A 41-year-old woman, 38 weeks and five days pregnant, came to labour and delivery with a headache, dizziness and a fetus that had stopped moving as much as it usually did. She had spent twelve hours next to a charcoal grill burning indoors for heat. Her carboxyhaemoglobin, drawn twelve hours after she left the room, was 7.4 per cent.
At most American hyperbaric centres that number does not open a chamber. Thresholds in common use sit at 25 per cent, or lower with syncope, acidosis or ischaemia; 7.4 per cent, half a day after exposure ended, is a value most clinicians would file under resolved. What changed the plan was not the mother’s blood. It was the fetal heart tracing, at 173 beats per minute with a late deceleration, and a biophysical profile of 2 out of 8. The obstetric team performed a caesarean section. The arterial cord gas came back with a carboxyhaemoglobin of 11.9 per cent and a pH of 7.05.
Both patients were then treated with 100 per cent oxygen at 2.4 atmospheres absolute for 90 minutes, beginning about two and a half hours after delivery; the infant was two hours old. After a single treatment the baby went to room air and was discharged three days later. The case, reported by Kreshak and colleagues at the University of California San Diego and the University of Utah in AJP Reports, illustrates a problem that runs the length of the hyperbaric literature: in a pregnant patient there are two circulations, and only one of them can be sampled.
Why the Fetal Compartment Behaves Differently
Foetal haemoglobin binds carbon monoxide more avidly than adult haemoglobin, and the fetus sits behind a placenta that is a barrier in one direction and a slow conduit in the other. Koren and colleagues put the consequence in a single sentence in a 1991 multicentre prospective study: “Because fetal accumulation of CO is higher and its elimination slower than in the maternal circulation, hyperbaric oxygen may decrease fetal hypoxia and improve outcome.”
That asymmetry explains the numbers in the San Diego case. The mother’s level had already had twelve hours to fall; the cord level, taken at delivery, was still 11.9 per cent. The two figures are not measurements of the same thing at the same time, and the gap between them is the entire clinical problem.
It also explains why the maternal carboxyhaemoglobin is such an unreliable trigger. The Undersea and Hyperbaric Medical Society’s fifteenth-edition chapter on carbon monoxide states that “the COHb level does not correlate with the development of neurological or cognitive sequelae”, and a 2015 review in Obstetrical and Gynecological Survey makes the same point for pregnancy, warning that “presenting carboxyhemoglobin levels may be poor indicators of severity of disease”. A threshold written around a value that does not track injury, in a compartment that is not the one at risk, is a rule of convenience, not a measurement.
Ozgok-Kangal’s 2021 series opens by citing a foetal mortality figure of 36 to 67 per cent for carbon monoxide poisoning in pregnancy, and the case literature contains outcomes that arrive late: a Tunisian infant diagnosed with hydranencephaly at 50 days of life after first-trimester exposure.
The Guidelines Agree on Treating and Disagree on Everything Else
The clearest statement of the maximalist position is Polish. In a 2025 position statement from the Section of Clinical Toxicology of the Polish Medical Society, Anand and colleagues write that “in pregnant women, HBOT is always indicated, irrespective of COHb concentration or clinical presentation”, and recommend extended normobaric oxygen before transfer “due to slower fetal CO elimination”, with the first session ideally inside six hours.
Italian toxicologists take the same destination by a different route. Writing in the Journal of Obstetrics and Gynaecology in 2022, Eleftheriou and colleagues from the poison centres of Bergamo, Pavia and Milan and four hyperbaric units describe hyperbaric oxygen in pregnancy as “proven to be safe” and “considered to be beneficial” — and then say the open question is precisely the one the Polish statement closes: which carboxyhaemoglobin level should trigger treatment. Their paper offers practical suggestions for units writing their own protocols. It does not offer a number.
A 2023 narrative minireview from Juntendo University is more careful about what the case reports show. Yanagawa and colleagues note that some treated women aborted, delivered prematurely or delivered infants with malformations, and argue these were sequelae of the poisoning rather than the treatment — presenting that as a hypothesis, not a finding. Arslan’s Konya series draws the boundary in the other direction: “HBO is not advisable for pregnant patients except for CO poisoning” — pregnancy is a reason to treat carbon monoxide, not a reason to enter a chamber for anything else.
Four documents, four countries, one shared conclusion — treat — and no shared threshold, because the value that would define one cannot be measured in the vessel that matters.
Four Cohorts, Forty Years, No Control Arm
The human evidence is a ladder of treated cohorts, and each rung has a gap in it.
Koren’s 1991 study was the first prospective multicentre attempt, collecting cases between December 1985 and March 1989 from faulty heating, car fumes and methylene chloride. Pregnancy outcome was adversely affected in three of five pregnancies with severe toxicity — two stillbirths and one child with cerebral palsy. All three adverse outcomes occurred in cases treated with high-flow oxygen, while the two severe cases with normal outcomes had received hyperbaric oxygen. That comparison rests on five patients: a signal worth following, nothing more.
The largest dataset comes from the Nord-Pas-de-Calais region of France. Wattel, Mathieu and Mathieu-Nolf followed 406 women and 412 children treated with hyperbaric oxygen between 1983 and 2008, using the children’s compulsory development assessments — 388 at day eight, 276 at two years, 232 at six — and matched them against anonymous local-authority records. They found no significant difference in psychomotor or height and weight criteria, and no malformations, and concluded that the findings support hyperbaric oxygen for all expectant mothers exposed to carbon monoxide.
The two Turkish series that followed are smaller and more recent. Arslan treated 32 pregnant patients at 2.4 ATA for 120 minutes, 23 of them with levels above 20 per cent and eleven with a history of syncope; three needed further sessions for fetal stress, there were no spontaneous abortions, four premature births, and two neonatal deaths from a cyanotic heart condition and necrotising enterocolitis. Ozgok-Kangal reviewed 28 pregnancies treated between 2013 and 2016, one of which had already lost the fetus before treatment began; three adverse events followed — an abortion, a premature birth and a limb malformation — and the remaining 24 delivered healthy term infants who were still free of diagnosed disease at a median age of 34 months. She treats the three events as likely incidental, on the basis of alternative causes in the mothers’ histories, and reports that trimester at exposure correlated with birth weight and that gestational week correlated with the age at which the children walked.
Read together, these are reassuring. Read strictly, they are all treated cohorts. Wattel’s comparison group is a set of unexposed children drawn from administrative files; nobody has followed untreated poisoned pregnancies alongside treated ones, and nobody has randomised a pregnant patient to oxygen alone. The question that matters clinically — whether the chamber adds anything to prolonged normobaric oxygen in this population — has never been put to a controlled test. The Polish statement’s “always” and the Italian letter’s “which level” are not two readings of the same dataset. They are two responses to the absence of one.
The Mouse Study That Kept the Question Open
The argument against unconditional treatment has always been biological rather than clinical, and it has a citation. In 1989, Van Hoesen and colleagues at Duke wrote in JAMA that “the dangers of hyperoxia to the fetus have been demonstrated in animal models” before arguing that the short exposures used clinically can be tolerated at any stage of pregnancy. Thirty-seven years later that sentence still contains the whole debate.
The animal work has since become more specific. A 2024 study in Developmental Dynamics from National Defense Medical Center in Taipei and Duke exposed pregnant mice to hyperbaric oxygen at the preimplantation stage and found blastocysts with raised reactive oxygen species and nuclear Nrf2, reduced Nf2 and Oct4 expression, raised DNA-damage and apoptosis markers, and ectopic nuclear signals for Cdx2, Yap and the Notch1 intracellular domain in the inner cell mass — induced apoptosis and impaired cell specification, in a paper framed around the underweight fetal body and overweight placenta seen after maternal hyperbaric exposure.
This is a mouse study at a gestational stage no clinician chooses. It is also, awkwardly, the stage at which exposure most often happens unobserved: before a woman knows she is pregnant, in a flat with a faulty boiler in the first cold week of winter. The human cohorts describe women who knew they were pregnant and reached a hospital. The animal signal describes a window those cohorts barely contain. That mismatch, rather than any disagreement about the data, is why the two positions have coexisted for four decades without either one losing ground.
What the Regulators Name, and What They Leave Out
In the United States, three documents define what hyperbaric oxygen is for, and all three name carbon monoxide while saying nothing about the patient carrying a second one.
The FDA’s consumer update, current as of July 2021 and now reachable only through an archived snapshot because the live page returns a “Page Not Found”, lists thirteen disorders for which hyperbaric chambers are cleared — among them carbon monoxide poisoning, decompression sickness, crush injury, gas gangrene and non-healing diabetic foot ulcers — and states that the agency has not cleared or authorised any device for conditions beyond that list. The word pregnancy does not appear.
Medicare’s National Coverage Determination 20.29, effective 3 April 2017, limits reimbursement to fifteen conditions and opens the list with “Acute carbon monoxide intoxication”. Across the full 14,437 characters of the determination, the string “pregnan” occurs zero times.
The Undersea and Hyperbaric Medical Society’s fifteenth-edition indications page lists seventeen numbered entries across fifteen indications, the second of which is carbon monoxide poisoning and its cyanide-complicated variant. Across the roughly 160,000 characters of text extracted from that page, “pregnan” also occurs zero times.
The FDA’s August 2025 letter to health care providers supplies the safety layer that applies to every patient in the building — fires with hyperbaric devices “that resulted in serious injuries and deaths”, a root cause the agency says “is not known”, requirements for training, grounding and monitoring for the duration of treatment, and compliance with NFPA 99-2024 and ASME PVHO-1-2023. It does not mention pregnancy either.
The consequence is uncomfortable but plain. The clearance, the coverage rule and the society’s indication all attach to the mother’s diagnosis; the fetus is treated inside a device cleared for somebody else, under a protocol assembled from physiology, four treated cohorts and a mouse experiment. That is how much of hyperbaric medicine’s difficult corners are handled, and exactly the kind of arrangement a wellness market misdescribes as settled.
The Trial That Has Not Been Registered
A query of the ClinicalTrials.gov registry combining hyperbaric, carbon monoxide and pregnancy returns no studies. Trials of hyperbaric oxygen in carbon monoxide poisoning do exist — a steroid-combination study and a study of cardiac toxicity among them — but the registry returns nothing for pregnancy as a defined population with fetal endpoints. For the most consequential single decision in hyperbaric medicine, it holds no randomised trial.
What such a trial would need is not exotic. Cord carboxyhaemoglobin at delivery, biophysical profile and umbilical artery Doppler through the third trimester, neonatal imaging, developmental assessment at two years — outcomes Wattel’s team already extracted from administrative records, collected prospectively instead. The comparator would be prolonged normobaric oxygen at 100 per cent, which the Polish statement already recommends for the reason the physiology gives, and enrolment would need stratification by trimester, because the mouse data and the human data do not describe the same weeks.
One caution belongs in that design. A 2025 case report in the American Journal of Emergency Medicine describes a neonate transferred to a tertiary centre with a raised carboxyhaemoglobin that proved physiological — the breakdown of foetal haemoglobin by haem oxygenase — and hyperbaric oxygen was appropriately deferred. In the compartment everyone argues about, even the assay needs a second look.
None of this argues against treating pregnant patients with carbon monoxide poisoning. The physiology points one way, the largest cohort points the same way, and the alternative to a chamber is a fetus clearing a poison on a timescale nobody has measured. It argues instead for precision about what kind of claim is being made. “Always indicated, irrespective of COHb” is a policy. It is a defensible one. But it is not a finding, it has never been tested against the alternative, and the second patient in the chamber has now been receiving it, on the strength of a mechanism and a case series, for nearly forty years.
Sources
- Kreshak AA, Lawrence SM, Ontiveros ST, Castellano T, VanHoesen KB. Perinatal Carbon Monoxide Poisoning: Treatment of a 2-Hour-Old Neonate with Hyperbaric Oxygen. AJP Rep. 2022;12(1):e113-e116.
- Koren G, Sharav T, Pastuszak A, et al. A multicenter, prospective study of fetal outcome following accidental carbon monoxide poisoning in pregnancy. Reprod Toxicol. 1991;5(5):397-403.
- Wattel F, Mathieu D, Mathieu-Nolf M. [A 25-year study (1983-2008) of children's health outcomes after hyperbaric oxygen therapy for carbon monoxide poisoning in utero]. Bull Acad Natl Med. 2013;197(3):677-94.
- Arslan A. Hyperbaric oxygen therapy in carbon monoxide poisoning in pregnancy: Maternal and fetal outcome. Am J Emerg Med. 2021;43:41-45.
- Ozgok-Kangal K. Long-term infant outcomes after hyperbaric oxygen treatment for acute carbon monoxide poisoning during pregnancy. Diving Hyperb Med. 2021;51(3):248-255.
- Anand J, Pawlas N, Schetz D, Kot J. Management of acute intoxication with carbon monoxide — Polish Medical Society, Section of Clinical Toxicology position statement. Int J Occup Med Environ Health. 2025;38(5):457-473.
- Van Hoesen KB, Camporesi EM, Moon RE, Hage ML, Piantadosi CA. Should hyperbaric oxygen be used to treat the pregnant patient for acute carbon monoxide poisoning? JAMA. 1989;261(7):1039-43.
- Li YM, Chung YL, Wu YF, Wang CK, Chen CM, Chen YH. Maternal exposure to hyperbaric oxygen at the preimplantation stages increases apoptosis and ectopic Cdx2 expression in mouse blastocysts. Dev Dyn. 2024;253(5):467-489.
- Friedman P, Guo XM, Stiller RJ, Laifer SA. Carbon Monoxide Exposure During Pregnancy. Obstet Gynecol Surv. 2015;70(11):705-12.
- Eleftheriou G, Butera R, Lonati D, et al. Open issues in management of carbon monoxide poisoning in pregnancy: practical suggestions. J Obstet Gynaecol. 2022;42(6):2540-2541.
- Yanagawa Y, Nunotani M, Abe K, et al. Narrative minireview of the current status of hyperbaric oxygen therapy for pregnant women. Acute Med Surg. 2023;10(1):e873.
- Hagemann J, Kozminski A, Miller S, Theiler CA, McCabe DJ. Normal physiologic carboxyhemoglobinemia of neonate masquerading as carbon monoxide poisoning. Am J Emerg Med. 2025;96:300.e1-300.e2.
- Undersea and Hyperbaric Medical Society — Hyperbaric Oxygen Therapy Indications, 15th edition (02a. Carbon Monoxide Poisoning)
- CMS — National Coverage Determination 20.29, Hyperbaric Oxygen Therapy (version effective 04/03/2017)
- FDA — Hyperbaric Oxygen Therapy: Get the Facts (consumer update, current as of 07/26/2021; archived snapshot)
- FDA — Follow Instructions for Safe Use of Hyperbaric Oxygen Therapy Devices: Letter to Health Care Providers, 25 August 2025