Knee Arthritis Is Where Hyperbaric Marketing Outruns the Evidence

Macro view of a translucent pale cartilage surface with a single dark fissure running across it, lit hard from the left and falling into darkness on the right

Osteoarthritis of the knee is the condition where hyperbaric oxygen therapy’s commercial language and its paperwork diverge most visibly. Three documents make the gap easy to measure. Medicare’s national coverage rule for hyperbaric oxygen, NCD 20.29, contains a section headed “Noncovered Conditions” that opens by stating no program payment may be made for any condition not listed in the covered section; among the entries is the phrase “Arthritic Diseases.” The Undersea and Hyperbaric Medical Society’s 15th-edition indication manual — seventeen numbered entries across fifteen indications, the specialty’s own statement of scope — does not contain the words osteoarthritis or knee anywhere in its text. And the Food and Drug Administration’s most recent communication on the devices, a provider letter dated 25 August 2025, is about fires, and reports “fires that occurred with HBOT devices that resulted in serious injuries and deaths.”

None of that has stopped the research pipeline. Since the middle of 2025 a steady trickle of knee papers has appeared: a rat study with a clinician’s title in Cartilage, a narrative review in Medical Gas Research, a mouse mechanism paper in the Journal of Orthopaedic Surgery and Research, one genuine randomized trial published in Scientific Reports, and a small Turkish retrospective that gets quoted more than the rest combined. Read together, they are a case study in how a plausible mechanism becomes a marketed indication while the human evidence stays exactly where it has been for thirty years: thin, surrogate-heavy and indirect.

A title that reads like a trial, and a methods section that says rats

The Cartilage paper is the one most likely to be screenshotted. Its title — “Hyperbaric oxygen therapy improves subchondral perfusion and cartilage integrity in established osteoarthritis” — reads like a clinical result, and the journal is a serious one. The design field is where the frame breaks: thirty-six Sprague-Dawley rats, anterior cruciate ligament transection to induce post-traumatic osteoarthritis, hyperbaric oxygen begun in week eleven, once the disease was established rather than the moment it was created, and continued for six weeks.

What the investigators at Tri-Service General Hospital in Taipei did with those animals is genuinely more careful than most preclinical hyperbaric work. They measured mechanical allodynia with von Frey filaments, imaged the joint on a 7-tesla scanner with T2* quantification and pharmacokinetic dynamic contrast-enhanced MRI, and then stained the tissue. Treated animals withdrew their limbs at higher forces than untreated ones, T2* values fell back toward normal, the clearance constants that DCE-MRI derives — kel and kep — recovered, and pathological blood volume fell. Histology showed lower OARSI scores and fewer alpha-smooth-muscle-actin-positive vessels crossing the tidemark, the boundary between cartilage and bone.

That is a coherent story about a specific compartment: the subchondral bone, where venous congestion and abnormal vessel growth are thought to feed cartilage loss. It is also, by design, a study of one model in one species in one laboratory, with a radiographic and histological endpoint rather than a patient-reported one. The authors’ own conclusion calls hyperbaric oxygen “a potent, non-invasive disease-modifying strategy.” Thirty-six rats cannot establish that, and the paper does not claim to have tried.

Two rat studies, the same operation, opposite verdicts

The mechanistic support is not uniform either, which is the part the reviews tend to skip. The Cartilage team’s reading is that hyperbaric oxygen normalizes pathological angiogenesis in subchondral bone. A separate mouse study from Northern Jiangsu People’s Hospital in Yangzhou reached a compatible conclusion by a different route: hyperbaric exposure raised PHD2, accelerated degradation of HIF-1alpha, suppressed VEGFA, and reduced the density of so-called H-type vessels in subchondral bone, with corresponding falls in MMP-13 and ADAMTS-5 and better preservation of Col2a1 and aggrecan.

Both papers therefore describe hyperbaric oxygen as anti-angiogenic in the osteoarthritic joint — a striking inversion of its usual role. In wound healing and radiation injury the therapy is used precisely because it promotes granulation tissue and new vessels; in the subchondral bone of an arthritic knee the same exposure is credited with shutting vessel growth down. Context dependence of that kind is biologically possible, but it is also the sort of claim that is easy to publish in one model and hard to confirm across several. A third paper makes the point bluntly. In 2021 a Turkish group ran nearly the same experiment — twenty-seven rats, ACLT in the right knee, a sham operation on the left, sacrificed at week ten — and compared untreated animals with a hyperbaric group breathing sixty minutes twice daily and an ozone group. Mean Mankin scores were 8.17 in controls, 6.22 with hyperbaric oxygen, and 4.72 with ozone. The ozone difference was significant. The hyperbaric difference was not: p = 0.086.

Three preclinical studies, two positive and one null, all using the same surgical model of the same disease. The protocols are not identical — the Cartilage group waited until week eleven to start and then treated for six weeks, while the 2021 animals were killed at week ten, so the two experiments were interrogating different phases of the same process — and the abstracts of neither report the pressure or the oxygen dose, which makes comparison harder than it should be. Whatever the explanation, the honest summary is that the animal literature has not converged. That is normal for a field at this stage. It is not a basis for telling a patient with a painful knee that the mechanism is settled.

The one randomized trial in knees is about the operation, not the arthritis

The strongest design in the cluster is also the least on-point, and the distinction matters. A team at West China Hospital in Chengdu randomized eighty patients undergoing total knee arthroplasty for primary osteoarthritis, forty to standard care plus hyperbaric oxygen and forty to standard care plus normobaric oxygen, then tracked markers of muscle damage — GOT, CK, LDH, myoglobin — along with inflammatory markers, limb circumference, quadriceps strength, range of motion and visual analogue pain scores at one, three and fourteen days. The trial is registered (ChiCTR2300072685).

The hyperbaric arm did better on several of those measures at day three: less rise in damage markers, faster return of quadriceps strength, less swelling, lower pain scores at rest and on movement at days two and three, better early range of motion, and no difference in adverse events. As evidence that a chamber helps a patient recover from a large elective operation, this is a real, randomized, comparatively unglamorous contribution.

It is not evidence about arthritis. Every participant had already reached the point of arthroplasty; the outcomes are postoperative and measured over two weeks; and the comparator was oxygen at normal pressure, which controls for the gas but not for the experience of lying in a pressurized vessel. Patients and assessors who know which arm they are in tend to report pain and swelling differently. Whether hyperbaric oxygen slows the loss of cartilage in a knee that has not yet been replaced — the question the Cartilage title implies — is untouched by this trial.

The retrospective that gets quoted for a disease it excluded

The human study most often cited in clinic brochures is smaller and older. In 2023 a team at Acibadem Kozyatagi Hospital in Istanbul reviewed patients aged 18 to 65 with bone marrow edema of the knee diagnosed on MRI, all treated with aspirin 100 mg daily, alendronate 70 mg weekly and instructions to avoid weight-bearing, and compared those who also received hyperbaric oxygen with those who did not. Healing was faster in the hyperbaric group, the authors reported, with no significant side effects.

Bone marrow edema is a real and painful condition, and it sits in the same bone compartment the rat studies were probing. But read the inclusion criteria: patients with osteoarthritis were excluded from that series, along with inflammatory rheumatic disease and malignancy. The most-quoted human knee result in the hyperbaric literature is therefore a retrospective, unblinded, single-centre comparison in people selected because they did not have arthritis, all of whom were also on a bisphosphonate. It supports a hypothesis about subchondral bone. It does not speak to osteoarthritis, and quoting it as though it does is a small, common and consequential sleight of hand.

Osteonecrosis is not osteoarthritis, and the rulebooks draw that line

There is a seam here that deserves naming, because it explains how the mismatch survives. The UHMS manual’s fifteenth indication is avascular necrosis, filed as “aseptic osteonecrosis,” and its pathology section describes reduced flow in “a terminal vascular bed, such as the one perfusing the femoral head, or similar vascular distributions such as femoral condyles.” Femoral condyles are the bottom of the femur — the top half of the knee. A reader in a hurry can get from that sentence to “the UHMS endorses hyperbaric oxygen for knee bone” without noticing that avascular necrosis is a focal death of bone from interrupted blood supply, while osteoarthritis is a whole-joint process in which the subchondral bone changes are one component among cartilage loss, synovitis, meniscal degeneration and mechanical load.

That difference is not academic, and the payment rules treat it as a difference. NCD 20.29 covers osteoradionecrosis and refractory osteomyelitis and crush injury, and separately declares that no payment may be made for arthritic diseases. Search the full text of that determination for the words osteoarthritis, avascular or osteonecrosis and the count is zero each; the only arthritis-adjacent string in the document is the exclusion. The same text contains no mention of knee. So an American clinic billing Medicare for hyperbaric sessions in a knee with osteoarthritis has no national coverage pathway, while a clinic treating a documented osteonecrotic lesion in the same joint is arguing from a different, and defensible, section of the same manual.

Around all of this sits the safety record that the field itself keeps trying to get clinicians to read. The FDA’s August 2025 letter does not question efficacy for the cleared uses; it notes that hyperbaric devices are Class II, cleared through the 510(k) process under product code CBF, and that the root cause of recent chamber fires is not yet known — then asks providers to review instructions for use, training, monitoring, cleaning, and to keep prohibited materials out of the chamber. The UHMS manual is blunter still about the commercial fringe: mild chambers, it notes, have “become widely available in so-called ‘wellness centers’ and health spas,” are “generally” not physician-prescribed or supervised, and are used for conditions “for which there is no scientific proof for any type of hyperbaric oxygen treatments,” with facilities that operate without fire-safety and chamber-construction standards “putting those exposed at risk for serious injury and even death.”

The osteoarthritis case is worth following for a reason beyond knees. It is a well-defined, extremely common, expensive condition with an acceptable animal rationale, a handful of human studies of the wrong design, and a marketing apparatus that has already arrived. That combination is where hyperbaric claims about longevity, athletic recovery and “cellular healing” will live or die over the next decade — not because the biology is absurd, but because the studies that would settle it have not been run. A sham-controlled trial in patients with Kellgren-Lawrence grade 2 or 3 disease, structural endpoints on MRI, and a year of follow-up would cost a fraction of what is spent annually on sessions that no payer covers. Until someone runs it, the reasonable position is the one the documents already take: a plausible mechanism, an unproven indication, and a knee that is not a femoral head.

Sources

← Back to blog