Same Papers, Two Verdicts: Duelling 2026 Reviews on HBOT and Breast Reconstruction

Dark anatomical diagram: an axial cross-section of a reconstructed breast with a thin skin-flap envelope over a smooth implant, a teal vascular pedicle branching from the upper left, and a rust-amber at-risk zone along the far arc of the flap, with two faint duplicate contours of the same slice offset at the right

On 30 October the journal Aesthetic Plastic Surgery published a short commentary on a meta-analysis it had carried in July, on hyperbaric oxygen for ischaemic complications after immediate breast reconstruction. Commentaries are the quiet part of the literature — a caveat, a caveat about the caveat, occasionally a correction. This one is worth noticing for a different reason: between the July review and the October commentary, the same journal published a second systematic review and meta-analysis on the same clinical question. It does not end up in the same place.

That is not a scandal. It is what a field looks like when the underlying evidence is thin and the analytical choices are consequential, and the pair together is more informative than either alone.

Two teams, two ways of adding up the same pile

The July review came from a group at the Tecnológico de Monterrey in Mexico City and the General Hospital of Mexico. It searched eight databases and registries through October 2025 and found five observational studies — four case series and one retrospective cohort — covering 409 patients, of whom 75 actually received oxygen. It then pooled the treated arms into single-arm estimates: flap salvage 82 percent (95% CI 72 to 90), pocket salvage 78 percent (65 to 88), reoperation 15 percent (9 to 23), explantation 8 percent (4 to 15), postoperative infection 3 percent (1 to 7). Treatment typically began inside the first ten postoperative days and ran 10 to 20 sessions at 2.0 to 2.5 atmospheres. The authors graded certainty low, assigned Level of Evidence IV, and concluded that oxygen “may be a safe and potentially useful adjunct,” pending prospective and randomised work.

The August review came from Hospital San Paolo in São Paulo with a co-author at Mohammed Bin Rashid University in Dubai. It screened 1,721 records down to seven studies and did two separate things: the same kind of single-arm pooling, and a comparison of oxygen against no oxygen. The pooled single-arm rates were seroma 16.1 percent, surgical-site infection 18.8 percent, return to theatre for debridement 12.6 percent and explantation 21.6 percent. The comparative odds ratios came out at 2.92 for haematoma (95% CI 0.47 to 18.06), 4.72 for skin necrosis (0.13 to 172.44) and 2.42 for reoperation (0.89 to 6.55) — none significant, at p = 0.25, 0.40 and 0.08. Its verdict: the comparative evidence is “insufficient and inconclusive”; oxygen “may be considered selectively within protocolized multidisciplinary pathways.”

Where the two numbers part company

Set the two single-arm results side by side and something obvious appears. Explantation is 8 percent in one review and 21.6 percent in the other. Return to the operating theatre for debridement is 12.6 percent in São Paulo; reoperation is 15 percent in Mexico City for what is nominally the same event. No arithmetic has gone wrong. Both papers are averaging proportions, and a pooled proportion inherits every decision upstream of it: which studies cleared the inclusion bar, which arm of a two-arm study was extracted, and how each series happened to define salvage.

The bibliographies make the point sharper. At least seven primary references appear in both reviews. These teams were not reading different worlds. They were reading substantially the same small pile of papers and applying different analytical lenses to it, and the headline percentage moved accordingly. Whatever 82 percent is, it is a property of a particular selection of five studies rather than a property of the therapy.

The odds ratios deserve their own pause. Printed above 1, they would ordinarily mean higher odds of the complication in the oxygen group, and the review does not take that up. With an interval running from 0.13 to 172.44 on skin necrosis, the sign is not meaningfully interpretable either way. An estimate that wide is not a finding; it is the shape of a dataset with almost no comparative information inside it. Which is, in more polite language, the São Paulo team’s conclusion. The commentary’s own argument sits behind the journal’s paywall, but its existence tells you the July numbers are still being argued over by people who read them closely.

The French series, and the trap hidden in a session count

In September, Bernard Najib and colleagues at the Centre Antoine Lacassagne in Nice published a retrospective series in the Journal of Gynecology Obstetrics and Human Reproduction: 39 women who had reconstructive breast surgery, carcinological or prophylactic, and then oxygen for scarring problems. Twenty-nine healed favourably (74.4 percent), ten did not (25.6 percent). Healing was independent of age, body-mass index and smoking. Half of the previously irradiated patients did badly. Nipple-areolar grafting made no difference. And favourable outcomes “seem to correlate” with a shorter interval between surgery and the first session, together with fewer sessions overall.

That last clause reads like an argument for doing less, and it is more likely an argument about which way the causation runs. A wound that closes in twelve sessions stops being prescribed sessions; a wound that fails for three months accumulates forty. Session count is downstream of healing, not upstream of it, and the Nice authors decline to push the point — no control group, therefore no causal inference, in their own words.

They are also pushing against an outcome that is not measured the same way twice. A 2022 systematic review from Duke, led by Nicholas Oleck, went through 59 studies covering 14,368 patients and 18,920 breasts and found the reported incidence of mastectomy skin flap necrosis running from 2.3 to 41.2 percent, with a pooled 10.4 percent per breast and 15.3 percent per patient. Studies defining necrosis by depth reported 15.1 percent; those defining it by area reported 6.3 percent, and the categories differed significantly. In a literature where the event itself is unstable, the denominator under every salvage percentage is a moving target.

The three-way comparison that did not flatter the chamber

The most awkward recent paper in this corner of surgery is not a meta-analysis at all. In June, a plastic surgery group at Kosin University Gospel Hospital in Busan published a propensity-matched three-way comparison in Cancers: 213 women with nipple-areolar necrosis after nipple-sparing mastectomy, of whom 57 received continuous topical oxygen, 59 hyperbaric oxygen and 97 standard wound care. After 1:1:1 matching on 15 covariates, 57 triplets remained. Time to complete epithelialisation was 30.5 days for topical oxygen, 36.4 for the chamber (p = 0.007) and 45.1 for standard care (p < 0.001). Topical oxygen emerged as an independent predictor of faster healing, hazard ratio 4.61, with the largest effect in wounds of 2 cm² or more.

Two details keep this from being the last word. The eight-week healing rates were 98.2, 96.5 and 87.7 percent — the chamber and the topical device arrive at nearly the same destination, and what separates them is speed. And it is one centre, retrospective, unblinded, with the therapy chosen by clinicians looking at the wound. That is precisely the design both meta-analyses had to work with, and precisely the design both said is not enough.

A case report in this quarter’s Undersea and Hyperbaric Medicine shows the other end of the spectrum. A 52-year-old woman developed a painful ulcer after bilateral reduction mammoplasty; it was treated as infection, debrided repeatedly, and got worse, which is what post-surgical pyoderma gangrenosum does when it is cut. Once histology and the clinical picture settled the diagnosis, she had corticosteroids, cyclosporine and infliximab, and healing still stalled for three months. Seventeen sessions at 2.4 atmospheres for 120 minutes, alongside atraumatic wound care, were followed by progressive granulation; complete epithelialisation came by postoperative week 38, with no recurrence at six months of follow-up. It is a case report. It shows what is possible, not what is typical — and cases like it are part of what lifts a pooled salvage rate.

What Medicare’s wording actually covers

Vocabulary matters here, and the two authorities do not use the same words. The Undersea and Hyperbaric Medical Society lists “compromised grafts and flaps” as indication 12 of its 15. Medicare’s national coverage determination 20.29 is narrower: it covers “preparation and preservation of compromised skin grafts (not for primary management of wounds).” The word flap does not appear anywhere in the determination’s live text, which we read directly rather than from a summary. Local contractors can and do decide differently, but the national benefit names grafts, excludes primary wound management, and attaches a condition to every use — no coverage for continued treatment without measurable healing inside any 30-day window.

So the common shorthand, that oxygen for a threatened breast reconstruction is “a covered indication,” is carrying more weight than the text supports. It sits in the right neighbourhood: the FDA clears these devices as Class II through the 510(k) pathway under Product Code CBF, and the specialty consensus list includes flaps. Consensus, coverage and proof are three different instruments, and only one of them is missing here.

The regulatory backdrop is also a reminder of how much remains unsettled. In its letter of 25 August 2025, the FDA told providers it was aware of fires involving hyperbaric devices that caused serious injuries and deaths, that the root cause is not known, and set out the fire-prevention basics — no prohibited electrical or static devices, cotton clothing rather than wool or synthetics, training, monitoring for the whole treatment. The same determination that covers grafts also carries an explicit non-covered list: cerebral vascular insufficiency, multiple sclerosis, arthritic disease, nonvascular causes of chronic brain syndrome including Alzheimer’s disease. That list is the closest thing in policy to an official statement that the chamber is not a general-purpose therapy, and it is why wellness and longevity marketing sits outside this discussion altogether.

What would actually settle it

Two things, and neither of them is a bigger chamber.

The first is a definition. Oleck’s review ended by asking for a consensus classification of mastectomy skin flap necrosis — depth, area, timing, whether an intervention was required — because until studies count the same event the same way, pooling them produces rates that swing threefold between two teams reading one literature.

The second is a comparator worth beating. The prevention side gives a hint of what that looks like. A 2023 meta-analysis from Monash Health covering 12 studies and 8,439 mastectomies found significant reductions in flap necrosis from topical nitroglycerin, closed-incision negative pressure wound therapy, topical dimethyl sulfoxide, oral cilostazol and local heat preconditioning. None of those is a chamber, and all of them are cheap. If oxygen is to be tested properly, it should be tested against an optimised bundle of that kind, in wounds stratified by a perfusion assessment — and even that diagnostic step is shakier than it looks, since the Cochrane review of indocyanine green angiography found only non-randomised cohorts and evidence of very low quality.

Which leaves a modest position, and it happens to be the one the São Paulo team wrote down: for a threatened reconstruction, oxygen is a reasonable adjunct inside a protocol somebody can audit, chosen for a wound that has not responded, and stopped when it is not working. What the literature cannot yet support is the number. Not 82 percent, and not zero.

Sources

← Back to blog