-exposure users were no more likely to have had their eyes examined than low-exposure ones, leaving nearly half of the people it classes as high-risk unscreened. Twenty-one years of US adverse-event reports put eye disorders at the top of both the volume and the signal strength, with a median time to onset of about four and a half years and sixty-eight percent of reports classed as serious.

Whether the drug causes it was, until recently, genuinely contested: a manufacturer-linked claims study found near-background incidence and argued causation was unsettled. The 2025–2026 literature has largely answered that. A single-centre study compared interstitial cystitis patients who had taken PPS with cystitis patients who had not: of a hundred and twenty-two patients imaged, all eight cases of pigmentary maculopathy had taken PPS and none of the unexposed had — which removes the disease itself as the explanation. A comprehensive review in Progress in Retinal and Eye Research, whose authors declare no conflicts, walks the causal criteria in order — strength of association, dose-response, absence of confounding by indication, biological plausibility from laboratory evidence of direct toxicity to the retinal pigment epithelium — and concludes for causation. The remaining dispute is about magnitude and incidence, which is a real and unresolved question; it is no longer about whether the effect exists.

Dose-response is now mapped in some detail. In a screened cohort of a hundred and eighty-seven exposed patients, the top exposure quartile at two kilograms or more carried an odds ratio near forty, and prevalence across the whole screened group ran between about four and twenty-five percent. Cumulative dose dominates, but dose per kilogram of body weight, older age, female sex and inflammatory bowel disease came out as independent factors — and among patients below eight hundred grams cumulative, those who developed maculopathy were lighter, older and had taken more drug per kilogram.

Then the finding that concerns injection users directly. In 2026 the first three cases were published of the same maculopathy after subcutaneous PPS for arthritis, at cumulative doses of forty-five to ninety-six grams over seven to ten years — a fraction of the oral exposures, which the authors attribute to roughly tenfold higher bioavailability by injection.

The arithmetic runs both ways, and both directions are worth stating. Taken at face value, a once-yearly course of six injections delivers roughly one and a half grams, so the lowest affected injection case had accumulated about thirty years' worth of such courses in under a decade — those patients were injecting continuously, not annually. But convert to absorbed drug and the two exposure figures stop looking different: a 1.6-kilogram oral course at around six percent absorption is roughly a hundred grams systemically, and forty-five to ninety-six grams injected at high bioavailability is the same order of magnitude. The oral and injectable cases are not two phenomena; they are one phenomenon reached by two routes, and injection is the efficient route. No exposure floor for safety has been established, the damage appears irreversible and can progress after stopping, and the reassuring industry data cover no injectable patients at all. Anyone running repeated courses has a concrete reason to get baseline retinal imaging.

Safety: blood

The second issue is the heparin family resemblance, and here injection is the route that carries the risk.

PPS prolongs clotting measures for hours after injection, dose-dependently. In the only randomized human trial of injected PPS against a standard anticoagulant — women receiving thromboprophylaxis after major gynaecological surgery — minor bleeding occurred in twenty-eight percent on subcutaneous PPS versus five percent on enoxaparin or fondaparinux, with no major bleeds in either arm. That is twice-daily dosing in the highest-risk bleeding window there is, not an arthritis course, but it is the clearest read available on what injected PPS does to haemostasis in people. The veterinary literature supplies the concrete events: four dogs bled from arthroscopy portals one to three hours after joint-injected PPS, with deranged coagulation parameters that resolved together with the bleeding, and one cat with undiagnosed heart disease collapsed and died after a single subcutaneous injection with severe subcutaneous and intermuscular haemorrhage found at post-mortem. Both involved off-label routes or doses; both show the drug reaching the whole animal.

PPS can also — rarely — trigger the same immune platelet reaction as heparin, which presents not as bleeding but as clotting: at least one published case produced a stroke, and 2026 laboratory work confirmed PPS forms the same kind of antigenic platelet-factor-4 complexes that cause heparin-induced thrombocytopenia, with binding strength between low-molecular-weight and unfractionated heparin and rising with molecular weight. Unexplained bruising or a platelet drop during a course is a stop-and-test event, not something to push through, and the manufacturer-linked authors of that work suggest platelet monitoring for parenteral use themselves.

The interaction picture is asymmetric in a way worth knowing. A dedicated crossover study found that oral PPS at the licensed dose left warfarin's pharmacokinetics and INR untouched — but oral PPS is the route that is barely absorbed, so that reassurance does not transfer to injection. Bleeding-risk caution and anticoagulant-interaction potential apply to the injected drug on its own pharmacology.

Course-level tolerability in trials has otherwise been acceptable in the sense that matters: no serious drug-related events in the phase 2, no case of heparin-induced thrombocytopenia, and a benign veterinary pharmacovigilance record. It has not been uneventful. Injection-site reactions ran at thirty-five and sixty-eight percent across the phase 2's two PPS arms against under five percent on placebo, six of its treated participants stopped the drug for adverse events where none on placebo did, and in the Ross River trial liver enzymes rose on PPS to nearly three times the upper limit of normal before settling back. PPS is not on the WADA Prohibited List.

What is in the vial

The regulatory geography explains most of the confusion around this drug. Oral PPS is approved for interstitial cystitis, not osteoarthritis. An injectable form has existed in Europe for decades from its earlier life as an antithrombotic. The veterinary injectable is a licensed product in the UK, Australia and elsewhere — for dogs. No injectable PPS is approved for human osteoarthritis anywhere; human arthritis courses run on off-label, compounded or special-access prescribing, which is exactly the population in which the 2026 maculopathy cases surfaced.

Product identity is not a formality here. PPS is a mixture of sulfated glucuronoxylans that physicochemical analysis cannot fully specify, and when ten commercial preparations from different manufacturers were compared with orthogonal methods and functional assays, they differed measurably in both structure and biological activity — enough that the authors, who work for two of the interested companies, concluded follow-on products are at best similar to the original rather than identical. That finding has teeth because the mechanism itself is size-dependent: the enzyme-bridging effect needs chains above a certain length, the anabolic effect on chondrocytes strengthens with molecular weight, and so does platelet-factor-4 binding. Gray-market "PPS peptide" vials are not the licensed human product, not the veterinary product, and not characterized at all.

Sponsorship, and what stands against it

Nearly all modern PPS-for-osteoarthritis data are Paradigm-sponsored, with employees and shareholders among the authors; the strongest independent evidence is a 2005 pilot that was never confirmed; and the veterinary record, read at the level of primary endpoints, is mostly null. Against that stand a real placebo-controlled symptom result, a placebo-controlled biomarker result its own authors call hypothesis-generating, a second randomized injection trial reporting joint symptoms, and a mechanism whose enzymology is better resolved than anything else in this section. The tier here has a scheduled examination date, which is rare: the phase 3's WOMAC-pain readout, due within roughly a year of this writing, is the named event that moves this entry up or down, with the first academic knee trial following behind it.