A 57 Percent Relative Drop in Deaths, From Thirteen Studies That Were Never Randomised

Data-art illustration of a dead-flat twenty-five-year mortality trend line across a dark navy panel, beside two stacked comparison columns showing a shorter oxygen-treated mortality bar against a taller control bar

Fournier’s gangrene announces itself badly. A patient — most often, though not exclusively, a man in his fifties — notices pain in the perineum or scrotum that does not match what the skin looks like. The infection is already moving along Colles’ fascia, and the skin above it stays deceptively intact while the tissue underneath dies. Diagnosis is missed for exactly that reason.

The mortality figures for this disease have been remarkably stable for a long time. When Radcliffe and Khan pooled 173 publications covering 6,152 cases treated between 1975 and 2018 for BJU International, they counted 1,220 deaths: 19.8 percent. They found no significant change in that rate across the quarter century (p = 0.996). Whatever has improved in critical care, antibiotics and sepsis protocols, roughly one patient in five still dies.

Which is why a paper published in Medical Gas Research deserves to be read slowly rather than quickly. Patel and Batura, working from urology at London North West University Healthcare NHS Trust, pooled 13 studies of adjunctive hyperbaric oxygen in Fournier’s gangrene. Their headline result: the chance of death in the oxygen arm was 0.43 times the chance in the control arm, with a 95 percent confidence interval of 0.21 to 0.86. A 57 percent relative reduction, clear of the null line, with moderate heterogeneity behind it (I² = 32 percent).

It is a genuinely interesting number, and one that requires an unusually careful reading of the methods.

What was actually pooled

The 13 studies were published between 2005 and 2022 and came from eight countries. Nine were retrospective comparative cohorts; four were case series. Together they contributed 322 patients who received oxygen and 366 who did not, at a mean age of 57 versus 56.

Grouped together, the mortality rates were 11.5 percent and 26.0 percent. The quality ceiling is set by the design: all 13 studies were retrospective, no randomised trial of hyperbaric oxygen in Fournier’s gangrene was found anywhere in the literature search, and the median score on the Newcastle-Ottawa scale was 6 out of 9 — “fair”, in the authors’ own word.

The detail that matters most sits in the methods rather than the abstract. The pooled relative risk of 0.43 was calculated from the nine comparative studies only. Four of the thirteen had nothing to compare against.

The four studies with no control arm

Those four were Rosa and Guerreiro’s 24 patients, Inácio and colleagues’ 23, Gülşen and colleagues’ 14, and Milanese and colleagues’ 6. Their mortality rates — 20.8 percent, 21.7 percent, 7.1 percent and zero — were set against a historical benchmark that Patel and Batura fixed at 25 percent on the basis of the published literature. That is a defensible way to put a number on a rare disease with a thin evidence base, and it is not a comparison. A published case series is not a random sample of the disease; units that write up their results tend to write up results they are pleased with. A benchmark fixed at 25 percent also carries no variance, so it cannot widen the confidence interval the way genuine disagreement between studies would.

The authors are straightforward about this, describing the manoeuvre in the methods and again in the limitations. It does not invalidate the review. It does mean the 11.5 against 26.0 summary is partly arithmetic rather than something anyone observed in 688 patients.

The patients who got the chamber were not the patients who did not

Here the article earns its place in the literature, because the review is unusually willing to point at its own weak spot.

In Feres and colleagues’ Brazilian series — 79 oxygen patients against 118 controls — mortality was 2.5 percent against 28.8 percent. That one study pulls the pooled estimate a very long way. Its oxygen group, the review notes, was younger (mean age 48 against 57) and carried substantially lower rates of diabetes and smoking. The comparison is between healthier men and sicker men, and no statistical adjustment in the paper repairs it.

Anheuser and colleagues’ multicentre German series makes the same point from the opposite direction. Their 17 oxygen patients were all septic and all treated in intensive care; among the 45 controls, 77.8 percent were septic and 69 percent reached intensive care. ICU stay ran 32 days against 9. Those 17 patients had a mean of 13 debridements against 5, and a hospital stay of 40 days against 22 — while their mortality was zero against 4.4 percent.

That single study explains why the review’s secondary finding — that the number of debridements “did not differ significantly” between groups — is less reassuring than it sounds. More operations in the oxygen arm is a marker of how sick those patients were, not evidence that the therapy fails to prevent operations. Patel and Batura say so themselves, then note that removing Anheuser’s cohort from the sensitivity analysis reverses the direction, producing a difference of 0.9 fewer debridements.

The series that went the other way

Mindrup, Kealey and Fallon, writing in the Journal of Urology in 2005, reviewed 42 patients treated between 1993 and 2002 and followed them for a median of 4.2 years. Mortality was 26.9 percent with oxygen against 12.5 percent without. Skin graft failure ran 8 percent against 6 percent. Average daily hospital charges were $3,384 against $2,552.

It is the only study of the thirteen in which oxygen looked worse, and it has the longest follow-up of any of them. The review’s explanation — that sicker patients were referred for the adjunct, and that a four-year window catches deaths a 30-day window misses — is plausible and unproven. It is an explanation, not a measurement.

The society that keeps it on the list

The Undersea and Hyperbaric Medical Society holds necrotizing soft tissue infections as an accepted indication, describing oxygen as “a recognized accepted adjunct to surgical debridements, antibiotic therapy and maximal goal-directed critical care therapy.” Its indications chapter sets out the case-series history on which that rests: Hollabaugh’s 26 patients across five Tennessee hospitals, with 7 percent mortality on oxygen against 42 percent without (p = .04); and 11 patients treated in Genoa with no deaths at all. The chapter also notes, carefully, that Fournier’s gangrene is usually studied and reported as its own group — which is exactly why a Fournier-specific meta-analysis was worth doing.

The same chapter also records the study that should temper all of it. Duke’s analysis of 50 consecutive patients over 15 years found a 20 percent mortality. On univariate analysis, three things predicted outcome: the extent of infection, the depth of infection, and treatment with hyperbaric oxygen. On multivariate regression, only the extent of infection remained an independent predictor. The society’s text puts the oxygen result at p = .06 — a signal that dissolves once you account for how much flesh the infection had already taken.

And then there is the sentence that tells you where the field has decided to stand: “With such strong case series evidence of reductions in morbidity and mortality for necrotizing fasciitis and the subset of Fournier’s Gangrene, it is difficult to envision ever seeing a controlled, double-blinded study of hyperbaric oxygen therapy.”

That is an odd argument to find in a guideline — the absence of a trial treated as an inevitability rather than a gap. Patel and Batura construct its mirror image from the other side, writing that “given the life-threatening nature of Fournier’s gangrene, prospective studies would be difficult, if not unethical, to carry out.” Both statements are sincere. Together they have left the evidence where it stood in 2005.

No two centres ran the same protocol

Sessions in the pooled studies ran from 30 to 180 minutes, at pressures from 2.4 to 3.0 atmospheres, for between 2 and 26 sessions. In several studies the protocol was not reported at all, and in others only in part. There is no international consensus on the role of oxygen in this disease, which the authors note with some understatement.

Severity scoring is the more consequential inconsistency. Some groups described their patients with the Fournier’s Gangrene Severity Index and others did not — and you cannot stratify by baseline severity across studies that never recorded it in a comparable way. The biggest confounder in the review is therefore the one thing the review cannot fix.

A scarce therapy, priced and rationed

The regulatory position is not ambiguous. Medicare’s national coverage determination 20.29 lists “progressive necrotizing infections (necrotizing fasciitis)” among the conditions for which it will pay, when the therapy is delivered in a chamber. That is a statement about coverage, not a finding of proven benefit; CMS does not require the second to authorise the first.

Access is the harder problem, and the review is blunt about it. In the United Kingdom, hyperbaric services are not routinely commissioned for any necrotizing soft tissue infection, Fournier’s gangrene included. The British Hyperbaric Association lists 10 registered facilities across the British Isles. Rosa and colleagues reported patients transferred from more than 100 kilometres away, which is a difficult thing to do to someone in septic shock. The authors put direct hospital costs above $35,000 for each course.

In a disease where delay to debridement is one of the few variables everyone agrees on, the adjunct that requires the transfer is being weighed against the transfer’s own risk.

What the FDA is worried about, which is something else

The Food and Drug Administration’s letter to health care providers of 25 August 2025 is not about indications. It is about the hardware. HBOT devices are Class II medical devices cleared through the 510(k) pathway under Product Code CBF, and the agency wrote that it was aware of “reports of fires that occurred with HBOT devices that resulted in serious injuries and deaths,” adding that the root cause in those cases is not known.

The letter asks for adherence to each device’s instructions for use, fire prevention measures, staff training, monitoring for the duration of treatment, cleaning and maintenance at the specified intervals, and no prohibited electrical or static-producing devices in or near the chamber — 100 percent cotton clothing, no wool or synthetics.

The warning lands twice in this context. The pressures used in the studies above, 2.4 to 3.0 atmospheres, are clinical-chamber pressures delivered with air breaks by staff trained to watch for oxygen toxicity. Soft-shell devices sold for home and wellness use run at 1.3 to 1.5 ATA and cannot reach them; nothing in this literature describes what they do, because nothing in this literature was done with them. The FDA has separately and for years listed conditions — cancer, autism, Alzheimer’s disease, multiple sclerosis and others — for which it says hyperbaric oxygen has not been shown to work, and which are marketed anyway.

What would change the answer

Not another case series. A prospective registry, with a severity index recorded at baseline for every patient, a protocol written down before the first session rather than reconstructed afterwards, and enough centres to make comparisons within strata of illness instead of across them.

The review’s own open questions are the right ones, and they are all still open: which pressure, which duration, how many sessions, and which patients — the ones scoring 4 on the severity index or the ones scoring 12. It also asks what the therapy costs a health system that has ten chambers for the whole of the British Isles.

None of that is an argument for withholding oxygen from a patient with necrotizing fasciitis who is in the right hospital at the right time, on a service that can deliver it safely and monitor for its complications. It is an argument about what the number 0.43 means. It means that across thirteen retrospective datasets, patients who went into a chamber died less often than patients who did not. It does not mean the chamber is the reason. Until someone builds the registry that can separate those two statements, the field will keep quoting a 57 percent reduction it cannot explain — and one patient in five will keep dying of a disease that has not changed its mind in twenty-five years.

Sources

← Back to blog