If you want to know whether a bovine hemoglobin oxygen carrier can sit in a bag on a shelf, be hung on a pole, and be given to a sick animal by a tired person at two in the morning, you do not have to guess. It has been happening for a quarter century. The product is called Oxyglobin, it is a polymerized bovine hemoglobin solution, and it was approved for dogs in the United States in the late 1990s, before any comparable product cleared a human regulator anywhere in the wealthy world.
That fact is mentioned in most reviews and then dropped. It deserves better, because the veterinary record is the only large, unglamorous, real-world data set we have on what a BHOC does when nobody is running a trial around it. This post reads that record for what it teaches, and for what it cannot teach.
The answer is not sentimental. It is structural.
A human oxygen carrier has to answer to a regulator that asks one brutal question: does it reduce death compared with red cells? Red cells are the comparator, they are excellent, and in the settings where they are available they are hard to beat. A carrier that merely matches them on paper still has to explain why anyone should take a new risk.
A dog with immune-mediated hemolytic anemia is in a different position. A veterinary clinic often has no matched canine blood on the shelf. Canine blood banks exist, but they are patchy, most practices do not stock units, and a unit of dog red cells has a short shelf life compared with a room-temperature bag of polymerized hemoglobin. The comparison is not red cells versus BHOC. It is frequently BHOC versus nothing in the next hour. The bar for benefit is lower, the question is simpler, and the regulator can approve on a clinical endpoint such as improvement in the signs of anemia without demanding a mortality win over an unavailable alternative.
Two other features helped. Dogs have no strong naturally occurring antibodies against a first transfusion of the common blood types, so the immunologic story is less pressing than in cats. And the bovine hemoglobin in Oxyglobin was purified, cross-linked with glutaraldehyde into polymers, and stripped of the red cell membrane, which is where the blood group antigens live. Nothing about blood type has to be matched. That single property, universal compatibility, is the reason emergency clinicians liked it.
Oxyglobin was supplied as a sterile solution with a hemoglobin concentration in the low teens of grams per deciliter, packaged in a flexible bag, and stored at room temperature for a shelf life measured in years. The hemoglobin sat in plasma, not in cells. It raised the oxygen content of blood without raising the hematocrit, which is a strange thing to see on a monitor. The dog’s packed cell volume could remain low while the plasma turned the color of red wine.
The half-life in the circulation was short, on the order of a day. This matters. It means the product is a bridge, not a repair. It carries oxygen while the dog’s marrow, or a transfusion, or the treatment of the underlying disease catches up. Every serious veterinary discussion of the product says the same thing in different words: it buys time.
And it did have a signature side effect, the one every clinician learned first. Free polymerized hemoglobin in plasma discolors things. Mucous membranes look yellow or brownish, the whites of the eyes tint, urine turns red-brown. Nothing is wrong with the kidneys in those cases, yet a new technician panics. Colorimetric laboratory tests can also be thrown off, because the pigment interferes with the reading. Chemistry values drawn while the product is circulating have to be interpreted with that in mind.
The most useful thing the veterinary record teaches has nothing to do with oxygen. It is about volume.
Polymerized hemoglobin is colloid-active. It pulls water into the vessels. In a dog with a healthy heart and a low blood volume, that is a feature. In a dog with heart disease, or a cat of almost any description, it is a hazard. Clinicians learned to give the product slowly, to watch respiratory rate, and to avoid it in animals already close to fluid overload. Cats in particular, with small blood volumes and a tendency toward hidden cardiac disease, showed a narrow margin. Pulmonary edema and pleural effusion became the reported adverse events that mattered.
Now translate that to a person. A large bolus of a colloid-active oxygen carrier, given to a patient already receiving other fluids, is a volume load. The human trials of hemoglobin products reported effects on blood pressure and heart function that are tangled up with this same physiology. The veterinary literature does not settle the human argument. It does show something clean: dose, rate, and cardiac reserve interact, and the people who use these products daily learned that from animals long before a human protocol wrote it down.
Here is the part I find most valuable, and I have not seen it written anywhere in the formal literature. When a product is on the shelf, practice changes.
A clinician who knows a compatible, ready oxygen carrier exists behaves differently at intake. The decision to start supportive care no longer waits for a donor, a crossmatch, or a courier. A dog in crisis gets something in minutes. The effect is not on the animal that would have been transfused anyway. It is on the animal that would have been euthanized, or referred too far, or lost in the delay while everyone searched for blood.
This is an argument about availability that human medicine understands well in the abstract and rarely measures. Every emergency physician has a story of the patient who arrived at the wrong hospital, on the wrong night, with the right blood one hundred kilometers away. A room-temperature bag that needs no typing turns geography into a smaller problem. Oxyglobin was proof, in a modest and cheap setting, that this is not a theory.
Oxyglobin did not have a clean commercial life. Its manufacturer went through financial trouble, ownership passed between companies, and supply became unreliable in later years. Veterinarians who had come to depend on it found themselves calling around for the last units in the country. Sometimes they used up expired stock with the owner’s informed consent.
Read that as a systems finding, not a scandal. A product that saves animals but is made by a thin business will vanish. The veterinary market was never large enough to keep a hemoglobin production plant profitable on its own, and the same plant was chasing a far bigger human indication that never arrived. The result is a paradox that anyone advocating for BHOC should hold in mind. The technology worked well enough to be missed. It was not sustained.
For a patient-advocacy audience, this suggests a practical lesson. Approval is not the same as durability. A carrier needs a manufacturing base, a supply agreement, and a market that pays for it, or it will be a good idea that lasts twelve years. The veterinary case shows that a niche, cleanly regulated indication can hold a product alive. It also shows that a niche alone cannot.
There is a further reason the veterinary field matters, and it involves working animals. Military and police dogs are treated with something close to trauma-medicine standards, and their handlers care intensely. Field veterinarians who operate far from blood banks have every reason to use a shelf-stable carrier, and published discussions of canine casualty care reflect interest in exactly that kind of product.
I want to be careful here. I am not going to claim detailed outcome figures for combat dogs, because I could not verify precise numbers, and inventing them would be worse than leaving the space blank. What can be said is narrower and still useful. Where an animal is wounded, remote, and hemorrhaging, the logistics look the same as they do for a soldier: no cold chain, no cross-match, and a helicopter that may not come. The reasoning that made a hemoglobin carrier appealing for dogs in a city practice is stronger in a forward position.
The transfer problem is the honest limit. Dogs are not small people. Their physiology of hemoglobin binding, their sensitivity to vasoconstriction, their immune response to a bovine protein, and their disease mix are all different. A drug that behaves well in a large, deep-chested animal with an autoimmune anemia is being tested in a very different circulation from a person with a torn femoral artery. The veterinary success tells us the product is manufacturable, storable, and tolerable at the right dose. It does not tell us it reduces human mortality.
There are four things I would carry out of the veterinary record and into any conversation about human access.
The first is proof of logistics. A polymerized bovine hemoglobin solution can be manufactured at scale, released, shipped, stored at room temperature, and used by ordinary clinical staff. This is no longer an argument. It has been done.
The second is a benchmark for tolerability. Discoloration, interference with laboratory tests, and volume overload are the predictable events. Any human program should list those first, with the veterinary experience as a floor.
The third is a warning about supply. A mission that depends on one manufacturer is fragile. If a foundation wants access to matter in five years, it needs to ask who makes the product, who funds continuity, and what happens if the owner changes.
The fourth is a caution about wording. When someone says “it has been used safely in animals,” the useful reply is: which animals, at what dose, in which disease, and safe compared with what? The answer for dogs with hemolytic anemia is quite good. The answer for a cat with a failing heart is more guarded. The answer for a person is not settled by either.
There is a sentence that gets said quietly in veterinary circles and rarely in human ones: an animal in acute anemia with no donor has a treatment, and a person in the same position, in most places, does not. That is not an argument that dogs matter more. It is a description of how regulation, liability, and market size interacted. Human trials had to prove that a product beat a very good standard, and the first generation of products did not always do it. The veterinary route only needed to prove that the product helped a patient who otherwise had little.
Whether that difference is right or wrong is a policy question for other posts. As a piece of evidence, it points in a clear direction. If there is a human setting where the alternative to a carrier is really nothing, because blood is unavailable, unacceptable, or unmatched, then the veterinary logic applies to that setting almost directly. That is why the conversation about expanded access, and about people who cannot receive blood for reasons of conscience or logistics, keeps returning to the same product family.
The dog got there first because the question asked of the dog was easier. The rest of us are still working out whether it should be.
Publicly available veterinary labeling and clinical reviews of Oxyglobin describe its indication, dosing, half-life, and adverse events, and are the best primary starting point. Readers should treat any figure quoted here as approximate and confirm it against the current label or a veterinary pharmacology source before relying on it.
Human hemoglobin is a small protein. Four globin chains, four hemes, about 64,000 daltons. It works because it is packed by the hundreds of millions into a red cell, where a membrane, a battery of enzymes, and a bath of the small molecule 2,3-BPG keep it in working order. Take it out of the cell and the whole arrangement falls apart. The molecule splits into halves, leaks through the kidney filter, grabs nitric oxide in the vessel wall, and rusts.
Almost every effort to build a hemoglobin-based oxygen carrier has been an effort to fix those four problems by force: cross-link the halves, polymerize the pieces, wrap them in polymer, or shrink them into a vesicle. Bovine hemoglobin (the BHOC route) inherits the same logic. It is a clever way to repair a small protein that was never meant to live alone.
There is a different starting point. Some animals never put their hemoglobin in cells at all. They dissolve it, in giant assemblies, directly in their blood. Those molecules have been doing what the engineers want for hundreds of millions of years. This post is about what they look like, why they work, and what a careful reader should and should not conclude from them.
The common earthworm, Lumbricus terrestris, carries an oxygen protein called erythrocruorin. It is not a hemoglobin in the sense we usually mean, although it is a relative. It has a molecular mass of roughly 3.6 million daltons, around fifty-five times heavier than human hemoglobin. It is built as two stacked hexagonal rings, with something on the order of 144 globin chains arranged around a scaffold of linker chains. Under an electron microscope the particle looks like a small, flat, six-fold flower about thirty nanometers across.
That size is the point. A molecule of that mass cannot pass through the kidney filter. It cannot slip between the cells lining a vessel and wander into the tissue, where it would scavenge nitric oxide and squeeze the surrounding vessels shut. It does not dissociate into dimers, because it is held together by many contacts, and in the earthworm it is stabilized by calcium. It carries dozens of oxygen molecules per particle. It is a molecule that behaves the way a small red cell would, minus the cell.
The worm has other advantages. Its erythrocruorin is cooperative, meaning it binds oxygen with the sigmoidal curve that allows efficient loading and unloading. Studies on the purified protein show it holds up to oxidation more gracefully than many engineered hemoglobins, and it does not require a cell to keep it reduced. I would treat that claim with normal care: the data come from a small number of laboratories and mostly from preclinical settings. But the direction of the evidence is consistent.
The marine lugworm, Arenicola marina, carries a similar giant hemoglobin. A French company, Hemarina, has developed a product from it, often called M101 or HEMO2life. The company reports that the molecule carries far more oxygen per particle than human hemoglobin and, notably, that it has an enzyme-like ability to neutralize reactive oxygen species. The stated rationale is protection of organs kept outside the body before use.
In 2016 the product was used in France in a first-in-human case as a preservation additive for a donated kidney. Since then it has been tested in a small number of further studies. I am deliberately keeping the description at that level. The published human experience is small, the design is not a randomized comparison of survival, and much of what is said about it comes from the developer. What deserves attention is not the marketing. It is that a giant, cell-free, non-human oxygen carrier has been put through a regulated human procedure at all.
Two things follow. First, the safety signal for the worm molecule in people, so far, is clean enough that studies have continued. Second, and more important for the BHOC conversation, it demonstrates the principle that a very large hemoglobin can be given to a human without the vasoconstriction that defined earlier products. That is a proof of mechanism, not of benefit.
The strangest example belongs to Riftia pachyptila, the giant tubeworm of deep-sea hydrothermal vents. It lives beside water rich in hydrogen sulfide, a compound that shuts down mitochondrial respiration in almost every other animal. It has no mouth and no gut. Its tissues contain bacteria that oxidize sulfide for energy, and the worm feeds the bacteria by delivering both oxygen and sulfide to them at the same time.
Doing that in a single bloodstream is a chemical trick. Sulfide and oxygen react with each other, and free sulfide poisons the heme. Riftia’s extracellular hemoglobins solve the problem by binding sulfide at a site that is not the heme. Work published in the mid-2000s identified zinc ions on the protein as part of that binding site, so oxygen occupies the iron while sulfide rides safely on the protein, held away from the place where it would do harm. The molecule is a two-cargo carrier with a built-in protective device.
Why should a program on bovine hemoglobin care? Because the problem the tubeworm solved is the general problem of a cell-free carrier: a reactive iron center exposed in plasma, where it meets things it would never meet inside a red cell. The vent worm’s solution was to place the hazard on a different part of the protein and to make the protein big and stable. That is an engineering hint, not a treatment. Nobody is going to infuse tubeworm hemoglobin into a patient because it handles sulfide.
Set the earthworm, the lugworm, and the tubeworm side by side and a design pattern emerges that has little to do with any one species.
They are large. A few million daltons keeps the molecule in the circulation, out of the kidney, and out of tissue where it could quench nitric oxide.
They are assembled from many chains. Redundancy makes the particle robust: the loss of a single chain does not destroy it, and the many contact points hold the complex together without artificial chemical cross-links.
They are protected by their own scaffold. The linker chains and the metal ions are not decoration. They tune stability, and in the tubeworm they carry a second, dangerous cargo.
They are cooperative. Even at that size they release oxygen on a curve that responds to tissue demand.
Compare this with polymerized bovine hemoglobin. A BHOC is a mixture of chemically linked polymers, with a spread of sizes. Some fraction is small enough to leak, and that fraction is where the historical concern about blood pressure and kidney injury comes from. Manufacturers have pushed to remove low-molecular-weight species for exactly that reason. The natural giants show what happens when the size distribution is not a mixture but a single, defined particle.
It is tempting to end here with a triumphant claim that worms have solved the problem and the industry should switch. That would be wrong, and the reasons deserve to be spelled out.
The first is immunity. A protein from a marine annelid is more foreign to a human immune system than a protein from a cow. Bovine hemoglobin is still foreign, but cattle proteins have a long history of clinical exposure, and the purified molecule is a globin with a recognizable human-like fold. A single dose of an invertebrate particle may be tolerated. The picture with repeated doses, in patients who need support over days, is far less certain. Antibodies could shorten the half-life or trigger reactions.
The second is supply and cost. Cattle are an established, regulated, large-scale source. Lugworms are harvested from tidal flats. Earthworms are farmed, but the protein has to be extracted from thousands of animals. Any of these can be scaled, but nobody has yet done so at the quantity that a national emergency program would require. The company behind the lugworm product has treated the supply as a proprietary manufacturing process, which is reasonable, and also means outsiders cannot audit it.
The third is oxygen affinity. A carrier that binds oxygen too tightly will not give it up where it is needed, and a carrier that binds too loosely will unload before it reaches the tissue. The natural giants evolved for the temperature, pressure, and oxygen environment of their owners. A worm that lives in cold mud, or on a vent chimney, works at a different oxygen tension from a person in shock. The equilibrium of the protein might need to be tuned, and the assembly is so complex that tuning it is difficult.
The fourth is evidence. Almost everything said about these molecules comes from a handful of laboratories and one developing company. That is normal for an early field. It also means that the claims have not been tested by the adversarial process that eventually broke earlier products.
There is a fifth issue that gets less attention: regulation. A regulator asked to approve a bovine protein has decades of precedent for cattle-derived biologics, including tests for prion contamination and viral clearance. A regulator asked to approve a worm protein has to write those expectations from scratch. That takes years, and it favors the developers who have already begun. It does not make the science wrong. It does mean that, for a patient who needs a product within this decade, the well-worn path is a real advantage.
I do not read the natural giants as a threat to the BHOC approach. I read them as a design specification written by evolution.
A bovine hemoglobin product can borrow from the worms in three ways. It can push toward larger, more uniform polymer sizes, which is what several developers are already trying to do. It can add stability through protein-based scaffolds rather than harsh chemical cross-links. And it can pair the carrier with protective elements that buffer its own chemistry, in the spirit of the sulfide-shielding tubeworm.
There is also a second-order use. Giant hemoglobins make useful controls in the laboratory. If a bovine product causes vasoconstriction in an animal model and a giant erythrocruorin does not at a comparable dose, the difference tells you which property of the bovine product matters. That is scientifically worth more than a claim in a brochure.
A patient community will hear enthusiastic descriptions of worm blood. A fair listener can ask five plain questions.
How large is the particle, and is it a single defined species or a mixture?
Where did the material come from, and can supply scale without harming an ecosystem?
What happens on the second dose?
How tightly does it hold oxygen, and at what temperature and oxygen tension was that measured?
Who besides the developer has tested it?
None of those questions is hostile. Answering them is what turns a beautiful fact about a worm into a medicine.
The earthworm sitting in a garden carries in its blood a molecule far larger and more elegant than anything a laboratory has made. It has never needed to be a red cell. That is worth knowing, and worth borrowing from, with caution.
Primary sources include published biophysical work on Lumbricus terrestris erythrocruorin, Hemarina’s public descriptions and clinical registrations for its lugworm product, and the mid-2000s structural and biochemical work on sulfide and zinc binding in Riftia pachyptila hemoglobins. Figures such as particle mass and chain counts are approximate and should be checked against those sources.
Most writing about hemoglobin-based oxygen carriers asks whether they work. This post asks a prior question that gets far less attention: who agreed to receive them, and how?
The answer is uncomfortable, because the setting where a BHOC is most needed is the setting where consent is hardest. A person losing blood in a road accident, a battlefield, or a remote clinic is often unconscious, in shock, or drifting in and out of understanding. The product is experimental or restricted. The normal machinery of informed consent, which assumes a calm conversation with time to think, does not apply. What replaces it is a set of legal exceptions and moral compromises that most patients, and most advocates, have never examined.
In the United States, federal regulation allows a narrow exception to informed consent for emergency research. The rule, found in Title 21 of the Code of Federal Regulations at section 50.24, permits an investigation to proceed without the patient’s consent when several strict conditions are met. The patient’s life must be at risk. Available treatments must be unproven or unsatisfactory. Obtaining consent must be impossible because of the patient’s condition, and there must be no time to reach a legal representative. The research must offer a real prospect of direct benefit. And, unusually, the sponsors must consult the community from which participants will be drawn, and publicly disclose the trial before it begins.
This is the framework under which the best-known trial of a hemoglobin carrier in trauma was run. The product was a human-derived polymerized hemoglobin called PolyHeme. Patients with severe hemorrhage were randomized in the field, in ambulances, to receive either the carrier or standard saline, and after arrival at hospital, either the carrier or red cells. Because the whole point was to test treatment in the first minutes, waiting for consent would have defeated the study.
The trial drew heated criticism from bioethicists, some of whom argued in print that community consultation had been thin, that the control arm meant some patients would receive an unapproved substance instead of blood available at the hospital, and that the risk information available at the time did not support an exception. Supporters replied that the exception existed precisely for cases like this, that public meetings had been held, and that people who did not want to participate could wear opt-out bracelets. Both sides described real things. Neither side was inventing.
What I want to highlight is a structural point. The exception treats consent as a property of a community rather than of an individual. Nobody in the ambulance said yes. A group of strangers in a city, at a meeting most residents did not attend, said yes on behalf of whoever might one day be put in that ambulance. That is a strange kind of permission, and it is worth understanding before anyone advocates for or against it.
The moral weight of the exception depends on how big the risk is. That risk was recalculated in 2008, when a group of researchers published a pooled analysis of trials of several hemoglobin carriers across many indications. It combined sixteen trials and more than three thousand patients, and reported that treatment was associated with a higher risk of death and of heart attack than control. The absolute numbers were small in some categories, and the products were not interchangeable. But the signal was enough for the exception’s basic premise, a prospect of direct benefit, to look shakier in hindsight.
It is important to be fair about what that analysis did and did not say. It pooled different molecules, made by different methods, at different doses, in different diseases. It did not show that every future product will cause harm. It showed that a class of first-generation products carried a hazard that individual trials had not been powerful enough to detect. From an ethics standpoint, this is the cleanest illustration of why the consent problem cannot be solved by quick permission. If the risk of a product is not known until years after a large trial ends, then a community that agreed to it in advance agreed to something nobody could describe.
Now turn to the reverse situation. There are people who can be quite sure they do not want a transfusion, and who are also in danger of bleeding to death. Jehovah’s Witnesses, on the basis of their reading of scripture, refuse transfusion of whole blood, red cells, white cells, platelets, and plasma. They regard some fractions, such as albumin, as a matter of individual conscience. Many accept hemoglobin-derived products, but not all, and the decision differs from person to person.
For this community, a hemoglobin carrier has a moral profile that does not resemble that of an experimental trauma drug. It is an alternative to a treatment they will not accept. The question of consent is reversed: instead of “how do we permit a treatment the patient cannot approve,” it becomes “how do we make sure the patient’s refusal is honored, and how do we offer a substitute that does not violate it.”
This is where the source of the hemoglobin matters. A bovine BHOC is not human blood. It contains no human cells or plasma. For many Witnesses, that distinction makes it acceptable, and it has been used in emergency cases where a patient with life-threatening anemia refused blood but agreed to it. In those cases the product was typically obtained by a single-patient emergency request, and the consent conversation took place with the patient, if awake, or through the documents the patient had carried.
That last phrase points to something practical. Witnesses commonly carry an advance medical directive, a card or form stating what they will and will not accept. In an emergency, that card is the consent. It is the one consistent place in medicine where a patient’s answer exists before the question arrives. It also shows a gap: the directive usually lists blood components, and rarely mentions a bovine hemoglobin product. A patient who ticked “no fractions” might or might not have meant this. The clinician who finds the card cannot ask.
Outside trials, the route to an unapproved product is expanded access, sometimes called compassionate use. A treating physician asks the manufacturer for the product and asks the regulator for permission, and in a true emergency the regulator can authorize by telephone. The paperwork that follows includes a consent form, an ethics committee review, and a report of what happened.
Two features make this route strange for oxygen carriers. The clock runs in hours, and a hospital that has never done it before must find a manufacturer, an ethics reviewer, and a pharmacist in the same afternoon. And the consent given is real only if the patient can give it. Many of the reported single-patient cases involved adults who were conscious enough to say yes, or whose relatives could speak for them. That is a much easier situation than the roadside. It also means the public record of these cases says little about the hardest scenario, the one that motivates most calls for access.
There is a second conscience issue that affects a much wider group. A bovine product comes from cattle. For a Hindu patient, cattle are sacred, and receiving a product made from cows may be a serious problem. For some observant Muslim and Jewish patients, the question turns on the animal’s slaughter and the status of a derivative, and religious opinions differ. Vegetarians and vegans may object on ethical grounds. A porcine product would raise parallel concerns for other groups. A lugworm product would raise almost none of these, though nobody has seriously studied how patients feel about that.
None of this is exotic. Hospitals already handle it for animal-derived medicines such as heparin from pigs or gelatin capsules. The relevant lesson is that a BHOC consent form that says only “this product is derived from bovine hemoglobin” is not enough for the emergency setting, because the person who needs to know is not there to read it. If a product is going to be stocked in ambulances or forward field units, its source must be visible at a glance and its presence must be recorded in a way that follows the patient into the hospital.
Emergency medicine has a partial answer to the timing problem, and it is worth borrowing. When treatment was given without consent, the patient or their family should be told as soon as possible, and the patient should be able to withdraw from further participation and, where research is involved, from data use. This is called deferred or retrospective consent. It does not undo the treatment. It does restore the patient’s standing as a person rather than a specimen.
I have noticed that discussion of deferred consent in the hemoglobin literature tends to be procedural: who sends the letter, on what day. The patient’s actual experience is rarely described. Imagine waking up in a hospital bed, being told you were given a red solution made from cow blood, and being asked to sign a form. The person may feel grateful, or violated, or both. How a system handles that moment determines whether communities support the next trial.
I would suggest five design principles that follow from the above.
Pre-consent should be encouraged where it is possible. People who anticipate risk, such as those who work in remote areas, who serve in the military, or who have conscience-based limits, should have a way to record their wishes about oxygen carriers specifically, not only about blood.
Labels and records should name the source. Species of origin, manufacturing route, and lot must travel with the patient.
Community consultation should be real. That means meetings with translation, with faith leaders, with groups who might object, and with people who have been through emergency care, not only with a committee.
Disclosure after treatment should be prompt and plain, with a clear route to ask questions and to opt out of research use.
Exception should not become habit. If a product has been used in enough patients to be described, then its risks can be explained in advance, and the emergency exception should give way to ordinary consent wherever a competent patient exists.
A foundation working on access to BHOC is usually asked for evidence of efficacy. It should also be prepared to answer questions on consent. Who gets to say yes? Who is protected when the answer cannot be given? How do people with conscience objections, religious or otherwise, find out what they are being offered?
An advocacy group that raises these questions first earns credibility. It signals that access is not being pursued at any cost. It also protects the patients who most need the product, because a trial that loses public trust will be closed, and the next one will be harder to open.
The oxygen carrier that saves a life at a roadside is a scientific accomplishment. The system that lets a stranger receive it is a moral one, and it has to be built with the same care.
The regulatory text is 21 CFR 50.24. The pooled safety analysis of hemoglobin carriers was published in the Journal of the American Medical Association in 2008. Bioethics commentary on the PolyHeme trial appeared in several medical and ethics journals during 2006 and 2007. Guidance for clinicians treating Jehovah’s Witnesses is published by hospital liaison committees and should be read alongside a patient’s own directive.
Suppose a patient in shock receives a BHOC and, an hour later, looks better. The blood pressure has come up. The skin is warmer. The lactate is falling. Did the oxygen carrier do that?
The honest answer depends almost entirely on what you measured, how, and with what instrument, and the instruments that we rely on were built for a world without free hemoglobin in plasma. This post walks through the measuring tools of oxygen delivery, what each can tell you, and the ways each can mislead when a cell-free carrier is present. It may be the most technical post on this site, but it is also the one that most affects whether a claim can be believed.
Oxygen delivery is the product of blood flow and the oxygen content of blood, and the aim of delivery is not the blood but the tissue. Everything measured at the bedside is an indirect reading of what the cells are actually receiving.
A carrier can raise blood oxygen content and still fail the tissue, if it changes flow. It can look useless on a global measure and still rescue a small region. And it can distort the test itself. Each of the following tools has a way of being wrong.
The most common lab value in medicine is the hemoglobin. Every clinician reads it as a count of red cell hemoglobin and treats it as a fuel gauge. When a BHOC is circulating, that reading combines two populations: the hemoglobin inside cells, and the hemoglobin freely dissolved in plasma. A total hemoglobin of nine may hide a red cell fraction that is far lower.
The hematocrit is even more revealing. It measures the volume fraction of cells and does not include plasma hemoglobin at all. A patient may have a hematocrit that looks alarming, and a total hemoglobin that looks acceptable, and both may be true at once. Clinical trial protocols have had to define separate plasma hemoglobin measurements, often by spinning a sample and reading the clear supernatant, to make sense of this.
The danger is behavioral. A physician who sees a low hematocrit after a BHOC may give a red cell transfusion that was not needed, or refrain from giving one that was. The result is a distorted picture of both safety and efficacy in the data.
The finger clip works by shining two wavelengths of light through tissue and comparing absorption. It assumes that the only significant absorbers in arterial blood are oxygenated and deoxygenated hemoglobin. Cell-free carriers add a third pigment and, if the iron oxidizes, a fourth, methemoglobin. The result is that a saturation reading may be inaccurate in ways that vary by device and by product.
Methemoglobin, the oxidized form that cannot carry oxygen, tends to pull the reading toward the low nineties regardless of the true value. A patient might be badly deprived of oxygen and see a stable number on the monitor, or be adequately oxygenated and see a worrisome one. A carrier that oxidizes quickly, which was a property of some first-generation products, makes this worse.
The laboratory answer is co-oximetry, a bench instrument that measures light absorption at many wavelengths and separates oxyhemoglobin, deoxyhemoglobin, carboxyhemoglobin, and methemoglobin. It is the reference for whether the pigment in the sample is still working. It also has to be calibrated for the specific carrier, since the spectrum of a polymerized bovine hemoglobin is not identical to that of human hemoglobin.
Many routine laboratory tests read a color change. Free hemoglobin in plasma tints the sample and can add to or subtract from the reading. Liver enzyme assays, creatinine methods, bilirubin, and some cardiac markers have shown interference in the presence of free hemoglobin, with the size and direction depending on the method.
This is not an academic footnote. A reported rise in a liver enzyme after a carrier could be genuine injury or an artifact of the assay. A troponin result, used to detect heart muscle damage, is central to the safety debate about hemoglobin carriers, and the reliability of that test in the presence of the product has been discussed in the literature. The way to handle it is to specify in advance which methods are validated for use with the product and to run a second method when the result matters.
Near-infrared spectroscopy shines light of several wavelengths through a small volume of tissue and estimates tissue oxygen saturation, usually written StO2. A common site is the thenar eminence, the fleshy pad at the base of the thumb. The device is noninvasive and can be kept on for hours. It offers something no blood test can: a continuous signal from the microcirculation.
A clever use is the vascular occlusion test. A blood pressure cuff is inflated on the arm until flow stops, tissue saturation falls at a rate that reflects local oxygen consumption, and when the cuff is released the rate of recovery is a reading of microvascular responsiveness. Slow recovery in shock predicts worse outcomes. If a carrier improves recovery, it is doing something at the level that matters.
The limitations are real. Light passes through skin, fat, and muscle, and the signal is a blend of arterial, venous, and capillary blood. Fat thickness changes the reading. Edema, which is common in resuscitation, changes it again. And because the algorithm assumes a certain set of absorbers, a plasma pigment can bias the result. Trend within a patient is more trustworthy than an absolute number, and comparisons between patients or between studies need caution.
The most elegant tool in the field is one that most clinicians have never seen. Certain molecules, when excited by light, glow, and the glow fades over a characteristic time. Oxygen quenches that glow. The more oxygen is present, the faster it fades. By measuring the decay time, one can compute the oxygen pressure directly, with no reliance on the color of the blood.
Researchers built probes from palladium porphyrin compounds designed to remain in the circulation or to stay in tissue, and they used them to map oxygen pressure in the microvessels of animals. In experiments, one can look at a single arteriole and a single venule and read the oxygen tension in each. This method is how the field knows, for example, that oxygen leaves the arterioles before it ever reaches the capillaries, a fact that changed how everyone thinks about where a carrier delivers.
It is a research instrument, not a bedside device. It requires a microscope, a probe, and a window into the tissue. But it establishes the physiological truth against which everything else is judged. When a BHOC claims to improve tissue oxygenation, this is the type of measurement that can support or refute the claim.
One of the most influential ideas in this field came from intravital microscopy. In a transparent skin fold chamber on a hamster, capillaries can be watched directly. The count of capillaries with flowing red cells per unit area is called functional capillary density. In hemorrhagic shock it falls, because capillaries collapse or clog, and tissue oxygen falls with it.
The finding that made the field pay attention was that resuscitation with fluids of different properties restored functional capillary density to very different degrees, even when the systemic numbers looked similar. Carriers with high oxygen affinity and low viscosity kept more capillaries open in these experiments, and the proposed reason was that they preserved flow rather than merely adding oxygen content. That idea became the microvascular hypothesis and drove a generation of product design.
The catch is translation. Functional capillary density is measured in a hamster chamber, which cannot be used in a person. The bedside surrogates, sidestream dark-field imaging under the tongue, are useful but variable and depend heavily on operator technique. Whether the hamster result predicts survival in a person remains unresolved.
Lactate is the traditional marker of an oxygen debt. When cells cannot get enough oxygen they switch to a less efficient metabolism that produces lactate, and blood lactate rises. A falling lactate after resuscitation is read as evidence that delivery has improved.
Lactate has flaws. The liver clears it, so a damaged liver can hold it high for reasons unrelated to oxygen. Stress and certain drugs raise it without hypoxia. And it lags: the number describes what has already happened, not what is happening now. Base deficit, a measure of acid load, has similar limits. Both are useful trends. Neither is a direct measurement of tissue oxygenation, and neither is immune to the changes a carrier can make to blood chemistry.
Put all of this together and a pattern emerges. The measurements that are easy, such as hemoglobin, saturation, and blood pressure, are the least reliable when a carrier is present. The measurements that are informative, such as phosphorescence quenching and capillary counts, are impossible at the bedside. The middle ground, near-infrared spectroscopy and lactate, is helpful but imperfect.
The regulatory consequence is that trials must decide what to call success. A surrogate endpoint, such as an improved tissue saturation or a fall in lactate, is a stand-in for the outcome that matters, which is survival and recovery. A surrogate is only as good as the evidence that changes in it predict changes in the real outcome. For oxygen carriers, that evidence has been thin, and there is a history of products that improved a surrogate and did not improve survival.
When a press release or a talk says a carrier improved oxygenation, ask which measurement was used, and whether it was validated in the presence of free hemoglobin. Ask whether the comparison was between patients or within the same patient over time. Ask whether the plasma hemoglobin was reported separately from the red cell hemoglobin. Ask whether the endpoint was a surrogate or a clinical outcome, and if it was a surrogate, what evidence links it to survival.
These are not obstructive questions. They are the ones a trial statistician would ask, and an advocacy organization that can ask them will be listened to.
The field would benefit from a shared measurement protocol: a defined set of instruments, calibrated for the specific carrier, with a plasma hemoglobin measurement, co-oximetry, a near-infrared trend with the occlusion test, and lactate, all reported in a consistent format. It would not be glamorous. It would let results from different studies be compared, which today they largely cannot.
Until that exists, every claim about a BHOC should be read as a claim about one instrument, on one patient population, in one laboratory. That is not a reason for despair. It is the ordinary discipline of science, and it is what will eventually separate the products that work from those that only look as if they do.
Foundational work on phosphorescence quenching oximetry comes from the laboratories of David Wilson and Sergei Vinogradov. The microvascular hypothesis and functional capillary density studies are associated with Marcos Intaglietta and colleagues. Reviews of near-infrared spectroscopy in shock and of interference by cell-free hemoglobin in laboratory assays are available in critical care and clinical chemistry journals. Specific numerical claims here are qualitative and should be checked against those sources.
Every conversation about a bovine hemoglobin oxygen carrier eventually reaches a question that the listener is usually too polite to ask. It comes from cattle. Cattle carry a disease that frightened a continent. Why should anyone be comfortable with a medicine that starts in a cow?
The polite answer is that the risk is negligible and the product is purified. That answer may well be right, but it is a conclusion and not an argument, and a patient community deserves the argument. This essay walks through the argument the way a supply-chain engineer would, from the pasture to the vial, because the safety of a BHOC is not a single property of the molecule. It is a property of an entire chain of custody, and every link in that chain has its own physics, its own failure modes and its own audit trail.
Bovine spongiform encephalopathy, usually shortened to BSE, was recognized in British cattle in the mid-1980s. It belongs to a family of diseases caused by prions, misfolded forms of a normal host protein that convert their normal neighbors into more of the misfolded form. Prions are not bacteria or viruses. They carry no genome. They resist heat, radiation, formaldehyde and most of the tools that sterilize ordinary pathogens, which is precisely why the disease became a public health crisis rather than an agricultural nuisance.
The turning point came in 1996, when British authorities reported that a new human disease, variant Creutzfeldt-Jakob disease, was linked to exposure to the BSE agent. The cattle epidemic had been amplified by feeding ruminant-derived protein back to ruminants, a practice later banned. The human cases followed dietary exposure, chiefly to neural tissue that entered meat products. The United States identified its first case of BSE in December 2003, in a cow that had been imported from Canada, and regulators on both sides of the Atlantic rebuilt their rules around a single insight: the infectious agent is not spread evenly through the animal.
That insight is the foundation of everything that follows. In an infected animal the agent concentrates in the brain, the spinal cord, the eyes, certain ganglia and the distal portion of the small intestine. It is not concentrated in muscle, and in the bioassay experiments that have been done, infectivity has not been detected in cattle blood. That last statement deserves care. Absence of detected infectivity in an assay is not proof of absence, because every assay has a limit of detection, and prion assays in particular are slow, expensive and species-dependent. It is evidence, not a guarantee, and a careful writer says so.
There is a comparison here that the bovine skeptic seldom hears, and I find it clarifying. The only documented cases of person-to-person transmission of variant Creutzfeldt-Jakob disease through blood came through human blood. In the United Kingdom, a small number of probable transfusion-transmitted infections were identified, on the order of a handful, in recipients of red cells donated by people who were later diagnosed with the disease. Those donors were incubating the infection silently, and the cells had not been filtered for white cells, which are thought to carry more of the agent than the red cells themselves.
The consequence is a strange asymmetry. Society accepted human blood, which has a demonstrated if rare route for prion transmission, and built screening rules around it: deferral of donors who had lived in the United Kingdom during the peak years, leukoreduction of cell products, and lifetime deferral of anyone who had received a transfusion there. Meanwhile, a bovine product has no demonstrated route for transmission at all, but its source animal carries the name of the disease. The perception of risk and the measured risk point in opposite directions.
None of this means a BHOC is risk-free. It means the risk should be examined where it actually lives, and it lives in three places.
The first defense is the choice of animal, and it has three parts.
The first is geography and history. A herd raised in a country or region with a very low or zero record of BSE, with documented feed controls, starts with a low prior probability of infection. When a manufacturer describes a herd as closed, it means the animals were born into it and no outside cattle entered, so that a contaminated animal cannot be introduced.
The second is age. BSE has a long incubation period, typically measured in years, and detectable disease in young animals is vanishingly rare. Restricting collection to animals below a certain age is a blunt but effective tool.
The third is feed. The epidemic was driven by rendered ruminant protein in cattle feed. A herd raised on pasture and plant feed, with an auditable record of what each animal ate, removes the mechanism that amplified the disease in the first place.
What I find striking about these controls is that they are cheap once they are designed into the operation and enormously expensive to add afterward. A company that builds a BHOC around a certified herd from the start has a very different risk profile from one that buys blood on the spot market. Regulators know this, which is why documentation of the source herd, its records and its veterinary oversight sits near the front of any submission.
The second link is where an unexpectedly physical hazard hides, and it is the detail I have never seen given the attention it deserves.
To collect blood from a slaughtered animal, the animal must first be stunned. One family of stunning methods uses a captive bolt, a device that drives a rod into the skull. Studies of that method have shown that it can push fragments of brain tissue into the circulation, where they can appear in the blood draining from the neck. Some designs inject compressed air into the cranial cavity, which increases the risk further. Regulators in several jurisdictions restricted or prohibited the more disruptive techniques for this reason.
Think about what that means. In an infected animal, the very tissue where prions concentrate could reach the blood at the moment of collection. The blood itself may carry little or no agent, but the process of obtaining it could add some. That is why a well-designed collection protocol treats stunning as a design variable. It specifies non-penetrating stunning devices, avoids methods that disrupt the brain, and monitors collected blood for markers of central nervous system tissue contamination. This is a subtle point, and it is one that an outsider would not think to ask about, yet it separates a careful supply chain from a casual one.
A related discipline concerns the animal’s inspection. Animals that are unwell, that cannot walk unassisted or that show neurologic signs are excluded, because those are the animals in which any residual disease risk concentrates. The rules regulators wrote for food after 2003 embody the same idea, banning material from what they call non-ambulatory disabled cattle and removing high-risk tissues from all older animals.
The third link is manufacturing. A purified hemoglobin solution is not blood. It is the product of a sequence of steps that begins with red cells and ends with a protein in a defined solution. The red cells are separated from plasma and washed. They are broken open. The membranes and other cellular debris are removed by filtration. The hemoglobin is then purified, often by chromatography, and concentrated. It is chemically polymerized and filtered again, and finally it is packaged.
Every one of those steps removes something. The question for prion safety is how much. The answer is measured in what regulators call log reduction. In a validation study, a manufacturer deliberately spikes a sample of intermediate material with a known amount of a prion preparation, runs it through the step, and measures how much survives. A step that removes a thousand-fold has a three-log reduction. Steps in series add their reductions, provided they act by independent mechanisms.
For prions, the workhorse mechanisms are physical partitioning, such as precipitation, filtration and chromatography, which physically separate the protein aggregates from the product, and chemical inactivation, such as treatment with strong alkali, which is one of the few reagents known to degrade prions meaningfully. Published work on the manufacture of human plasma-derived proteins has found that some individual steps deliver several logs of clearance, which is why plasma product makers rely on those numbers. A BHOC manufacturer borrows the same toolkit and the same validation logic.
I want to be circumspect here. Clearance studies are model experiments. They depend on the spiking material resembling the natural agent, on the scaled-down process resembling the real one, and on assays that are sensitive enough. Regulators accept them because there is no better alternative, and because the evidence base for individual steps has grown over three decades. But a thoughtful advocate should describe them as a layer of risk reduction that adds to the earlier layers, not as a proof.
Prions dominate the conversation because of their cultural weight, but they are not the only concern in a bovine product, and it would be dishonest to leave the others out.
Cattle harbor viruses. The one most familiar to pharmaceutical manufacturers is bovine viral diarrhea virus, which has long been a known contaminant of bovine-derived reagents such as fetal calf serum used in cell culture. It does not infect people readily, but its presence would signal a failure of upstream controls, and manufacturers test for it. Others include bovine parvovirus and a group of enveloped viruses, and the manufacturing process needs steps demonstrated to remove or inactivate both enveloped and non-enveloped types.
There are also endotoxins, fragments of bacterial cell walls that can trigger fever and hypotension even in tiny amounts. Blood collected from a slaughterhouse floor is exposed to environmental bacteria, and a product intended for intravenous use must meet strict endotoxin limits. Endotoxin is a plausible confounder in the early hemoglobin trials, because contamination in a first-generation product can produce a syndrome that looks like the effects of the molecule itself. The field has spent years learning to separate the two.
Finally there is the problem of the protein itself. Bovine hemoglobin is a foreign protein. It is more similar to human hemoglobin than a worm protein would be, but it is not human. Antibodies against it can develop, and a patient who received it once may react differently the second time. This is an issue of immunology rather than infection, and it belongs in the same conversation because it is part of what it means to be a bovine product.
Engineers speak of single points of failure, and a BHOC supply chain can have several. The most obvious is a single source herd. A disease outbreak in cattle, whether or not it involves BSE, could halt collection for months. A regulatory change in the source country could make the herd ineligible. A single processing plant, a single supplier of the polymerizing reagent and a single sterile filling line are further examples.
The history of Biopure, the American company that developed Oxyglobin and Hemopure, is instructive. It began in the 1980s by sourcing bovine blood from regional slaughterhouses, and over time it moved toward tightly controlled herd sourcing. Its later story, ending in a bankruptcy filing in 2009 and a sale of assets, was driven by financial and regulatory problems rather than a prion event. Yet even that lesson applies: when a company is thin, its supply chain is thin, because redundancy costs money.
An advocate who wants access to a BHOC should therefore ask not only whether it is safe today, but whether the chain that makes it can survive a shock. Does the manufacturer have more than one qualified herd? Are there stored reserves of intermediate material? Is there a plan for the day an animal in the herd tests positive for anything? A supplier that cannot answer those questions has a resilience problem even if the product is excellent.
Consider what a single vial of a BHOC should be able to tell you if you interrogated it. The lot number on its label leads to a batch record, and the batch record leads back through the purification steps to the pooled starting material, and the pool leads to the individual collection days, and each collection day leads to a list of animals with ear tags, birth dates, herd of origin and inspection results. In a well-run operation, that chain can be walked in either direction within hours. If an animal were later found to be infected with anything, the manufacturer could name every lot that contained its blood and every patient who received those lots.
This is the property called traceability, and it is the reason I trust documentation more than reassurance. A system that can be audited backward can also be corrected forward. It is worth noticing that this is exactly the property human blood banks acquired only after painful lessons in the 1980s, and that it was imposed on them from outside. A bovine product designed from the beginning with the same discipline starts ahead of the game.
The reverse detective exercise is equally revealing. Imagine a manufacturer whose records show blood pooled from many slaughterhouses with no animal-level identification. Even if every animal was healthy, the manufacturer could not prove it, and any adverse event would be untraceable. Regulators are right to treat that pattern as a serious deficiency, and patient advocates are right to ask about it. The question to put to any developer is short and plain: if a problem were found in one animal five years from now, how many minutes would it take you to find every patient exposed to its blood?
The regulatory frame in the United States is a layered one. The Food and Drug Administration has rules, first written in interim form and later finalized, restricting the use of certain cattle materials in food and cosmetics, and it has guidance on cattle-derived materials in medical products that requires manufacturers to document the country of origin, the age of the animals, the tissues used and the steps that reduce risk. Its position on the food supply is that the risk of human exposure through regulated products is negligible given the controls in place. A medical product goes through a stricter dossier than a food, because it is injected.
I would summarize what a regulator wants from a bovine-derived injectable in five sentences. Show me where the animals came from. Show me what they ate. Show me how they were inspected and how the blood was collected. Show me what your process removes, with validation data. Show me how you would know if something changed. Every element is documentation, and every element can be audited.
A patient community will hear two competing simplifications. One says cows carry mad cow disease, so bovine products are dangerous. The other says purification makes everything safe, so there is nothing to worry about. Both are wrong in the same way: they collapse a chain into a single link.
The accurate position is more useful. The prion risk of a BHOC is the product of several independent probabilities: that the animal is infected, that infectivity reaches the collected blood, that it survives the process, and that a dose sufficient to infect reaches a recipient by the intravenous route. Each factor is small, and the layered controls are designed so that no single failure is decisive. That does not equal zero. It equals a risk small enough that regulators and manufacturers of other bovine-derived injectables have accepted it for decades, and it is watched by a surveillance system that would notice a signal.
I would encourage three practical requests for any organization advocating access. Ask for the herd documentation, in summary form if the details are proprietary. Ask for the prion and viral clearance data, at least the number of logs achieved by each validated step. And ask for the plan for lot traceability, so that if a problem emerged years later, each vial could be traced back to an animal and a date. A manufacturer that answers those three questions readily is signaling a mature operation. One that deflects them is signaling the opposite.
The strangest lesson from this subject is a reversal of the usual intuition. We tend to trust what feels natural and fear what feels foreign, and human blood feels safer than cow blood. The record says the opposite is at least arguable. Human blood has carried a demonstrated prion route, while a carefully sourced bovine product has none on record. Feelings are a poor instrument for measuring that, and audit trails are a good one. The pasture is where the argument starts, and the paperwork is where it ends.
Regulatory background is available from the US Food and Drug Administration on bovine spongiform encephalopathy and on cattle-derived materials in medical products, including the 2007 rule on materials derived from cattle used in medical products and the 2016 final rule on cattle materials in food and cosmetics. Reviews of prion clearance during plasma protein manufacture describe the validation approach. Details of stunning methods and central nervous system emboli appear in the veterinary and food safety literature. Figures for the small number of transfusion-transmitted variant Creutzfeldt-Jakob disease cases come from United Kingdom surveillance reports and should be confirmed there.
Ask a room of physicians why a person’s blood carries oxygen and most will say hemoglobin. Ask why hemoglobin outside a red cell fails at the job, and the room grows quieter. The answer is a small molecule that most clinicians met once in physiology class and forgot: 2,3-bisphosphoglycerate, usually abbreviated BPG or, in older texts, DPG.
This essay is about that molecule, about its absence from a cow’s red cells, and about how that absence turned bovine hemoglobin into the raw material of choice for the oxygen carriers that reached the clinic. It is a story about allostery, the way a protein changes shape in response to something binding at a distant site, and it is one of the cleaner examples I know of nature doing a favor to an engineer without meaning to.
A red cell is a bag of hemoglobin at a concentration of roughly thirty-odd grams per deciliter of cell contents, together with the enzymes that keep the iron reduced and a bath of small molecules that tune the protein’s behavior. The most important tuner in humans is BPG. It binds in the central cavity of the deoxygenated hemoglobin tetramer, wedging between the two beta chains, and stabilizes the low-affinity form. The effect is to make hemoglobin release oxygen more readily. Without BPG the protein clings.
The numbers make the point. Hemoglobin inside a normal human red cell half-saturates with oxygen at a partial pressure, called the P50, of around 26 to 27 millimeters of mercury. Take the same protein out of the cell, put it in plasma, and the BPG washes away. The P50 falls to something like 12 to 14 millimeters of mercury. That is a hemoglobin that will load beautifully in the lung and then refuse to give its oxygen up until the surrounding tissue is nearly anoxic. For a carrier meant to feed starving tissue, this is a fatal design flaw.
There is a second problem. Outside the cell, the four-chain tetramer can split into two-chain halves. These dimers are small enough to slip through the kidney filter and out of the circulation, which shortens the useful life of the protein and delivers a burden of free hemoglobin to a delicate organ. So a successful carrier has to solve two problems at once: fix the oxygen affinity, and glue the pieces together.
The first generation of engineers attacked the two problems by brute chemical force applied to human hemoglobin.
One approach, pyridoxylation, attaches a small phosphate-bearing group to mimic the effect of BPG and lower the affinity. Another, the diaspirin cross-link used in the product known as HemAssist, bridges the two alpha chains with a chemical clamp, locking the tetramer together and, as a side effect, shifting the P50 up toward the thirties. A third, polymerization with glutaraldehyde, stitches many tetramers into larger assemblies, addressing the filtration problem, though it can also trap the protein in a fixed conformation.
These interventions worked, more or less, in the sense that the resulting molecules could hold together and give up oxygen. They also carried consequences. The chemical modifications were sometimes crude, and the final product was often a mixture. The trial history of the class, which I have discussed in other posts, includes serious safety signals, and it would be an oversimplification to blame any one design decision. But it is fair to say that starting from human hemoglobin meant starting from a protein whose native regulation had been taken away, and rebuilding it by hand.
Now look at the ruminants. In cattle red cells, the concentration of BPG is very low, far lower than in humans. Bovine hemoglobin has evolved to work in a red cell that does not depend on BPG to tune its affinity. Instead its oxygen affinity is regulated largely by chloride ions, which are abundant in plasma at a concentration of about a hundred millimolar, and by the ordinary effects of pH and carbon dioxide.
This is a startling fact when you first take it in. A protein that is tuned by chloride is a protein that is already tuned when you put it into plasma. Chloride does not wash away. It is the most common anion in extracellular fluid. Bovine hemoglobin dissolved in a chloride-containing solution therefore has a P50 in the neighborhood of thirty or forty millimeters of mercury, which is in a physiologically useful range, without any chemical addition.
Reviewers of the field have noted this advantage repeatedly: chloride rather than BPG regulates affinity, giving bovine hemoglobin an edge over the human protein in cell-free form. It is not the only reason the bovine protein was chosen. Cattle are plentiful, slaughterhouses offer a large and regulated supply of blood, and bovine hemoglobin has been reported to be more thermally stable than the human protein. But the allosteric difference is the most elegant one, because it means the protein arrives at the manufacturer already behaving like a decent oxygen transporter.
The chloride trick does not solve the dissociation problem. The bovine tetramer still splits into dimers in dilute solution, and dimers still leak through the kidney. The manufacturers of the two best-known bovine products solved it with glutaraldehyde, a small molecule with an aldehyde at each end that reacts with the amino groups of lysine residues on the protein surface.
Treated correctly, glutaraldehyde does two things. It forms bridges within a tetramer, holding it together, and it forms bridges between tetramers, building polymers. The resulting mixture contains molecules of many sizes. Both bovine products are named, in their formal nonproprietary names, for their average molecular mass: the product marketed for humans in some countries is hemoglobin glutamer-250 (bovine), and the veterinary product is hemoglobin glutamer-200 (bovine). The numbers refer, roughly, to average masses in kilodaltons. A single human tetramer is around 64 kilodaltons, so these products average three to four times heavier.
I emphasize the word average because it hides the interesting engineering question. A polymerization reaction is statistical. It does not produce a single molecular species. It produces a distribution, from unpolymerized tetramer up through dimers of tetramers, trimers and larger assemblies. The shape of that distribution matters enormously, and it is the reason process chemists in this field are obsessed with size-exclusion chromatography, the technique that sorts molecules by size and lets them read the distribution like a bar chart.
The reason is the two ends of the size curve. Molecules that are too small leak through the kidney and slip out of the circulation into the space beneath the endothelial lining of vessels, where free hemoglobin scavenges nitric oxide and provokes vessel tone changes. That is one mechanism proposed for the blood pressure elevation seen with earlier products. Molecules that are too large increase the viscosity of plasma and may behave differently in the microcirculation.
So a manufacturer wants to minimize the low end, and the specifications reflect that. The process aims to reduce unpolymerized tetramer and small oligomers to a low percentage through purification steps after the reaction, commonly ultrafiltration or chromatographic removal of the smallest species. I want to be careful here: the exact release specifications for the two bovine products are proprietary in part, and figures quoted in the literature vary. What can be said safely is that a lower fraction of low-molecular-weight material was a deliberate design goal, that the products were characterized by size-exclusion analysis, and that the distribution is a critical quality attribute.
A second critical attribute is the fraction of hemoglobin in the oxidized form, methemoglobin. Ferric hemoglobin cannot bind oxygen, and it also releases heme, which is toxic. A manufacturer measures it in every lot and sets a ceiling, because it rises as the product ages. The stability of the ferrous form during storage is a reason packaging and storage conditions are chosen with such care. The engineering of shelf life turns out to be an engineering of iron chemistry.
Now for a controversy that shows how much remains unsettled. Bovine glutamer products have a P50 in the thirties, roughly like or slightly above that of a person’s red cells. Is that ideal?
A school of thought associated with the microvascular hypothesis argues that it is not. The argument runs as follows. In a healthy circulation, oxygen leaves red cells while they are still in the arterioles, the small upstream vessels, and the arteriolar wall senses the oxygen level and constricts to protect downstream tissue from excess. A carrier with a low affinity, like a bovine glutamer, releases oxygen early, triggering that constriction reflex more strongly and reducing flow. A carrier with a high affinity, which holds oxygen until the last moment, would deliver it to the capillary bed with less premature loss, avoiding vasoconstriction and preserving the perfusion of small vessels. Under that logic, the best oxygen carrier for shock is one with a P50 well below that of blood.
The opposing view points out that a high-affinity carrier delivers oxygen only in tissue that is severely hypoxic, which is exactly where you need it in shock but is a poor design for resuscitating a patient who is anemic and otherwise stable. A carrier with a P50 in the thirties will offload in a broader range of conditions. The disagreement is not resolved. What is worth noticing is that the P50 of a bovine product is not an accident of chemistry alone; it is a physiological bet, and the bet can be argued either way.
For a foundation thinking about which indications to advocate for, this matters. A carrier with a mid-range P50 sits comfortably in an anemia niche, such as a patient who cannot receive blood. A carrier with a very high affinity might sit better in an ischemic niche, where the task is to reach the last few oxygen-starved cells. They are different jobs, and they may call for different molecules.
One more piece of regulation deserves mention, because it is a beautiful piece of physiology that a cell-free carrier keeps. Hemoglobin releases oxygen more readily when the surrounding tissue is acidic and rich in carbon dioxide. That is the Bohr effect. It is why exercising muscle, which makes acid and carbon dioxide, gets extra oxygen delivered exactly where it is needed.
The Bohr effect does not depend on BPG. It depends on the protonation of particular amino acid side chains and on carbamate formation at the ends of the chains. A cell-free bovine polymer retains it, at least in part, though polymerization can dampen the response by restricting the protein’s movement. This means that even outside a cell, the molecule can respond to local acidity by releasing more oxygen, which is the outcome you want in a tissue that is failing. The extent of the retained response is a measurable property, and it varies among products and among lots.
I mention it because it undermines a caricature. People sometimes describe a BHOC as a dumb sack of hemoglobin that can only hold and release oxygen according to one fixed curve. The truth is that it inherits much of the responsiveness of its parent protein, dampened by chemistry. That is a reason to take the design seriously, and a reason to measure it carefully.
Physiologists draw the relation between oxygen pressure and hemoglobin saturation as an S-shaped curve, and the P50 is simply the pressure at the curve’s midpoint. It is more useful to think of the curve as a map with three regions.
The upper flat region, above about sixty millimeters of mercury, is the lung. Here the curve barely rises with pressure, so hemoglobin loads to near-complete saturation even if the inspired air is thin. This flatness is a safety margin: a person on a mountain at reduced pressure still loads almost fully.
The steep middle region, between roughly twenty and sixty, is the working zone of the tissues. A small drop in pressure produces a large release of oxygen. A carrier’s usefulness depends on where its steep region lies relative to the pressure of the tissue it feeds. Move the curve to the left, as high-affinity molecules do, and the steep zone slides toward the lower pressures, so the molecule gives up its cargo only in the most oxygen-poor territory. Move it to the right and it gives up cargo earlier.
The lower region, below about twenty, is the reserve. A carrier with a curve that is too far to the left leaves much of its oxygen locked in this region, and no amount of extra hemoglobin will help a tissue sitting there. This is the arithmetic behind the early failure of unmodified human hemoglobin in plasma. It was not that the molecule carried too little oxygen, but that it declined to release it.
Seen this way, the choice between a bovine glutamer and a high-affinity engineered molecule is a choice about which zone to serve. And chloride, in the bovine case, simply put the curve in the middle of the working zone by default.
For readers who advocate for access, the chemistry converts into a short list of questions that separate a serious program from a hopeful one.
What is the P50 of the final product, measured in plasma-like conditions at body temperature, and how much does it vary from lot to lot?
What does the size-exclusion profile look like, and what fraction of the material sits at the small end?
What is the methemoglobin content at release and at the end of the labeled shelf life?
How much of the Bohr response survives polymerization?
None of these needs a doctorate to ask, and every serious manufacturer has the numbers.
It would be a mistake to leave the impression that the chloride story solves the problem. It solves one problem, the affinity problem, and only partially.
It does not solve nitric oxide scavenging, which depends on the availability of free hemoglobin to the vessel wall and on the size of the molecule. It does not solve auto-oxidation, the tendency of the iron to lose its electron and become useless. It does not solve immunogenicity, since bovine hemoglobin remains a foreign protein. And it does not solve the tension between a longer circulating half-life, which favors bigger polymers, and viscosity and vascular effects, which disfavor them.
What it does is remove a whole category of chemical modification from the manufacturing process. A process that does not have to attach anything to mimic BPG has fewer reactions, fewer side products and fewer opportunities for variability. This is an unglamorous benefit, and in pharmaceutical manufacturing the unglamorous benefits are frequently the ones that decide whether a product survives scale-up.
Every P50 quoted in a brochure is measured at body temperature, and every P50 changes when the temperature does. Cooling shifts the curve to the left: hemoglobin holds oxygen more tightly in the cold. For a patient in shock or a surgical patient cooled on purpose, this is a modest correction. For a donor organ resting in ice, it is a large one, and it means that a molecule tuned for warm blood behaves like a different molecule at four degrees. That is the subject of a later post, but the seed is here. The chloride trick sets the default. It does not make the curve immune to the thermometer.
Someone will ask why cattle, not pigs or horses. I do not have a definitive published answer, and I would be suspicious of anyone who claims one. The likely reasons are practical. Cattle produce a large volume of blood per animal, there is an established slaughter and inspection infrastructure, and the biochemistry of the bovine protein, with its chloride regulation, is convenient. Porcine hemoglobin does not offer the same allosteric advantage, as far as I know, and it introduces cultural and religious considerations of its own, discussed in another post on this site. The choice of species was probably made by a combination of chemistry and logistics, and both pointed toward the cow.
If you take one idea from this essay, take this. The success of the bovine approach was never a triumph of design so much as a gift of evolution accepted with gratitude. A cow’s red cell does not use BPG, so a cow’s hemoglobin is tuned by something that plasma already contains. Human hemoglobin needed to be repaired before it could be used outside its cell. Bovine hemoglobin needed only to be held together.
That distinction should shape how patient communities talk about the technology. The choice of a bovine source is not a curiosity of the marketplace. It has a chemical reason, and the reason is one a non-specialist can understand: the cow’s protein works in plasma because it never needed the molecule that plasma lacks.
The rest of the story, the questions of safety, of vessel tone, of oxidation and of immune response, remains open. The chloride, at least, is settled.
Reviews of hemoglobin-based oxygen carriers describe chloride regulation of bovine hemoglobin affinity, glutaraldehyde polymerization and the naming of the glutamer products; one accessible open-access review is titled Hemoglobin-Based Oxygen Carriers: Current State-of-the-Art and Novel Molecules. Numeric values for P50 and molecular weight distributions vary across sources and lots and should be verified against product documentation before use.
Medicine likes to tell the story of its failed products as a story about the products. The molecule was toxic, the trial was underpowered, the mechanism was misunderstood. Those explanations are sometimes correct. But they leave out a second story running in parallel, which is about the organizations that made the molecules. Companies have anatomies. They have capital structures, burn rates, single points of failure and moments at which a bad quarter turns into an existential one.
This essay reads the history of the oxygen-carrier industry as a series of corporate autopsies. I will not pretend to a level of detail that public sources do not support, and where I am uncertain I will say so. The aim is not gossip. The aim is to identify the patterns that repeat, because those patterns tell a patient community what has to be different this time if a BHOC is going to be available in five years and not merely promised.
Begin with the structure of the problem. A company that wants to sell an oxygen carrier for human use has to do four expensive things in sequence, and it has to pay for each before it earns a dollar.
First it has to build a manufacturing plant capable of producing a sterile injectable from an animal or recombinant source, at a scale that regulators will accept as representative of commercial supply. This is a capital project measured in tens of millions of dollars. Second, it has to run clinical trials in emergency and surgical settings, which are among the most difficult and costly trials in medicine, with sites that must screen patients in minutes and enroll them without consent. Third, it has to survive regulatory review, which for a first-of-class product with a troubled predecessor history is slow and unforgiving. Fourth, it has to build a market in hospitals that already have a working alternative in the blood bank.
The difficulty is that none of these steps is optional and none can be shortened by cleverness. The fixed costs are borne before there is a revenue stream to service them. In the financial vocabulary, the business model demands long duration equity and offers a binary outcome. Investors call this the valley of death, and it swallowed more than one company in this field.
The following sketches are drawn from public reporting and reviews. They are intentionally brief and I have kept to facts I could source.
Baxter and the diaspirin hemoglobin. The best-known early product from a major pharmaceutical firm was a human hemoglobin cross-linked between its alpha chains, sold under the developmental name HemAssist. A trauma trial published in the Journal of the American Medical Association in 1999 reported that 24 of the 52 patients given the product died within 28 days, 46 percent, against 8 of the 46 patients given saline, 17 percent. The program ended. What is notable is that this was a large, well-financed company with regulatory expertise, and the failure was scientific rather than financial. It set the tone for the entire class in the eyes of regulators and journal editors, and the memory of that trial is still invoked when new products are discussed, even though the molecule is chemically different from a bovine polymer.
Hemosol and the raffinose-linked hemoglobin. A Canadian company developed a human hemoglobin cross-linked with a sugar-derived reagent. According to the same review, the program advanced to late-stage trials and was discontinued in 2003 after patients receiving the product experienced adverse cardiac events. The company’s own finances deteriorated in the years that followed, and it entered insolvency proceedings in the middle of the decade. Here, financial death followed scientific death by a short interval, which is the typical pattern for a company whose entire value is a single product.
Northfield and the polymerized human hemoglobin. This Illinois company pursued the field’s most ambitious trial, the prehospital trauma study discussed elsewhere on this site, which used an exception from informed consent. The regulatory outcome was unfavorable and the pooled safety analysis published in 2008 hurt further. The company filed for bankruptcy in 2009. What distinguishes Northfield’s case is that it was bold. It attempted the hardest indication first, in a setting where the comparator was standard emergency care and the endpoint was mortality. If it had succeeded, it would have transformed emergency medicine. It did not, and there was no second product to fall back on.
Biopure and the bovine glutamers. The Massachusetts company that developed the two bovine products, one for dogs and one for people, is the most relevant case here, and its history is documented in some detail. It was founded in 1984 and began by sourcing bovine blood from regional slaughterhouses. In 1999 the Food and Drug Administration and the European Commission approved the veterinary product for anemic dogs. In April 2001 South Africa approved the human product for surgical anemia. It was also approved in Russia. But the American regulator placed a hold on a proposed trauma study in April 2003 and rejected an application for use in orthopedic surgery in July of that year. Later that year the stock fell sharply after the Securities and Exchange Commission found misleading statements about the regulator’s stance, and in 2005 the commission brought charges against the company’s chief executive and other officers. In 2008 the pooled safety analysis appeared. On July 16, 2009 the company filed for Chapter 11 bankruptcy protection, and its assets were bought by a private entity that September.
The successor entities. The assets passed through more than one owner. By 2014 the human product was being made available through the regulator’s expanded access mechanism, and in 2022 a successor company filed a registration statement seeking about 30 million dollars in an initial public offering, drawing a skeptical response from at least one financial journalist who called the attempt a long shot. I do not know how that filing ended, and I would rather say so than guess.
Read the five stories side by side and five patterns emerge.
The single-product company. Every one of these organizations was built around one molecule. When the molecule stumbled, there was nothing to absorb the shock. Big pharmaceutical companies that took on hemoglobin projects, like Baxter, could shelve the effort and continue. A small company cannot. The lesson is not that small companies should not try, but that a small company needs a structure that separates its survival from any single trial result. Options include a platform with several indications, a diversified funding base, or a partnership in which a larger organization shares the downside.
Judgment by class. When one product harmed patients, regulators, reviewers and journalists learned to treat the whole class with suspicion. The pooled analysis in 2008 was an example: it combined trials of several different molecules and produced a single alarming number. That is a fair scientific act, because the products shared a mechanism of concern. It was also a commercial catastrophe for the companies whose products differed. A bovine polymer is not chemically the same as a cross-linked human dimer. But a company cannot argue its way out of a class reputation in a single meeting. It has to generate its own data, patiently, over years, and it has to be able to afford the years.
Improving comparators. The blood supply became safer over the same period, in large part because of the diseases that first motivated the search for substitutes. Screening for viruses improved, leukoreduction became routine, and transfusion practice grew more restrained. Every improvement raised the bar for a product whose main pitch was safety from infectious disease. A product designed in the 1990s to offer protection against contamination that then became rare had a shrinking advantage.
Governance failures. The Biopure case includes an element that no chemistry explains. The commission’s findings about misleading statements to investors and the criminal charges against officers destroyed trust in the company’s public communications. It is difficult to overstate how corrosive this is. In a field where the science is already contested, the credibility of the sponsor becomes a scientific variable, because reviewers must decide how far to trust the sponsor’s reading of its own data. A company that has been caught shading its statements pays a tax on every subsequent claim.
Survival through niches. And here is the most useful pattern of all. The bovine product did not vanish. It survived, in a battered form, because three small niches kept it alive: a veterinary market that valued it, a handful of countries that had approved it, and a compassionate access route for patients who could not receive blood. None of these markets could support a company on its own, but together they preserved the manufacturing know-how, the regulatory dossier and a trickle of clinical experience. When a new owner acquired the assets, it acquired something that a start-up could not have built: a product that had already been made, shipped and used. The lesson for planners is that niches are not a consolation prize. They are scaffolding. A niche that pays even modestly keeps a validated process running, keeps trained people employed and keeps a pharmacovigilance record accumulating, and those three things are the hardest to rebuild once lost. An ephemeral company can leave behind an enduring capability, provided someone is diligent enough to preserve it through the wreckage.
Suppose a patient organization were asked to design the funding and governance structure for a BHOC program that could survive the next fifteen years. What would it insist on?
I would begin with the customer. The natural early customers for an oxygen carrier are not consumers or even hospitals, but institutions that carry responsibility for people in places where blood cannot follow. Military medical commands, remote-region health services, disaster response agencies and organizations serving people who refuse blood on grounds of conscience all fit. Such buyers can fund development on contract, purchase in bulk and value a shelf-stable product for reasons the blood bank market does not. Because they pay before the commercial market matures, they reduce the valley.
Next comes the regulatory strategy. A rare-disease or orphan designation, where it applies, offers incentives that improve the economics of a small market. Expanded access, used carefully, generates a record of real-world use and can be a modest revenue source, though a regulator will not allow it to substitute for trials. Foreign approvals, which have some precedent for this product, can open smaller markets and produce clinical data, provided their standards of evidence are transparent.
Third is manufacturing. A plant that can only make one product at one scale is a fragile asset. Modular, smaller-scale facilities that can produce clinical lots and low-volume commercial supply reduce the capital at risk. I do not know which developers are pursuing this route, and I would ask them.
Fourth is data discipline. The lasting damage to the field came from unfulfilled promises and disputed interpretations. A program that pre-registers its trials, publishes negative results and invites independent analysis builds a kind of credit that money cannot buy. This is not idealism. It is a strategy for lowering the tax on every future claim.
Fifth is governance. A company whose communications have been audited by an independent committee, whose disclosures are conservative and whose leadership does not overstate its regulatory position offers an investor something rare in this sector: predictability. A patient organization has leverage here. It can decline to lend its name to promotional claims that go beyond the data, and it can say so publicly.
A worked example makes the structure vivid. The figures below are invented for illustration and do not describe any real company, but the shape is faithful to what the histories show.
Imagine a developer that needs 40 million dollars for a pilot plant, 25 million for two pivotal trials, and 8 million a year in overhead for five years before a decision. The total before any revenue is roughly 105 million dollars. Suppose the chance that the first trial succeeds is one in three, that regulators approve given success is one in two, and that the first year of sales is small. The expected value of the enterprise at the start is dominated by a one-in-six event. An investor who understands that will demand a return of many times the money, and a founder who understands it will be tempted to shade the probabilities.
Now add the peculiar feature of this field. A failure of one trial does not merely lose that trial’s cost. It changes the prior for the whole class, so the probability of success for the next company falls too. In economic language, each failure imposes a negative externality on competitors. That is why capital in this sector has come in bursts and then vanished for years: the informational damage of a failure lingers long after the balance sheet is cleaned up.
The remedy is not obvious, but one piece is: reduce the amount that has to be raised at the riskiest point. A program that can reach an early human data point cheaply, through a small compassionate-use series or an ex vivo organ perfusion study where the drug never enters the patient’s bloodstream, spends a few million dollars to learn something that would otherwise cost tens of millions. Cheap early information is the most valuable asset a small developer can buy.
It is easy to write an autopsy after the death. The harder and more useful task is to ask what an outside observer, such as a patient organization, could have noticed while the patient was still alive.
Watch the ratio of announcements to publications. A developer that issues frequent press statements and rarely publishes peer-reviewed data is telling you something about where its evidence lives.
Watch the language used about the regulator. Confident statements about what an agency has said or will say, unsupported by documents, are a warning. The Biopure episode with the securities regulator turned on exactly that kind of statement.
Watch the trial design. A first trial that tries to prove survival benefit in the most severe emergency setting, without earlier smaller studies, is an all-or-nothing wager. It may be brave. It is also how a company puts its whole future on a single roll.
Watch the manufacturing. If the company cannot describe its process, its lot release criteria and its supply redundancy in plain terms, its plant is either immature or its story is.
Watch the cash runway. A company with fewer than eighteen months of funding and a trial that will read out in twenty-four is a company that will negotiate its next financing from a position of weakness.
None of these indicators is proof of failure. Together they are a checklist that an advocacy group with access to public filings can apply, and it is a checklist that would have raised flags in several of the histories above.
There is a temptation for advocates to see the industry’s failures as an argument for more urgency and less caution. I think the history says the opposite. Every failure in this field included a moment of enthusiasm that outran the evidence: the early trial that was expanded before the signal was understood, the announcement that reached investors before regulators, the confidence that a favorable meeting meant approval.
What advocacy can offer is something the market underprices: patient legitimacy. A trial that patients trust enrolls faster and drops out less. A program with a credible community voice is more likely to obtain the community consultation that emergency research demands. An organization that has read the history and can say what went wrong last time is better placed to make sure it does not happen again.
It can also do something quieter. It can keep the record. The reason I can write this essay at all is that other people, journalists, regulators, reviewers, took the trouble to document how the companies rose and fell. Without that record, each new generation of developers would rediscover the same mistakes at the same cost. A patient organization that maintains a plain-language archive of what happened, product by product, is doing work the industry has no incentive to do.
Five companies, three bankruptcies or insolvencies among them by my count, and one cow protein that limped on. If you had to summarize the pattern in a sentence, you could say that this field has been unusually hard on its organizations and, oddly, kinder to its molecules. The bovine glutamer outlived the company that made it.
That should reassure and warn in equal measure. It reassures because it suggests the molecule is not the weak point; a product that keeps finding a use after its sponsor collapses is doing something worth doing. It warns because a molecule without a durable steward has no future, however good it is. The next chapter will be decided less by chemistry than by whether the people who care about it, including the patients, insist on an organization worthy of the molecule.
Company histories are drawn from public sources including reviews of blood substitutes, the published account of Biopure and its assets, and news reporting on the 2022 registration filing by its successor. The 2008 pooled safety analysis appeared in the Journal of the American Medical Association. Specific figures, such as the 46 and 17 percent mortality figures for the diaspirin-linked product, come from a published historical review and should be checked against the original trial reports.
There is a product in the oxygen-carrier family whose central design feature will sound, on first hearing, like a mistake. It is a bovine hemoglobin, chemically wrapped in polyethylene glycol, and it is manufactured and stored with carbon monoxide already bound to its iron. Carbon monoxide is the gas that kills people in garages and closed rooms precisely because it takes the place of oxygen on hemoglobin. A medicine that begins by filling every binding site with carbon monoxide seems to be built backward.
The product is known as Sanguinate, a PEGylated carboxyhemoglobin from bovine blood, developed by Prolong Pharmaceuticals. Reading how its developers justify the design is one of the more instructive exercises in this field, because it forces a rethinking of what an oxygen carrier is for. This essay walks through the argument, states clearly which parts are supported and which are hopeful, and draws the lessons that matter for patients following the wider BHOC story.
Start with a fact that surprises many people. Your body makes carbon monoxide continuously. An enzyme called heme oxygenase breaks down heme, the iron-containing ring from old hemoglobin, and one of the products is carbon monoxide, along with iron and a green-yellow pigment called biliverdin. This is not a rare or pathological event. Every day a healthy adult breaks down a substantial quantity of hemoglobin and generates a small, steady flow of the gas.
Over the past few decades, physiologists have shown that this endogenous carbon monoxide behaves as a signaling molecule. It relaxes blood vessels, dampens certain inflammatory responses, and influences how cells handle stress. It belongs to a small group of gaseous transmitters, sometimes called gasotransmitters, along with nitric oxide and hydrogen sulfide. The three share a history: each was known first as a poison, and each was later found to be a messenger the body makes on purpose. Readers of another post in this series will recognize hydrogen sulfide from the deep-sea tubeworm, which uses a protein to carry it safely.
The step from that science to a drug is a small one conceptually. If carbon monoxide at low doses can calm inflamed, oxygen-starved tissue, then a delivery system for small, controlled amounts of the gas might help. Chemists developed several classes of carbon monoxide-releasing molecules to test the idea. The Sanguinate design takes a different route: it uses hemoglobin, the body’s own carbon monoxide carrier, as the delivery vehicle.
There is a second, more prosaic reason to load hemoglobin with carbon monoxide, and it is pure chemistry. Free hemoglobin is fragile. In the oxygenated state it slowly loses an electron from its iron and turns into methemoglobin, which cannot carry oxygen and which sheds toxic heme. The oxygenated form is the most susceptible. In the deoxygenated state the iron is reactive in other ways, including toward nitric oxide.
Carbon monoxide bound to hemoglobin protects the iron. The carboxy form is reported to be considerably more resistant to oxidation than the oxy form, because carbon monoxide binds tightly and holds the iron in its ferrous state. A manufacturer who stores the product as carboxyhemoglobin therefore has a molecule that stays useful on the shelf. When the product is infused, the carbon monoxide is gradually displaced by oxygen in the lung and in tissue, since oxygen at high concentration can compete with carbon monoxide, and the released carbon monoxide can act as a signal. The gas is at once a preservative and a payload.
Those are two independent justifications, one from stability chemistry and one from pharmacology, that point to the same molecule. When two independent arguments converge, it is worth paying attention, though convergence is not proof that either is right.
The second element of the design is the polyethylene glycol. PEGylation attaches long, water-loving chains to the protein surface. It has been used for decades to extend the circulation time of protein drugs and to shield them from immune recognition. In the context of a hemoglobin carrier, the chains do several things at once. They enlarge the molecule’s effective size in solution, which reduces leakage through the kidney. They add a hydration shell that increases viscosity and colloid osmotic pressure without requiring a high concentration of protein. And, as the developers argue, they lower the molecule’s tendency to trigger vasoconstriction, which in earlier products was linked to the scavenging of nitric oxide in the vessel wall.
PEGylation also changes oxygen affinity. Reports on this class of PEGylated hemoglobins describe a high-affinity carrier, with a P50 well below that of normal blood. That is the opposite of the bovine glutamer products discussed in another post on this site. A high-affinity design is consistent with the microvascular hypothesis: a carrier that holds oxygen tightly delivers it only where tissue is severely hypoxic, and does not trigger arteriolar constriction upstream. The developers presented this as an advantage in conditions such as sickle cell crisis and ischemia, where blood vessels have narrowed and red cells cannot reach the tissue.
It helps to line up the three gasotransmitters against the single protein that meets all of them, because the comparison shows why hemoglobin is both the problem and the opportunity.
Nitric oxide is made by the vessel lining and diffuses a short distance to the smooth muscle around it, where it relaxes the muscle and widens the vessel. Free hemoglobin in plasma is an avid scavenger of nitric oxide, which is the root of the vasoconstriction that undermined earlier carriers. The problem is that hemoglobin removes a gas the body wants to keep.
Carbon monoxide is made in small amounts by heme breakdown, binds hemoglobin about two hundred times more tightly than oxygen does, and relaxes vessels by a different route, through the same enzyme that nitric oxide activates and through potassium channels. Here the protein holds a gas that the body may want to release. Loading it in advance turns the scavenger into a depot.
Hydrogen sulfide is made by several enzymes, is toxic at high levels, and in the tubeworm is carried on zinc sites of the protein without harming the heme. It shows that a hemoglobin can be engineered, by evolution, to carry a second gas safely.
Seen together, the three gases suggest a design vocabulary. A future carrier might scavenge less nitric oxide, deliver a dose of carbon monoxide, and shield its iron from sulfide. That is speculative. It is also a coherent research program, and the PEGylated carboxy product is its first tangible example.
Because the safety argument depends on dose, a numerical thought experiment is helpful. The numbers below are illustrative only and are not drawn from any trial.
Suppose a patient has five liters of blood with a total hemoglobin of ten grams per deciliter, so fifty grams circulating. Suppose an infusion adds three grams of a carboxy carrier, six percent of the total. If every molecule of the carrier were saturated with carbon monoxide, then in the first moments the carboxy fraction of the patient’s total hemoglobin would rise by about six percentage points from the carrier alone. In practice the carbon monoxide begins to exchange almost at once with oxygen in the lung, so the peak would be lower and the fraction would decline over hours.
Compare this with a heavy smoker, who can have baseline carboxyhemoglobin of five to ten percent, and with a person in a garage, in whom levels above twenty or thirty percent produce headache, confusion and collapse. Under this admittedly rough arithmetic, a small dose sits in the range that a smoker tolerates daily, and a very large one would approach the range of clinical concern. That is why the protocol details matter, and why the honest summary has to include the phrase depends on the dose.
The reason to go through this is not to reassure. It is to show the kind of reasoning a reader can apply to any claim about safety. Ask for the dose, ask what fraction of hemoglobin it represents, and ask how that compares with a familiar exposure. Numbers turn a scary word into a quantity.
The published descriptions of the product’s development list a set of targets that reads like a map of the places where oxygen fails locally rather than globally. The company obtained an orphan drug designation from the United States regulator for sickle cell disease. The designation was reported to be based on what the manufacturer described as the product’s ability to reverse sickling in laboratory settings, and it was cited in connection with complications including vaso-occlusive crisis, acute chest syndrome, leg ulcers and stroke. Trials were reported in other oxygen-deprivation conditions, including delayed cerebral ischemia after bleeding around the brain and delayed function in kidney grafts from donors. Early trials were also reported in beta-thalassemia outside the United States.
There is also the use that made the product visible outside the specialty literature: emergency compassionate use for patients with life-threatening anemia who refused blood. These single-patient cases, like others in this field, generated case reports rather than controlled data.
I want to be exact about what I know and do not know. I have read the developers’ summary papers and the regulator’s public designation notice. I have not seen complete, peer-reviewed results of a randomized trial showing that the product improves a clinical outcome in any of the listed conditions, and I will not imply otherwise. Where a phase 2 trial has been run, the responsible response is to consult the trial report itself. My purpose in this essay is to examine the logic of the design, not to certify its success.
Sickle cell disease is worth a moment because its logic is a good test of the whole idea. In sickle cell disease a mutation in the beta chain makes deoxygenated hemoglobin polymerize inside the red cell, deforming it into a rigid crescent that jams small vessels. The jam causes a local oxygen deficit, which promotes more deoxygenation, which promotes more sickling. It is a feedback loop.
The argument for a cell-free, high-affinity carrier is that it can slip past the blockage. A red cell is a large, stiff object when sickled and cannot squeeze through a partly closed vessel. A dissolved polymer or a PEGylated protein moves with the plasma, and plasma still flows through narrow channels that cells cannot enter. If the protein delivers even a little oxygen to the tissue beyond the blockage, it might interrupt the loop. The addition of carbon monoxide may contribute an anti-inflammatory and vessel-relaxing effect that eases the vessel narrowing.
This is a beautiful hypothesis. It is also a hypothesis, and a hard one to test, because sickle cell crises are unpredictable, patient reports of pain are subjective, and the natural course of a crisis is to improve on its own. Trials in this area are difficult to design and easy to misread. A modest positive signal in a small trial may be noise. A negative one may reflect the wrong dose or the wrong timing. That does not make the idea wrong. It means the honest summary is: interesting mechanism, limited proof.
Anyone who has read the earlier post on measurement will see a difficulty at once. If a product is loaded with carbon monoxide, the instruments we use to read oxygenation become unreliable in a specific way.
A standard finger pulse oximeter, which uses two wavelengths of light, cannot distinguish oxyhemoglobin from carboxyhemoglobin; both absorb red light similarly, so it reads a falsely reassuring saturation when carbon monoxide is present. This is the same reason the instrument fails in real carbon monoxide poisoning. A laboratory co-oximeter, which measures at many wavelengths, can separate the two species, and so any trial of a carboxyhemoglobin product has to specify that arterial samples be analyzed on a co-oximeter and that finger readings not be relied on.
Here is the rub. The total carboxyhemoglobin level in the blood after an infusion of a carboxyhemoglobin product is a mixture of the product’s own carbon monoxide and the patient’s native levels, and the interpretation depends on the dose. A small dose adds a fraction of a percent to the total carboxyhemoglobin fraction. A large one could raise it to a level that in a smoker or a poisoned patient would prompt concern. A careful protocol therefore has to define dose limits in terms of the carbon monoxide delivered, monitor carboxyhemoglobin by co-oximetry, and educate staff to disregard the pulse oximeter’s number. Any of these steps can fail in a busy emergency department, and that is a real operational hazard.
Is it safe to give carbon monoxide to a patient? The pharmacologist’s answer is that dose determines the poison, as with most drugs. Carbon monoxide is dangerous when it displaces a large fraction of the body’s oxygen-carrying capacity, typically levels of carboxyhemoglobin well into the double digits. The amount delivered by a modest infusion of a carboxyhemoglobin carrier is a fraction of that, and it is diluted into a patient who is receiving less than a full replacement of their hemoglobin.
But the pharmacologist’s answer is incomplete for three reasons.
First, the patients most likely to receive the product are already compromised. A sickle cell patient in crisis, a patient with a brain bleed or a patient with a failing organ may have less reserve to tolerate any reduction in usable hemoglobin. Second, carbon monoxide binds not only to hemoglobin but also to other heme proteins, including cytochromes in mitochondria, and its net effect on cellular respiration depends on dose in ways that are still being mapped. Third, carbon monoxide has a long history as an occupational and environmental hazard, and the public and clinicians carry a strong prior against it. A hospital pharmacy committee may struggle to approve a drug whose active ingredient is a poison by common understanding, even if the pharmacology is sound.
None of these objections is decisive. All of them are the kind of question a regulatory reviewer would ask, and the public documents suggest the developers took them seriously by monitoring carboxyhemoglobin levels in their early trials. I will simply state that a careful reader should treat the safety case as plausible but incompletely established.
Step back from the product itself, because I think its importance is broader than any one program.
The mainstream story of oxygen carriers is a story about substitution: a molecule replaces a red cell, and success is measured by whether it carries as much oxygen as blood. The carbon monoxide design proposes a different category. The molecule is a dual-purpose object, an oxygen transporter that is also a drug. It exploits the fact that hemoglobin is the natural carrier of gaseous signaling molecules as well as of oxygen, and it treats that as a feature.
If the idea is right, it opens a path for the whole family. A carrier could be loaded with a therapeutic amount of a gas, or with nitric oxide in a protected form to counter the scavenging problem, or with a stabilizing ligand that extends shelf life. The bovine glutamers discussed in another post carry oxygen and nothing else. A future generation might carry oxygen and a dose of signal.
If the idea is wrong, it teaches something too. It would mean the vessel-relaxing and anti-inflammatory actions of carbon monoxide at these doses are too weak to matter, and that the protective value of the PEG shield and the high affinity did the work. Either way, the comparison of PEGylated carboxy products with unmodified bovine polymers offers a natural experiment: two designs with related chemistry and very different philosophies. A well-organized field would design its trials to tell them apart.
The lesson for advocates is largely about vocabulary. When someone tells you a product is a bovine hemoglobin oxygen carrier, ask which kind. The category contains molecules that differ in size, in oxygen affinity, in what they carry besides oxygen, and in what disease they aim to treat. A bovine glutamer of average mass around two hundred fifty kilodaltons with a mid-range affinity is not the same object as a PEGylated carboxyhemoglobin with a high affinity. Lumping them together, as public discussion often does, makes evidence for one seem to apply to the other, and makes the failure of one seem to condemn both.
It is equally important to be honest about the difference between mechanism and outcome. A design can be clever and still fail. The history of the field is littered with clever designs. What earns trust is a patient, transparent record of trial results, and until that record exists the appropriate stance toward any of these products is curiosity, not conviction.
The poison bound to that iron is not a paradox. It is a proposition, stated in chemistry: that the body’s oldest toxin, given in a controlled dose and carried by its own transport protein, might be an ally. The proposition has been made. The test is the trial data.
The developers’ summary of mechanism and clinical experience appears in an Artificial Organs paper from 2017 on PEGylated carboxyhemoglobin bovine. The regulator’s orphan designation for sickle cell disease was reported in 2015. Background on carbon monoxide as an endogenous signaling molecule, produced by heme oxygenase, is available in general physiology and pharmacology reviews. Readers should consult trial registries and peer-reviewed reports for current clinical results, which I have not attempted to summarize.
The standard pitch for an oxygen carrier goes like this: blood is scarce, perishable and dangerous, so give the patient something better. It is a pitch that made sense in 1985, when the blood supply had been shaken by a new virus and by a shortage of donors in a war zone. In 2026 it needs a more careful hearing, because the thing it claims to beat has changed. Blood in the field is getting simpler, closer to the patient and better studied. The competitor is no longer a hospital blood bank a helicopter ride away. It is increasingly a cooler in the back of an ambulance.
This essay is a stress test. It asks, with as little wishful thinking as possible, where a BHOC still has a defensible place given what the simpler alternatives can now do. I believe the answer is real and narrower than promoters suggest, and that a patient organization gains more from stating the narrow answer plainly than from repeating a broad one that experienced clinicians will pick apart.
In the wars of the early twenty-first century, military medicine relearned an old lesson. Patients who die of hemorrhage die early, often within the first hour, and the intervention that changes their odds is giving blood products promptly and in something like the proportions in which they were lost. Whole blood, which contains red cells, plasma and platelets together, fits that principle naturally.
Two developments then moved whole blood from the battlefield to the civilian ambulance. First, the concept of low-titer group O whole blood took hold. Group O red cells can be given to almost anyone. Group O plasma contains antibodies against A and B antigens, which can harm recipients of other groups, but if the antibody level is low the risk is small. Blood banks began screening group O donors for low antibody titers, with a common threshold in the trial described below being a titer below 256, and issuing that blood as a near-universal product. Second, small teams in helicopters and ground ambulances began carrying it, along with portable refrigerators.
Meanwhile, plasma moved forward on a parallel track. A large randomized trial of plasma given during air medical transport, known as PAMPer and published in 2018, found lower mortality at 30 days in trauma patients who received plasma early. In the trial’s control group, mortality was about 33 percent, and in the plasma group roughly 23 percent. A post hoc analysis in a later paper suggested that the benefit was concentrated in patients whose transport time was longer than about 20 minutes. Dried plasma, which can be stored at room temperature and reconstituted with sterile water, offered a way to bring the same principle to places without refrigeration.
The most recent large study of prehospital whole blood, called TOWAR, was published in the New England Journal of Medicine in May 2026. It enrolled 1,020 patients across 44 air medical bases and 11 trauma centers, between May 2022 and June 2025, using a design in which bases were randomly assigned, two to one, to carry whole blood or the usual component therapy. The primary outcome was death within 30 days.
The result surprised many people who expected a benefit. According to the published summary, mortality at 30 days was 25.9 percent in the whole blood group and 20.5 percent in the component group, an adjusted odds ratio of 1.24 with a confidence interval running from 0.87 to 1.76, and a p-value of 0.24. In plain language, the trial did not show that whole blood helped, and the point estimate leaned the other way, though the uncertainty was wide enough that a benefit or a harm remained possible. The authors noted limitations: crossover between arms, variable volumes, an inability to mask the clinicians and limited data on coagulation. They also noted that overall mortality in the trial, 24.3 percent, was lower than in the earlier trial, which suggests that modern trauma systems have become better at controlling bleeding and delivering blood quickly.
I mention this trial for two reasons. First, it should humble anyone, including a BHOC advocate, who assumes that a physiologically attractive product will beat the alternative. Whole blood is about as physiologically attractive as a product can be, and in a rigorous test it did not win. Second, it shows what the bar looks like. A modern trauma system, delivering component therapy quickly, achieves a mortality figure around a fifth to a quarter in a severely injured population. A new product that hopes to demonstrate a mortality advantage in the same setting faces a very large sample size and a very narrow margin.
If the civilian trauma pathway is increasingly well served by blood, where is the gap? I would organize the answer around three functions that human blood products, however good, cannot perform. These are the defensible territory for a BHOC.
The first is to reach patients who will not accept blood. For patients whose conscience forbids transfusion, blood products are not competitors at all. They are unavailable by the patient’s own decision. For those patients the relevant comparison is between a BHOC and no oxygen-carrying support other than what their own bodies can manage, which may mean life-threatening anemia treated with iron, erythropoietin, oxygen and hope. This is the setting in which emergency compassionate access has been most often sought, and it is the one where the ethical case is strongest. The population is small, but the need in each case is acute.
The second is to reach patients who cannot be matched. Some patients have developed antibodies against many red cell antigens, through repeated transfusions or pregnancies, so that compatible blood is hard or impossible to find. Some have rare blood types. In such a patient, a product with no red cell antigens and no need for cross-matching is not a marginal convenience but the only feasible option. The number of such patients is again modest, and their situation often unfolds slowly enough that the search for compatible blood can continue while a bridge is given.
The third is to survive where the cold chain cannot. Whole blood is stored at refrigerator temperature, and in the trial discussed above the shelf life was up to 21 days. Dried plasma travels better, but it supplies clotting factors and volume without red cells. Neither delivers oxygen-carrying capacity in a stable room-temperature form. A BHOC, on the other hand, has been described in product literature as storable at room temperature for years. There are places, remote clinics, disaster zones, small vessels at sea, long-range military patrols, where a cooler with a monitored temperature is a logistical burden that cannot be met. For those places, a shelf-stable oxygen carrier is not a substitute for blood so much as a way to keep a patient alive until blood can be reached.
Logistics is a matter of grams and degrees, and a short calculation shows why the room-temperature niche is more than sentiment. The figures are illustrative round numbers and do not describe any specific product or unit.
A medic on foot can carry perhaps a few kilograms of resuscitation supplies among everything else. A unit of refrigerated whole blood weighs roughly half a kilogram before packaging, and the insulated container and cold packs required to keep it between one and six degrees Celsius can multiply that several times over. The container must be opened rarely, monitored, and replaced when the ice runs out. A patrol that lasts three days in a hot climate has to plan for the ice.
A bag of a room-temperature carrier of comparable volume weighs about the same as the blood, without the container, and can ride in a pack beside the bandages. It also has no expiry that arrives in three weeks. If the patrol does not need it, it comes home and goes back on a shelf. Whole blood that was not used may be wasted or returned with its clock running down.
None of this makes the carrier better clinically. It makes it lighter and more forgiving, and in logistics that is sometimes the entire argument. Military planners talk about the burden of weight and the burden of refrigeration in the same breath, and any advocate who wants to be taken seriously by them will speak in those units.
It is instructive to imagine how a hospital blood bank director hears the BHOC pitch. Her whole professional life is organized around the safety of a product that is difficult to keep safe: matching, screening, storage, expiry, tracing. She is proud of the record, and she should be. She hears a proposal to add a new product to the shelf, made from another species, with a troubled history and a large question mark over its cardiac safety.
What would persuade her? Not a claim about the future of medicine. She would want four things. A defined list of patients for whom she can use it, so that her staff have a protocol and not a judgment call. A monitoring plan for the laboratory tests that a plasma pigment will disturb, so that chemistry results are not misread. A clear statement of who owns the decision to use it, and who documents consent. And a supply that does not run out, since a product that vanishes after she has written it into her protocols is worse than no product.
This is a useful test for a patient organization: if you cannot answer those four questions in a page, you are not ready to ask a blood bank director for anything. If you can, you are speaking her language, and she is likely to listen. Blood bankers are, for the most part, people who would like their patients to have more options, provided the options are safe and the paperwork is honest.
It is just as useful to name the claims that a fair-minded clinician will not accept.
The claim that a BHOC can replace blood in general trauma is unsupported. The history of the field, including the pooled analysis published in 2008, gives reasons to be cautious, and the modern comparator has improved.
The claim that a BHOC is safer than blood because it lacks infectious risk is weaker than it was. The modern blood supply is screened for the known viruses, and the residual risks are very small. The bovine product introduces its own risks, addressed in another post in this series.
The claim that a BHOC is a universal solution to blood shortage is unrealistic. Shortages are real, but they are chronic and regional, and the oxygen-carrying capacity of a few units of a carrier is a small fraction of what a bleeding patient needs. A carrier has a half-life measured in hours to a day, and it does not replace clotting factors or platelets. A patient bleeding from a torn vessel needs the bleeding controlled and clotting support, and a BHOC provides neither.
That last point is worth dwelling on, because it changes the framing. Hemorrhage is a problem with three parts: stop the bleeding, restore clotting, restore oxygen delivery. The literature of the past two decades has shown that the first two have the greatest effect on survival, and that blood products address all three at once. A BHOC addresses only the third. It is a component, and a good use for it is as a bridge alongside other measures.
Here is a different way to see it. Suppose that the future of forward resuscitation lies in a combination of products chosen for their logistical properties. Dried plasma provides clotting factors and volume. A hemostatic agent controls bleeding. Whole blood, where cold chain exists, provides everything. And a shelf-stable oxygen carrier provides the oxygen-carrying capacity that plasma lacks when blood is unavailable.
The Army has published thinking about a lyophilized product strategy along these lines, combining freeze-dried plasma with a freeze-dried red cell or platelet component. That work is aimed at closing the gap between what a medic can carry and what a patient needs. In that architecture a BHOC is one of several possible oxygen components, and the question is whether it wins on weight, stability, cost and safety against the other candidates.
Stated that way, the comparison is more tractable and more honest. A BHOC does not have to beat blood. It has to beat the nearest alternative in a specific place. In a helicopter with a fridge, the alternative is blood. In a village three days’ walk from a road, the alternative is nothing.
If a patient organization wanted to press the case in the sensible territory, what evidence would it ask for?
For the conscience-based population, a registry of every emergency access case, recorded consistently, with outcomes at discharge and follow-up, safety events, dose, hemoglobin measured both in plasma and in red cells, and the reasons for choosing the product. These cases are already occurring. Without a common form, the information is scattered across case reports.
For the unmatched population, a similar registry, together with a defined pathway under which a hospital blood bank can request the product when compatible blood cannot be found within a set time.
For the cold-chain-limited setting, a field-feasibility study: stock the product at remote sites, train staff, track storage conditions and use, and measure whether it is used appropriately and whether patients survive to reach definitive care. This is not a randomized trial, but it produces something regulators and funders need, which is proof that the logistics work.
For each, the outcome should be defined ahead of time and the comparator should be honest. Where the comparator is no oxygen-carrying therapy, a historical or registry control is acceptable if openly described. Where blood is available, the comparison should be to blood.
One more distinction protects everyone from confusion. Saying that a BHOC belongs where blood cannot go is different from saying that blood should not be used where it can. A patient who is bleeding in a well-equipped hospital should receive what the best evidence supports, which is blood in the right proportions. Nobody serves the field by suggesting otherwise, and nobody in the field who is honest with themselves believes it.
There is a temptation to answer the whole-blood trial by dismissing it, arguing that it was underpowered or that the population differed. The sounder response is to accept it and adjust. A field that has spent decades trying to beat blood in a general population has spent decades on the hardest version of the question. The easier version, and the one in which a BHOC has a clear reason to exist, is to reach the people that blood cannot reach.
That reframing also improves the ethics. A product aimed at patients who have no alternative has a favorable risk-benefit calculation even if its safety is imperfect, because the alternative is worse. A product aimed at patients who could receive blood must clear a higher bar. Being clear about which is which protects patients and protects the credibility of the advocates.
It also protects the science. If the only trials ever run are large trials against blood in trauma, the likely result is another disappointment, and the class will suffer for it. If the trials are designed around the populations with no alternative, the likely result is a clear signal, one way or the other, at a manageable sample size.
Imagine that a trauma surgeon who has read the recent trial asks why anyone still cares about a cow protein. A fair answer might go like this. You are right that in your system, with a helicopter, a blood bank and a cooler, the case for a carrier is weak. But you do not treat all the patients. Some refuse blood, some cannot be matched and some are in places your system does not reach. For those patients your options are thinner than you would like. A stable, unmatched, room-temperature product that keeps someone alive until blood is possible is worth having. I am not asking you to replace your cooler. I am asking you not to let the better answer for most patients become the enemy of a workable answer for the rest.
I think a surgeon might accept that. I think many would.
The simpler competitor has raised the bar, and in doing so it has clarified the mission. The story of the last fifteen years is not that oxygen carriers lost. It is that blood got better, and the space that remained is the space where blood cannot go. That space is smaller than the promoters once imagined, and it is real. Every person in it has a name and a family, and every one of them is a reason to do the work properly.
A cow protein will not rescue the general trauma patient. It might rescue the patient who has no other door. That is a modest ambition, and it happens to be the one the evidence can support.
The prehospital plasma trial is Sperry and colleagues, New England Journal of Medicine, 2018 (PAMPer). The whole blood trial discussed here is TOWAR, published in the New England Journal of Medicine in 2026. A review of prehospital whole blood and plasma appeared in 2025, and the United States Army has published on a combined lyophilized product strategy. Numerical results quoted are from published summaries and abstracts and should be confirmed against the full reports.
The name of the foundation that hosts this blog contains a technical term that most visitors read as a slogan. Cold ischemia is the interval during which an organ removed from a donor sits cooled and without blood flow, waiting to be placed in a recipient. It is measured in hours. It is one of the few variables in the whole process that a human being can shorten, and it is the reason that every organ is packed in ice and rushed in a cooler.
This essay is about the physics and biochemistry of that interval and about a question that follows from it. If oxygen carriers are to help a donor organ, they must work at temperatures far below the ones for which they were designed. Does a molecule tuned for the warmth of a living body behave sensibly at four degrees? The answer turns out to be interesting, partly counterintuitive and central to how one might build a preservation product around a BHOC.
Cooling an organ is a bargain with biochemistry. Metabolic reactions slow as temperature falls, following a rule of thumb that rates drop by a factor of roughly two to three for every ten degrees Celsius. By the time a kidney or liver sits at about four degrees, its energy demand is a small fraction of what it was at body temperature; figures around a tenth or less are commonly quoted, though the exact fraction differs among tissues and conditions. The organ is not dead. It is in a kind of suspended animation.
The bargain has a cost. Even at low demand, the cells still consume some energy, and in the absence of blood flow they have no oxygen to make it with. They burn through their stores of adenosine triphosphate, the molecule that powers pumps and repairs. The pumps that keep sodium and calcium in balance falter. Cells swell. Mitochondria, the small structures that make energy, begin to accumulate metabolic intermediates. The longer the interval, the more of this damage builds.
One of the most striking discoveries in this area concerns a molecule called succinate. During ischemia, mitochondria in many tissues accumulate succinate, a compound normally consumed in the energy-producing cycle. When blood flow returns and oxygen floods back, the built-up succinate is oxidized rapidly and drives a burst of reactive oxygen species from the mitochondria. That burst is thought to be a major driver of the injury that occurs on reperfusion, the moment the organ is reconnected. In other words, the damage is not simply from being without oxygen. It is from the way oxygen returns.
This changes the question. The problem is not only how to keep cells alive in the cold. It is how to prepare them for the moment of reconnection so that the return of oxygen is orderly rather than explosive.
Now consider oxygen itself. Two physical facts collide at low temperature.
The first is that gases dissolve better in liquids when the liquid is cold. Oxygen is more soluble in water at four degrees than at thirty-seven, by a factor that is significant though not enormous. So a cold perfusion fluid can hold more dissolved oxygen at a given pressure. Combined with a lower metabolic demand, this means that the dissolved oxygen in a properly oxygenated cold solution may suffice, without any carrier, to supply an organ at rest.
The second is that hemoglobin binds oxygen more tightly in the cold. Cooling shifts the oxygen-binding curve to the left, lowering the P50, the pressure at which half the hemoglobin is saturated. The shift is substantial. A hemoglobin with a P50 in the thirties at body temperature has a much lower P50 at four degrees, so that it holds onto its oxygen until the surrounding pressure is very low. Physiologists know this effect from cold-water diving and from animals that let their extremities cool.
Put these two facts together and the picture is odd. In the cold, dissolved oxygen is more abundant and hemoglobin is more reluctant. A carrier that was designed to release oxygen at a moderate pressure in a warm body becomes, in ice, a molecule that grips its cargo. That is not necessarily fatal for the approach, because the organ needs so little oxygen and because the perfusate can be oxygenated to a very high pressure. But it means the role of a carrier in cold preservation is not a copy of its role in the body. It has to be reconsidered from first principles.
Before asking what a carrier adds, it helps to know how much oxygen alone accomplishes. Here the clinical evidence is unusually good, at least for the liver.
A randomized trial published in the New England Journal of Medicine in 2021 compared livers from donors after circulatory death that were preserved by static cold storage, the ice-cooler method, with livers that received a period of hypothermic oxygenated machine perfusion before implantation. The perfusion machine circulated an oxygenated solution through the liver at low temperature. Without going beyond what I could confirm, I recall that the trial found a substantially lower rate of symptomatic bile duct strictures in the perfused group, roughly 6 percent against roughly 18 percent. A careful reader should confirm those figures in the original report, but the direction and rough size of the effect are widely cited.
The important point is what the machine did not use. It did not circulate red cells. It did not use a hemoglobin carrier. The oxygen was dissolved in a cold, acellular fluid. That is a demonstration that in the cold, with a resting organ, dissolved oxygen can be enough to change outcomes. The proposed mechanism is that supplying oxygen during the cold interval lets mitochondria restore some of their energy stores and clear the succinate that would otherwise fuel the burst on reperfusion.
Any argument for a carrier in cold preservation must therefore start with humility. The baseline is not ice alone. The baseline is a machine that already delivers oxygen well, at least in the liver. The carrier would have to show a benefit beyond it.
There are at least four situations in which a carrier could plausibly earn its place.
The first is static cold storage with an additive. Many organs, especially kidneys in many centers, travel in a cooler with no machine at all. Machines cost money, need staff and are not available at every donor hospital. A preservation solution that included an oxygen carrier could give a slow supply of oxygen to an organ sitting in a bag, without any pump. A developer of a lugworm-derived giant hemoglobin has pursued exactly this idea, and reported a first-in-human use in a kidney in France in 2016. The evidence base is small and largely from the developer, and I would treat it as suggestive.
The second is the warmer end of the range. Between the cold of four degrees and the warmth of body temperature lies a zone, roughly twenty to thirty degrees, that some groups use for what they call subnormothermic perfusion. At those temperatures the organ has substantial metabolic demand but dissolved oxygen alone falls short, and a carrier’s affinity shift is smaller. This is the region where a carrier’s properties fit the task best.
The third is the normothermic machine. At body temperature, machines that perfuse an organ for hours use red cells as the carrier, and those red cells must come from donated blood. A cell-free carrier that could replace them would remove a dependence on blood supply, ease logistics and, in principle, allow perfusion to be offered in places without blood banks. This is perhaps the most straightforward application, because the temperature matches the body, and the molecule is working in conditions it was designed for.
The fourth is the reperfusion moment. If a carrier could be present in the perfusate at the moment the organ is reconnected, it might buffer the return of oxygen. This idea is more speculative, but it matches the mechanism: the injury comes from a burst, and a carrier could in theory smooth it.
One argument in favor of using a carrier in an isolated organ is powerful and deserves more attention than it gets. The history of oxygen carriers in the bloodstream is a history of systemic side effects: blood pressure elevation, effects on the heart, effects on the kidney, effects arising from the scavenging of nitric oxide in the vessel wall. Those effects are what stopped several programs.
An organ on a machine is outside the body. The carrier circulates through the organ and back to a reservoir. If the organ is flushed before it is connected to the recipient, most of the carrier is washed out, and the recipient receives little of it. In effect the drug is administered to the organ, not to the patient. The exposure that harmed trial subjects in the past is largely avoided.
I want to be careful. The claim is that systemic exposure is reduced, not eliminated. Some carrier remains in the small vessels and tissue, and its fate after reperfusion needs to be studied. The organ’s own vessel lining, exposed to a hemoglobin solution, may show local effects. But as a way of reintroducing an oxygen carrier to medicine while limiting its risk, the ex vivo setting is one of the most sensible I know. Ask the right question of any program: how much carrier remains in the organ after flush, how was it measured, and what happened to the recipients?
Abstractions are easier to judge against a concrete timeline. The following sequence is a composite for illustration, not a record of any real case.
The donor is declared, and the kidneys are recovered in an operating room while the donor is still on support. Cold preservation fluid is flushed through the vessels until the organ blanches, then the kidney is trimmed, bagged in sterile layers and packed in a cooler with ice. From that moment the clock of cold ischemia is running.
The kidney travels. It may go by road to an airport, by air to another state, by road again to a hospital. Each handoff is a chance for delay: a missed connection, a weather hold, a courier who cannot find the loading dock. Meanwhile, at the receiving hospital, a recipient is called in, tested and prepared, and a surgical team is assembled. If the recipient turns out to be unsuitable at the last minute, the kidney is redirected, and the clock keeps running.
When the kidney arrives it is inspected, and someone decides whether it looks well enough to use. That decision, made under time pressure, on appearance and a few numbers, is a place where a lot of quality and a lot of anxiety meet. Then comes the surgery, during which the kidney is warmed by the body and stitched into place, and finally the clamps come off and blood flows. That moment of reconnection, the reperfusion described earlier, is when the injury from the cold interval expresses itself.
Nothing in that sequence gives the organ oxygen. If a preservation system could add even a modest supply of oxygen during the transport, in a cooler with no pump, it would touch every stage of the journey without asking the couriers to do anything differently. That is why the static cold storage additive is, to my eye, the most attractive application of a carrier, even though the evidence is thinnest there.
A last caution concerns generalizing from the liver trial. Organs differ in their metabolism, their vascular architecture and their sensitivity to cold. The liver is a metabolically busy organ with a dual blood supply, and its injury pattern after preservation involves the bile ducts, which are exquisitely sensitive to oxygen deprivation. The kidney has a different profile. Its tubules consume large amounts of energy to reabsorb salts, and its characteristic post-preservation problem is delayed function, when the organ is slow to begin working. Machines for kidneys have been studied for years, and randomized trials have found reductions in delayed function with hypothermic machine perfusion compared with ice storage, although the size of the benefit varies among studies and the machines used are usually not oxygenated.
The upshot is that a carrier for kidneys would be entering a field with a well-established machine comparator, in which the addition of oxygen itself is still being tested. A prudent developer would design a kidney study with both questions in view: does oxygen help, and does a carrier add anything to oxygen delivered by simpler means? An advocate should want both questions asked, because a positive answer to the second would be far more persuasive than a positive answer to the first alone.
A cold perfusion carrier raises measurement problems that echo an earlier post. Clinicians assess a machine-perfused organ by looking at it. Color, texture and the way it blanches and refills are used as signs of quality, and an experienced surgeon judges viability partly by eye. A solution full of red-brown hemoglobin changes the look of the organ and may change those judgments. Perfusate chemistry tests, such as lactate or enzymes released from damaged cells, are used to judge injury, and a pigment can interfere with colorimetric assays.
The remedy is not to avoid carriers but to validate the assessments in their presence, and to agree on markers that do not depend on color. This is the sort of unglamorous protocol work that decides whether a technology is adopted.
If I were advising a developer about preservation in the cold, I would ask which molecule and why.
A carrier with a very high affinity would be a poor unloader at four degrees. A carrier with a moderate affinity at body temperature may become high-affinity in the cold, which is acceptable if the organ needs little oxygen but is wasteful if the goal is delivery. A carrier with a large size, such as a giant worm hemoglobin, stays in the vessels and may show good antioxidant behavior, which suits reperfusion. A bovine polymer sits between. It has a long track record of manufacturing and human use, a middling affinity and a modest size distribution.
There is also the option of a cold-tuned design, a carrier engineered so that its affinity at low temperature is closer to what the organ needs. Whether such a design is feasible is an open question. I raise it because the cold has been treated as a constraint to be endured rather than a design parameter to be used.
The choice of comparator deserves a sentence of its own. A carrier study that compares against nothing but ice will look better than one that compares against a simple oxygenated machine, and sponsors will be tempted by the easier comparison. Patients are better served by the harder one, because the harder one tells them whether the carrier is worth the added complexity, cost and regulatory burden.
For a patient waiting for a donor organ, the cold interval is invisible. It happens to somebody else’s organ, in a cooler, on a road or a runway, and is measured in hours nobody sees. Yet its length and quality shape how long the organ works after it arrives, and whether it works at all. A meaningful share of organs that are recovered are never used, for reasons that include concern about quality. Improved preservation is one of the few interventions that could widen the supply without asking a single additional person to donate.
That is the reason a foundation devoted to the cold interval should follow oxygen carriers closely and demand clear evidence. It should also be a wary audience. The strongest current data show that oxygen, delivered by a machine in a simple fluid, already helps. The carrier’s role is to extend that benefit to places without machines, to warmer perfusion and to the moment of reconnection. Those are worthwhile goals, but each requires its own evidence.
To close, imagine the study I would want to see. It would take organs from donors of comparable type, randomize them to standard preservation or to preservation with a defined carrier, and record the time in the cold. It would measure the carrier remaining in the organ before implantation. It would follow recipients for function of the organ, injury markers, delayed function, complications and survival of the graft at one year. It would report all results, positive and negative, and it would be overseen by an independent committee.
It would not be a large study at first. A small, careful, transparent study of that kind would be worth more than a dozen glowing case reports. And it would be a study a patient community could support without hesitation, because it asks a modest question: does adding oxygen to the cold help?
The randomized trial of hypothermic machine perfusion in liver preservation appeared in the New England Journal of Medicine in 2021, and follow-up and review articles discuss the mechanism. The role of succinate accumulation and reperfusion injury was described in work by Chouchani and colleagues in 2014. The developer of a lugworm-derived carrier has published descriptions of its preservation use. Quantitative claims about temperature effects on metabolism and hemoglobin affinity are approximate and should be checked in physiology references.
Nobody designs a blood substitute by looking at a Himalayan village, and that is the first mistake. Every chemist working on hemoglobin-based products carries an assumption so deep that it rarely gets stated: more oxygen-carrying capacity is better. If a patient is anemic, add hemoglobin. If tissue is starved, add hemoglobin. If the first product fails, make a more concentrated one. The assumption feels like arithmetic.
Populations who have lived for thousands of years at four thousand meters, and animals that dive for an hour on a single breath, have been running a much longer experiment on that assumption, and their answer is more complicated. In several of them, evolution pushed hemoglobin concentration down, not up. In others it pushed the storage of oxygen out of the blood altogether. Reading their results carefully does not give a designer a blueprint. It gives a set of warnings and a set of questions, and I think that is the more useful thing.
Consider two large human populations who have lived at high altitude for a long time. The Tibetan plateau sits above four thousand meters, and the high Andean altiplano is similar. The air at those altitudes contains the same fraction of oxygen as at sea level, but the pressure is roughly forty percent lower, so each breath delivers correspondingly less. Every resident faces the same physical problem.
Their solutions differ. Andean highlanders tend to respond the way lowlanders do when they climb: their bodies make more red cells, and hemoglobin concentration rises, in some individuals to very high values. This is an adaptation of a kind, since more hemoglobin means more oxygen per liter of blood. Tibetans do something else. A review of the research on hemoglobin at altitude summarizes their pattern as a relatively low erythropoietic response with a correspondingly low hemoglobin concentration. Tibetans living on the plateau have hemoglobin values that are strikingly close to those of people at low elevation, or only modestly higher.
If you had to bet before looking at the data, you would have bet on the Andean strategy. It is the one a physician would recommend. The puzzle is that the Tibetan strategy appears to work better.
The explanation begins with a property of blood that clinicians know but rarely feel: viscosity. Blood is a suspension of cells, and as the fraction of cells rises, the fluid thickens, slowly at first and then steeply. Thicker blood is harder to pump, flows more sluggishly through small vessels, and can reduce delivery to tissues even though it contains more oxygen per milliliter.
The relation between cell fraction and oxygen delivery therefore has a peak. Below the peak, adding cells helps, because capacity dominates. Above it, adding cells hurts, because viscosity dominates. Physiologists have estimated where the optimum falls in various circumstances, and it is lower than most people expect, and it moves with conditions.
The Andean pattern can overshoot. Chronic mountain sickness is a syndrome of excessive red cell production at altitude, and it brings headaches, fatigue, blue lips, and an increased risk of stroke and heart attack. The review I mentioned reports that among Peruvian Andean males the condition can reach rates as high as 18 percent, while among Tibetans it is around 1 percent. In other words, the population that made the intuitively sensible response paid for it in disease, and the population that held back did not.
This is not a story about one group being better than the other. Genetic and developmental factors differ, and Andeans have other adaptations, including changes in blood vessel function. It is a story about the shape of a curve: there is a range where more is better, and beyond it there is a range where more is worse. That shape is the first instruction the mountain gives.
The molecular story is unusually clear for a human adaptation. A gene called EPAS1 encodes a protein known as HIF-2 alpha, a transcription factor that helps cells respond to low oxygen, and among its targets is the erythropoietin gene, the master regulator of red cell production. Tibetans carry variants of EPAS1 that appear to reduce its activity. The review states that these variants function as loss-of-function alleles that downregulate the protein product, and that they reach a frequency of roughly 72 percent in Tibetan populations.
The origin of the variant was a surprise. In 2014 researchers reported that the Tibetan version of the gene sits within a stretch of DNA that closely resembles the genome of Denisovans, an extinct group of ancient humans known from a cave in Siberia. The finding suggests that a variant useful for surviving on the plateau was inherited through interbreeding with an archaic population and then favored by natural selection in the mountains.
The ironic part is worth stating clearly. A gene variant that reduces the body’s response to low oxygen was selected for in people living where oxygen is scarce. That sounds paradoxical until you recognize that the body’s response to low oxygen is a blunt instrument. Ramping up red cell production in response to every dip in oxygen level is a strategy that suits a short trip to altitude, and can be counterproductive over a lifetime.
If Tibetans do not compensate by adding hemoglobin, they must compensate some other way, and the research points to several. Their resting ventilation tends to be high, they show relatively strong oxygen delivery through increased blood flow to certain tissues, and studies have reported higher levels of nitric oxide in their circulation than in lowlanders at the same altitude, which would dilate vessels and improve flow. Some evidence suggests differences in how they use oxygen within cells.
I would not lean too hard on any single item on that list; the field is active and the picture is still being filled in. What the collective evidence suggests is that the Tibetan strategy favors delivery through flow and efficiency rather than through capacity. It keeps blood thin, keeps vessels open, and makes each molecule of oxygen count.
For a designer of oxygen carriers, this is instructive. The tissue does not need more hemoglobin. It needs oxygen delivered at the right place, at the right rate, through vessels that stay open. Those are three separate targets, and adding hemoglobin addresses only the first.
Move now from mountains to the sea. A Weddell seal can dive for tens of minutes on one breath, and some whales can dive for far longer, while a human freediver, with training, can reach several minutes. The seals do not have magical lungs. In fact, diving mammals often carry relatively little oxygen in their lungs, because compressed air in the chest is a hazard at depth. They carry it elsewhere.
They store oxygen in three places. First, they have unusually large blood volume relative to body size. Second, their blood is rich in hemoglobin, and in some seals a contracting spleen squeezes a reservoir of packed red cells into the circulation during a dive, raising the concentration temporarily, a mechanism that lets the animal hold thin blood at rest and thick blood when it is needed. Third, and most strikingly, their muscles contain very high concentrations of myoglobin, the single-chain relative of hemoglobin that stores oxygen inside muscle cells. Seal and whale muscle is nearly black from it.
The design principle is different from that of the mountain dweller. Here the animal maximizes stored oxygen, and it does so by putting the storage where it is used, inside the muscle, so that the oxygen does not have to travel. The blood serves as a supplemental reservoir and a transport line, but the emergency supply sits at the point of need.
Storage and transport are usually confused in discussions of oxygen carriers, and the confusion has practical consequences.
A transport system moves oxygen from where it is plentiful to where it is needed. Its value depends on flow. If flow stops, the transport system delivers nothing, however full its carriers.
A storage system holds oxygen where it is needed. Its value depends on how much it holds and how fast the tissue draws on it. If flow stops, the store lasts until it is empty.
Human blood evolved as a transport system with a modest reserve. The venous blood returning to the lungs is still about three-quarters saturated at rest, which is a reserve, but not a large one. When flow stops, as in a cardiac arrest or a clamped vessel, the reserve is spent within minutes.
A cell-free carrier is also a transport system, and its main potential contribution is to reach places that red cells cannot. But it is worth asking whether there are situations in which a carrier could act more like storage. An organ held cold on a machine is one. If the preservation fluid holds a high dissolved oxygen content in the form of a carrier, the tissue has a small reserve to draw on even without flow. A tissue starved by a blockage, briefly, could use the same reserve. I raise this not as a claim about existing products but as a conceptual tool: when you evaluate a carrier for a particular use, ask whether the job is transport or storage, because the answer changes what properties matter.
Another lesson from altitude concerns oxygen affinity, the tightness with which hemoglobin grips oxygen. In another post I described how the P50 is the pressure at which hemoglobin is half saturated, and how the tissue’s needs favor different values in different jobs.
Animals at altitude show the effect in a beautiful way. The bar-headed goose migrates across the Himalayas at altitudes that would incapacitate most birds, and its hemoglobin is reported to have a higher oxygen affinity than that of lowland geese, a difference traced to changes in the amino acid sequence. Llamas, adapted to the Andes, also carry hemoglobin with higher affinity than other mammals of their size. The logic is plain: at low pressure, a hemoglobin that grips harder loads more completely in the lung, and the tissue still receives what it needs because the pressure drop from lung to tissue is steep.
Humans at altitude do something related. A shift in the oxygen-binding curve toward higher affinity is seen in some highland populations and in pregnant women, and in the fetus, whose hemoglobin has a higher affinity than an adult’s so that it can extract oxygen across the placenta.
The design implication runs against the intuitions of a first-time designer. The best affinity is not fixed; it depends on the environment. A product intended for a patient breathing thin air in an ambulance at altitude, or a patient with a lung problem that impairs loading, might want a higher affinity than one intended for a patient with a normal lung and a starved limb. There is no single right P50. The mountain tells you that the answer depends on where you live.
There is a further link, and it involves a therapy that emergency physicians know well: hemodilution. When a patient loses blood and receives fluid, the remaining red cells are diluted. The patient’s hematocrit falls, and blood becomes thinner. It is easy to think of this as pure harm, and if it goes too far, it is.
But moderate dilution improves flow through small vessels by reducing viscosity, and the body compensates for the lower capacity by pumping more blood. There is a range of hematocrit, lower than normal, in which delivery is about as good as or better than at normal values. The Tibetan pattern, seen from this angle, is a permanent version of controlled hemodilution.
An oxygen carrier dissolved in plasma is a way to have it both ways. It adds oxygen capacity without adding cells, so it raises capacity while leaving viscosity low. In principle, it lets a patient be thin-blooded and well-supplied. This is the deepest of the arguments for the whole class of products, and the mountain reinforces it. The corresponding risk is that the dissolved carrier brings its own problems, discussed elsewhere on this site, including the scavenging of nitric oxide, which is exactly the molecule that the Tibetan body appears to keep at higher levels. The highlander’s physiology and the dissolved carrier pull in opposite directions on that one variable.
A small calculation makes the shape of the curve concrete. The numbers are illustrative simplifications and should not be read as physiological constants.
Suppose oxygen delivery is proportional to two things multiplied together: the oxygen content of a liter of blood, which rises in direct proportion to the fraction of cells, and the flow rate the heart can sustain, which falls as viscosity rises. Take a hematocrit of forty-five percent as the baseline, with an oxygen content of 1.0 in arbitrary units and a flow of 1.0. At a hematocrit of thirty, the content falls to about 0.67, but the flow rises, because thinner blood moves more easily. If the flow improves by, say, thirty percent, the product is 0.87, a modest loss. At sixty percent, the content rises to 1.33, but if flow drops by forty percent because the blood has thickened badly, the product is 0.80. Both extremes are worse than the middle.
The exact peak depends on the vessel, the heart and the person, so this is only an outline. But it shows why a physician who sees a low hemoglobin should not automatically see a problem, and why a physician who sees a very high hemoglobin should not automatically see an advantage.
The sport of endurance racing supplies a second natural experiment, and an uncomfortable one. Athletes have long known that training at altitude raises red cell mass, and some have used shortcuts, including transfusion of their own stored blood and injections of erythropoietin, to raise hematocrit further. These practices are banned in sport for good reasons, including fairness and safety.
The safety story is a direct illustration of the viscosity problem. Endurance athletes who pushed their hematocrit beyond natural limits were reported in the 1990s to have suffered serious cardiovascular events, and the sport introduced hematocrit limits, initially as a health measure. The lesson is the same as the mountain’s: raising capacity by making blood thicker has a ceiling, and exceeding it can be lethal.
The athlete’s case also points at a design opportunity. If capacity can be raised without thickening the blood, the ceiling moves. That is precisely what a dissolved carrier promises, and precisely why the sporting world has watched such products with anxiety. Some carriers have appeared on lists of prohibited substances, as they were used or considered for use in sport, and their detection is a live topic in anti-doping laboratories. It is an unexpected reminder that the same property that might help a bleeding patient could distort a race.
It would be irresponsible to end on the note that nature has shown the way. It has shown several ways, and some of them do not translate.
Populations differ from patients. A highlander’s body has been shaped over hundreds of generations to match its environment. A patient in shock has had minutes. Adaptation is not the same as resuscitation.
The observations are correlational. We see a low hemoglobin and a low rate of chronic mountain sickness, and we infer a link, but many other differences accompany the pattern, from diet to lung size to genetics elsewhere in the genome. The gene story is stronger than most, because the variant has been mapped and its effect on hemoglobin quantified, but even here the wider consequences are still being studied.
Finally, extrapolating from a healthy population to a sick one is a classic source of error. A lower hemoglobin is adaptive in a healthy Tibetan with intact vessels and lungs. It may be dangerous in a person with heart disease.
If I were sitting with a group of developers, I would bring four questions from the mountain and the sea.
Is your target patient limited by capacity, by flow, or by loading? A product that raises capacity does little for a patient whose problem is a blocked vessel, and a product that improves flow does little for a patient who is simply low on hemoglobin.
Where on the curve of viscosity and capacity does your patient sit? A patient already thin-blooded may gain little from added capacity, and one already thick-blooded, for example after a long stay at altitude or with a blood disorder, may be harmed by it.
What affinity does the environment call for? A single value for all uses is a compromise, and the compromise should be stated.
Are you transporting or storing? If the answer is storing, the product might look more like myoglobin than like blood.
The mountains do not answer those questions. They teach how to ask them.
The comparison of hemoglobin concentration across highland populations is reviewed in a 2025 article in the American Journal of Human Biology on adaptive hemoglobin concentration at high altitude. The Tibetan EPAS1 variants were reported in a 2010 paper in the Proceedings of the National Academy of Sciences, and the Denisovan-like origin of the variant in a 2014 paper in Nature. Values quoted for disease prevalence come from the 2025 review and vary by study and definition. Statements about diving mammals reflect standard comparative physiology and should be confirmed in a specialist source before use.
A hemoglobin-based carrier is a chemical solution to a physical problem. Oxygen barely dissolves in water, so blood evolved a molecule that binds it. Everything about the BHOC story, from the affinity curve to the iron chemistry to the trouble with nitric oxide, follows from that fact.
There is another way to carry oxygen, and it involves no binding at all. Certain synthetic liquids, chains and rings of carbon in which nearly every hydrogen has been replaced by fluorine, dissolve oxygen the way a sponge soaks up water. They are called perfluorocarbons, and they have a long and peculiar history as oxygen carriers. This essay lays out how they work, where they fit, and what the comparison teaches about the hemoglobin products that dominate the conversation on this site.
When oxygen dissolves in a liquid, the amount that dissolves is proportional to its partial pressure. This is Henry’s law, and it is one of the simplest relationships in physical chemistry. Double the pressure and you double the dissolved oxygen. In ordinary plasma, the solubility is low: at the pressure of arterial blood, a deciliter of plasma holds only a fraction of a milliliter of dissolved oxygen.
A perfluorocarbon liquid dissolves far more, on the order of twenty times more than water by the figures usually quoted, and a pure liquid of this type can hold something like forty or fifty milliliters of oxygen per hundred milliliters when exposed to pure oxygen at atmospheric pressure. I offer those as rough figures from the literature, and they vary by compound.
But the rule is a straight line. Hemoglobin follows an S-shaped curve: it loads almost fully at the pressure of ordinary lung air and then holds its oxygen tightly until the pressure falls near that of the tissue. A perfluorocarbon does neither. It picks up oxygen in proportion to how much is around and gives it up in proportion to how much is missing. The consequence is the defining feature of the whole class: to load a perfluorocarbon with a useful amount of oxygen, you need a high oxygen pressure. The patient must breathe oxygen-enriched air or oxygen. That is what I mean by needing a mask.
This is not a small caveat. It is the opposite of hemoglobin’s design. A hemoglobin carrier can be loaded by breathing room air. A perfluorocarbon carrier needs an oxygen supply and a delivery apparatus in addition to the infusion, which is a logistical and clinical complication in the very settings where a carrier is most wanted.
Perfluorocarbons are hydrophobic and lipophobic, meaning they mix with neither water nor oil. You cannot inject them as they are. Making an intravenous product requires turning the liquid into a fine emulsion, tiny droplets suspended in water with the help of an emulsifier. The droplets in these products are on the order of a fifth of a micrometer across, smaller than a red cell by a factor of about thirty.
That small size is what allows the droplets to travel through capillaries, and in principle to reach tissue that red cells cannot. It is the same rationale that motivates the hemoglobin carriers, achieved by a different route.
The emulsions carry their own problems. Emulsions can be unstable, and the early ones had to be stored frozen, and thawed before use. The droplets are removed from the blood by cells of the reticuloendothelial system, mainly in the liver and spleen, so the product has a half-life measured in hours to days and leaves behind a burden of fluorinated material that is slowly exhaled over weeks to months. Recipients of early products commonly experienced flu-like symptoms, a reaction thought to be related to uptake of the droplets by macrophages and the release of inflammatory signals.
The most notable perfluorocarbon product was called Fluosol, developed in Japan and approved in the United States around 1989 for use in a specific setting: protecting heart muscle during balloon angioplasty, when a vessel is briefly blocked. The idea was clever. During the few minutes that a balloon was inflated in a coronary artery, blood could not pass, but the small droplets could squeeze past the balloon through collateral channels, and if the patient was breathing high concentrations of oxygen, they carried a useful amount to the starved heart tissue.
The product was subsequently withdrawn from the market. The reasons were commercial and practical as well as clinical: it needed to be stored frozen, prepared with several components before use, and required oxygen breathing, while advances in angioplasty equipment, such as balloons that allow blood to flow past, reduced the need for it. The details of its commercial life deserve better sourcing than I have gathered here, so I will keep to the shape of the story. It was approved, it was used, and it faded.
A later generation of emulsions, made with a different compound and stabilized with a phospholipid, reached late-stage clinical trials in surgical patients. One of them, developed under the name Oxygent, was tested in surgery to reduce the need for blood transfusion, and the program encountered safety concerns in a trial and did not reach the market. I mention these to make a general point about the class. A physically elegant idea did not translate into a successful product. The reasons were mostly practical, and practical reasons are where the field lives.
Set the two designs side by side and each shows what the other lacks.
Hemoglobin gives a lot of oxygen capacity at low oxygen pressure. A concentrated hemoglobin solution holds many times the oxygen of an equal volume of plasma at ordinary arterial pressure, without needing the patient to breathe enriched oxygen. That is the great advantage of chemistry over physics.
The perfluorocarbon droplets, by contrast, hold oxygen in a way that responds smoothly to local conditions. In the lung, when the patient breathes oxygen, the droplets load; in the tissue, where pressure is low, they release. Because the relationship is linear, a perfluorocarbon delivers a large fraction of its oxygen in the tissue range, while hemoglobin at high affinity may not. In tissue that is very low in oxygen, the sponge gives up cargo readily.
The perfluorocarbon has no iron and therefore no oxidation problem. It does not scavenge nitric oxide, so it does not cause the vasoconstriction associated with free hemoglobin. It is not a biological protein, so it does not raise concerns about foreign proteins or prions. It can be stored, if the emulsion is stable, at room temperature.
The hemoglobin, meanwhile, has the S-curve, the ability to load in ordinary air, a greater carrying capacity per unit volume in the relevant pressure range, and the cooperative behavior that has been refined over hundreds of millions of years.
I find the pairing instructive because it shows that the problems of hemoglobin carriers are not properties of oxygen carrying in general. They belong to a specific chemistry. If a patient community worries that all oxygen carriers cause vasoconstriction, the perfluorocarbon record is a counterexample. If it worries that all oxygen carriers require a breathing apparatus, the hemoglobin record is a counterexample. The properties that define a class are chemical.
The straight-line rule can be turned into arithmetic, and the arithmetic is sobering. The figures below are round numbers chosen for illustration, not values for any product.
Start with ordinary blood. Hemoglobin at 12.5 grams per deciliter, fully saturated, holds about 17 milliliters of oxygen per deciliter, using the usual figure of 1.34 milliliters per gram. Plasma, by dissolved oxygen alone, adds about 0.003 milliliters per deciliter for each millimeter of mercury of pressure, so at an arterial pressure of 100 it adds roughly 0.3. Tissues normally extract about 5 milliliters per deciliter on a single pass.
Now suppose we infuse a perfluorocarbon emulsion until it makes up a tenth of the blood volume, and that the emulsion is 30 percent perfluorocarbon by volume, so that 3 percent of the blood is the liquid itself. Take a liquid that dissolves about 45 milliliters of oxygen per 100 milliliters at one atmosphere of pure oxygen. While the patient breathes room air, with an arterial pressure near 100 millimeters of mercury, the perfluorocarbon adds about 0.18 milliliters per deciliter. That is smaller than the oxygen dissolved in plasma and negligible against the 17 in hemoglobin.
Now let the patient breathe pure oxygen, raising arterial pressure to about 600. The perfluorocarbon load climbs to roughly 1 milliliter per deciliter. That is a six-fold increase, and it is meaningful for a patient whose hemoglobin has fallen very low. But it is still a small fraction of what healthy blood carries. The sponge does its best work when the patient has almost no red cells left, or when the flow is so poor that a little extra oxygen makes the difference.
The arithmetic explains both the appeal and the disappointment. It shows why perfluorocarbons were tried as a bridge for patients who refused blood, and why the amounts involved could never replace a large hemoglobin deficit. It also shows why the concentration of the emulsion and the volume infused mattered so much, and why doses large enough to matter tested the limits of what patients tolerate.
There is a practical detail of the first commercial emulsion that illustrates how much clinical adoption depends on mundane things. The product was supplied frozen, and it had to be thawed, combined with additional solutions, and prepared shortly before use. Hospital pharmacists and perfusionists had to schedule the preparation around the procedure. A drug that requires a checklist and a countdown is a drug that is used only when someone has planned for it.
The comparison with a room-temperature bovine hemoglobin bag makes the point vividly. A shelf-stable product can be grabbed and hung. It does not depend on a pharmacy. Convenience is not a footnote in emergency medicine. It is often the deciding factor between a product used and a product admired.
Now the paradox that gives the essay its title. If a patient in shock is already receiving high-flow oxygen through a mask, as is standard in emergency care, then a perfluorocarbon may load in the lung and be effective. The need for supplemental oxygen is not an extra burden in a patient who would receive it anyway.
The complication arises in the field. A soldier or an accident victim in a remote place has no oxygen cylinder, or has only a limited supply. The very setting that argues for a room-temperature, shelf-stable carrier, the field, is the setting where the extra requirement is hardest to meet. And in a patient who breathes room air, the droplets add little.
There is a second complication in the opposite direction. High concentrations of oxygen are not benign. Prolonged breathing of very high oxygen levels can injure the lungs, and it changes the balance of reactive oxygen species in tissue. A treatment plan that depends on a high inspired oxygen fraction inherits those risks.
The paradox teaches something about how to evaluate any carrier. Ask what the carrier requires in order to work. Hemoglobin requires that the lung function and that the iron remain reduced. Perfluorocarbons require enriched oxygen. Encapsulated hemoglobin requires that the vesicle remain intact. Each dependency is a potential failure point in a particular setting.
The oddest and most encouraging uses of perfluorocarbons have been in places where the mask problem does not arise, because the oxygen is delivered directly.
One is organ preservation. In some preservation methods, an organ is stored in a two-layer arrangement in which a perfluorocarbon layer at the bottom is bubbled with oxygen and the organ, usually a pancreas, floats at the interface, receiving oxygen by diffusion. The perfluorocarbon acts as an oxygen reservoir, with no need for a patient to breathe anything. This method has been studied in pancreas preservation for islet cell isolation, and the results have been described as encouraging in some studies and modest in others. I raise it because it demonstrates a use in which the sponge is well suited: an organ at low temperature and low metabolic demand, in a bath oxygenated from outside. This is exactly the scenario described in a companion essay on cold oxygen, and it shows that the perfluorocarbon approach might have a natural home in preservation.
Another is eye surgery, where certain perfluorocarbon liquids are used as a temporary tool to flatten a detached retina. That use has nothing to do with oxygen carrying, and it is a reminder that these compounds are chemically inert and safe enough for direct contact with delicate tissue.
A third, more speculative, is liquid ventilation, in which the lungs are partially filled with an oxygenated perfluorocarbon liquid. It was tested in premature infants and adults with lung injury, and it has been a favorite of science fiction. The clinical trials did not show the benefit that supporters hoped for. It nonetheless shows how far the field was willing to go.
Return to the cold. The physics of dissolved oxygen changes with temperature, and the change favors the sponge. Gas solubility in liquids increases as temperature falls. A perfluorocarbon at four degrees dissolves more oxygen than at body temperature. Hemoglobin, in contrast, grips oxygen more tightly in the cold, which makes it a poorer donor. In the cold, then, the relative merits reverse: the sponge gets better and the chemical carrier gets worse.
That observation, which follows directly from Henry’s law and the temperature dependence of hemoglobin affinity, deserves attention from anyone thinking about preservation products. It suggests that a preservation fluid for an organ in the cold might do better with a physical carrier than a chemical one, or with a mixture in which each does the job it is best at.
I would go further. The most interesting design for a warm perfusion might combine both: a bovine hemoglobin carrier for capacity at ordinary oxygen pressures, and a perfluorocarbon component to buffer oxygen in the microvessels. I am not aware of such a combination being tested in the clinic, and I mention it as a question rather than a proposal.
There are three points to take away for a patient audience.
First, oxygen carrying is not synonymous with hemoglobin. When someone says a BHOC is the only way to carry oxygen without red cells, they are wrong, and the history of perfluorocarbons is the proof. That matters because it places the BHOC in a competitive landscape. An honest advocate says which of the alternatives it is meant to beat, and on what.
Second, the history is a lesson in the difference between physical elegance and clinical success. Perfluorocarbons had beautiful physics and an ugly logistics profile. Hemoglobin has an ugly set of side reactions and a beautiful mechanism. Neither won. A patient community that understands both stories is less likely to be dazzled by a single clever property.
Third, it shows how much depends on the setting of use. Every one of the perfluorocarbon successes was a niche with special conditions: a balloon in a coronary artery, a bath for a pancreas, an eye. The general-purpose blood substitute remains unsolved. This does not mean that niche uses are failures. On the contrary, they may be the only path by which a technology reaches patients at all. A BHOC that finds a defensible niche, whether organ preservation, patients who cannot receive blood, or remote care, stands a better chance than one that aims at replacing blood.
One more observation belongs here, because it is easy to miss. The perfluorocarbon record is a reminder that regulators and clinicians judge a product by the whole system it lives in: the frozen box, the thaw, the oxygen tank, the mask, the monitoring. A carrier that is chemically brilliant and operationally clumsy loses to one that is chemically ordinary and operationally easy. That is worth remembering when the BHOC story is told in terms of molecules alone. The molecule is one part of the product. The bag, the label, the storage temperature and the training of the person who hangs it are the rest.
Borrowing from the comparison, I would suggest a short list for any product.
What does the patient have to breathe, and what does the site need to supply?
What is the carrier’s half-life, and what happens to it when it leaves the circulation?
What is the physical form: a solution, an emulsion, a particle, and how stable is it on the shelf?
What does the product do to the tissues that clear it, since every carrier is cleared by something?
And what is the honest niche, the place where the property that limits the product is not limiting?
Those are ordinary engineering questions, and the perfluorocarbon field, for all its disappointments, taught the community to ask them.
Reviews of perfluorocarbon-based oxygen carriers, including a 2010 review of products and trials and a review that traces the field from physics to physiology, describe oxygen solubility, emulsion properties and clinical experience in detail. Figures for solubility, droplet size and half-life vary by compound and formulation and are quoted here as rough orders of magnitude. The history of Fluosol’s approval and withdrawal and of the Oxygent trial deserves confirmation in primary sources before it is repeated.
Every hemoglobin-based oxygen carrier that has reached a clinic is a bare protein. Whether the protein is cross-linked, polymerized, or wrapped in polymer chains, it floats free in plasma with nothing between it and the vessel wall. That design carries the direct exposure that produced the troubles of the class, and a persistent minority of researchers has argued that the way forward is to take a step back: put the hemoglobin in a package again.
A red cell is not a bag of hemoglobin. It is a machine. A membrane keeps the protein in and the plasma out, a set of enzymes keeps the iron reduced, a bath of small molecules tunes oxygen affinity, and a flexible skeleton lets the whole thing fold through a capillary narrower than itself. Recreating any of that outside the body is difficult. Three programs, each aiming at a different piece of the machine, have advanced far enough to teach us something. This essay describes them, and it puts them beside the bovine polymers to ask a fair question: what is a packaged carrier for?
Start with the logic of the package, because it is the same logic that evolution followed. A membrane does four jobs for a red cell.
It keeps hemoglobin away from the vessel wall. Free hemoglobin scavenges nitric oxide, a gas the wall uses as a relaxing signal. A cell keeps the scavenger inside, at a distance.
It keeps hemoglobin away from the kidney. A protein small enough to leak through the filter is lost, and can be toxic there. A package that is much larger than the filter’s pores cannot leak.
It supports a chemical environment. Inside a cell, enzymes and small molecules keep iron in the working form and regulate affinity. Inside a synthetic package, at least some of those functions can be included by adding the right compounds.
And it controls the concentration. A packaged carrier can hold hemoglobin at the same high concentration as a red cell, so the oxygen capacity per particle is large, while the fluid in which it travels stays thin.
These are real advantages. They have to be weighed against the costs of a package: it must be made, it must be stable, and it must not trigger the immune system or clog the very vessels it hopes to reach.
The most advanced packaged carrier is the hemoglobin vesicle, developed over decades by research groups in Japan. It consists of a highly concentrated hemoglobin solution enclosed in a bilayer of phospholipids, the same class of molecule that forms a cell membrane. The particles are small spheres about a quarter of a micrometer across, roughly thirty times smaller than a red cell. The surface is decorated with polyethylene glycol to reduce recognition by the immune system and to prevent the particles from sticking to one another.
The hemoglobin inside is purified from outdated donated human red cells, and the vesicles are formulated with additional compounds that mimic the small-molecule environment of a cell, including a molecule to tune oxygen affinity and reducing agents to protect the iron. The developers have reported that the product can be stored for long periods, and have described it as a candidate for emergency use where blood is unavailable. A first-in-human phase 1 trial was published in 2022 in a hematology journal. It enrolled a small group of healthy volunteers, gave them single doses, and reported on safety and tolerability. I have not been able to confirm its detailed results, and I would treat any claim about its performance in patients as unproven.
The design has a few clear attractions. It is built from human hemoglobin and lipids, so it introduces no animal proteins. The particle size is well above the range that escapes the kidney. The iron is protected inside a packaged environment. And its behavior in animals has been extensively studied, including the observation, reported in rodents, that injection did not cause the anaphylactic reactions that some lipid-based particles are known to provoke.
The design has clear drawbacks too. The raw material is human blood, so it depends on the same donor supply it hopes to supplement, although it is made from expired units that would otherwise be discarded. Lipid particles are cleared by the reticuloendothelial system, the same scavenger cells that take up perfluorocarbon emulsions, and this clearance limits the half-life and raises questions about long-term effects on those cells. And manufacturing a uniform lipid particle at scale is a specialized process.
A second approach attempts to build the package from synthetic materials, with no cells and no donor blood in the loop. One design in this family uses hemoglobin held in a nanoparticle with a polymer shell, together with molecules that mimic the enzymes and affinity regulators of the red cell, and a formulation that can be freeze-dried into a powder. The stated goal is a product that can be stored at room temperature and reconstituted with water in the field.
I want to be very careful about how I describe this. It is a research program from a small number of laboratories, with published animal data and a strong narrative but no completed large human trial that I could verify. I mention it because it shows the direction in which the idea is being pushed: from a natural package to an engineered one, with each function of the red cell replaced by a designed component. The appeal is obvious. If a synthetic package works, it removes the dependence on blood entirely. The uncertainty is equally obvious. Each added component adds an interaction to be tested, and complex particles have a history of surprising behavior in the body.
Here is a way to think about the difference between the two programs. The vesicle is a conservative design: it keeps nature’s blueprint, the phospholipid bilayer, and works to stabilize it. The synthetic nanoparticle is a radical design: it keeps the function and replaces the structure. Conservative designs tend to fail slowly and modestly. Radical designs tend to succeed dramatically or fail quietly. Neither can be judged in advance.
The third approach avoids substitutes entirely. Instead of building a package, grow a genuine red cell in the laboratory. Blood stem cells, taken from donors, can be coaxed in culture to mature into red cells, which are then harvested and transfused.
A study in the United Kingdom took this beyond the bench. In a trial conducted in the early 2020s, healthy volunteers received small volumes of laboratory-grown red cells that had been tagged so their survival could be measured against the volunteers’ own cells. The aim was not therapy but a demonstration that the cells could circulate and survive in a human. Reports described it as the first trial of its kind in the world, and the results were being analyzed at the time I gathered this material. I cannot give a verified account of the findings, and I will not guess.
The theoretical attraction is immense. A cultured cell has the membrane, the enzymes, the affinity regulation and the flexibility, because it is a red cell. If cultured cells could be made in volume, they could supply patients whose blood types are rare or who have developed antibodies against common types, a group for whom compatible blood is hard to find.
The practical obstacle is scale and cost. A single unit of red cells contains on the order of a trillion cells. Producing that number in culture, with the nutrients and growth factors required, and doing so sterilely and reproducibly, is orders of magnitude beyond what current laboratory methods deliver. The realistic near-term use is small-volume, high-value: rare-type patients, not trauma. For emergency oxygen, a cultured cell is not a competitor to a shelf-stable bag. It is a complementary tool for a different job.
Scale deserves a worked example, because it is the point at which many elegant ideas quietly die. The numbers here are round figures for illustration.
A unit of donated red cells, roughly 300 milliliters, contains on the order of two trillion cells. A single bleeding adult may need several units, so the raw count for one patient runs to ten trillion or more. Now suppose a culture system can expand one starting stem cell into a hundred thousand mature red cells, a generous assumption. To make one unit, you would need twenty million starting cells. To make a hundred units, two billion. Those are numbers a bioreactor can handle in principle, but each of those cells needs weeks of nutrients, growth factors and sterile handling, and the cost of the growth factors alone can run into thousands of dollars per unit at laboratory prices.
The point is not that the arithmetic makes the idea impossible. It is that it sorts applications by value. A unit that costs thousands of dollars to grow makes sense for a patient with a rare blood type who would otherwise go untreated. It makes no sense for a trauma patient who needs six units in an hour. That is why I placed cultured cells beside the shelf-stable carriers as a complement and not a competitor.
The same reasoning applies to vesicles and synthetic particles, in a different form. Their cost is not in growth factors but in purified hemoglobin, in lipids of pharmaceutical grade, and in the yield of a multi-step manufacturing process. A packaged carrier that costs many times as much as a bare polymer per gram of hemoglobin has to earn the difference through a benefit that patients and payers can see.
Packaged carriers pose a particular challenge for preclinical testing. The reason is that many of their problems depend on the immune system and on the cells that clear particles, and these differ between species.
Consider complement activation. Some lipid and polymer particles trigger a chain of proteins in the blood that can cause flushing, chest tightness, breathing difficulty and a drop in blood pressure within minutes of the first infusion. The reaction does not require prior exposure, and it does not involve the antibodies of classical allergy. Its strength varies enormously among species. Pigs are notoriously sensitive, and rodents are far less so. A finding of no reaction in mice is therefore reassuring only up to a point.
Consider clearance. The scavenger cells that take up particles are more or less abundant in different species, and their capacity to handle repeated doses is uncertain. A single injection in a rodent cannot tell you what a series of injections in a person will do.
Consider size and shape. A particle that passes easily through a rodent capillary may behave differently in a human vessel, and rigid particles may be filtered by the lung.
The upshot is that the first human doses are a genuinely informative experiment, more so than for a simple protein, and the design of that experiment deserves care: small starting doses, slow infusion, monitoring of complement markers and of inflammatory signals, and staged escalation.
The mention of healthy volunteers in the vesicle and cultured cell trials invites a question worth asking. Why test an emergency product in people who are not in an emergency?
The reason is safety. Before a carrier is given to a critically ill patient, in whom side effects are hard to separate from the disease, it is given to healthy people so that its effects can be seen cleanly. This is standard practice for new drugs. But an oxygen carrier is a particular case, because its most likely side effects, changes in blood pressure, in organ function and in laboratory tests, are exactly the ones that a healthy volunteer will experience without benefit.
The ethics therefore turn on transparency and on choosing an appropriate dose. Volunteers must be told clearly that they will receive an unapproved product, that it will not help them, and that measurements will be taken to look for harm. The doses should be low enough to make serious injury very unlikely and high enough to be informative. Some carriers turn the urine or plasma a strange color, and volunteers should be prepared for that. And follow-up should last long enough to catch delayed effects, particularly for products that are cleared slowly by scavenger cells.
There is a further consideration that applies to patient advocacy. When a foundation supports a trial, it should ask who is being enrolled, how they are paid, and whether the payment is proportionate. A volunteer trial that recruits from vulnerable groups because they need money is ethically different from one that recruits broadly, and the difference shows up in the quality of the consent.
It helps to line the three programs up beside a bovine polymer, on the problems that the whole field cares about.
On nitric oxide scavenging, the package helps: a membrane keeps the protein away from the wall. The polymer relies on size and chemistry to reduce the effect, and only partly succeeds. The cultured cell solves it completely.
On kidney leakage, all the packaged particles are large enough to avoid filtration. The polymer manages by raising its average size, with the small end of the distribution as a lingering risk.
On iron oxidation, the package can carry protective enzymes and reducing agents. The polymer is exposed and depends on the manufacturing chemistry and on storage conditions.
On immune recognition, the vesicle and the nanoparticle use polyethylene glycol to reduce it, while the bovine polymer relies on the protein’s chemistry and on the fact that hemoglobin is a familiar molecule. The cultured cell is the most natural but requires blood group matching.
On shelf life and convenience, the vesicle and the freeze-dried nanoparticle aim at room temperature. The polymer already achieves it. The cultured cell is the least convenient.
On maturity, the polymer leads by decades of human data. The vesicle is at the start of human testing. The nanoparticle and the cultured cell are earlier still.
That last line is the crux. The bare polymer is the least elegant design and the most tested one. It has been given to thousands of patients, manufactured at scale, sold in three markets, and used in emergencies. The elegant packages have not.
I want to resist the temptation to conclude that a package is simply better. It may prove to be, in the long run. But a patient reading this in the near term should recognize two things.
First, the packaged carriers inherit a new set of problems, which are not zero. The immune system is especially wary of lipid and polymer particles, and there is a known reaction to some such particles, in which complement activation produces a sudden, sometimes dangerous response on first exposure. Animal studies with hemoglobin vesicles have found no anaphylactic reactions in rodents, which is encouraging but does not settle the question in humans. The reticuloendothelial system may be burdened by repeated doses of particles. The long-term fate of the lipids is not fully described.
Second, the timeline is different. A bovine polymer exists. It can be manufactured today and was administered to patients on a compassionate basis. A packaged carrier is a development program whose completion is measured in years. For a patient who needs an option now, the difference is decisive.
There is a reasonable middle view. The bare polymer may be the bridge. It occupies the space until something better arrives, and the experience gained with it, in manufacturing, in regulation, in trial design, in measurement, in the ethics of emergency access, will be inherited by whatever succeeds it. Every technology in this essay will draw on the path that the polymer cleared, or failed to clear.
For a reader who wants to follow this field with a critical eye, here are the markers I would track.
For the vesicle: results of larger human trials, data on repeat dosing, and the identity of a manufacturer with the capacity to produce at scale.
For the synthetic nanoparticle: published human data of any kind, and independent replication of the animal results.
For cultured cells: the yield per donor cell, the cost per unit, and the number of patients treated outside a trial.
For the polymer: the outcome of any new trial, the continuity of manufacturing and the pattern of adverse events in expanded access.
None of these will arrive quickly. All of them will be visible in the public record if someone is paying attention.
Evolution solved the oxygen problem by putting hemoglobin in a cell, and most of what has gone wrong in the history of oxygen carriers can be read as the consequences of taking it out. The three programs here attempt to put it back in some form. It is entirely possible that the future of the field lies in a package, and equally possible that the bare polymer, refined and used in the right niche, will serve for decades. The prudent position for patients is to support the science that tests both, and to insist that each be judged on the same terms: what it does, in whom, compared with what.
The hemoglobin vesicle work is described in reviews by its developers, including a 2022 article on storable, ready-to-use artificial red blood cells for emergency medicine and a first-in-human phase 1 report in Blood Advances the same year. A study of hemoglobin vesicle injection in rodents reported an absence of anaphylactic reactions. Programs on synthetic and cultured red cells are described in the recent literature and in press releases from the research groups involved. Claims about progress in these programs should be checked against primary publications, since I have not verified trial results for any of the three.
There is a sentence in the press materials of nearly every company that has tried to sell a bovine hemoglobin oxygen carrier: it is approved in such-and-such a country. The sentence is true. It is also one of the most misunderstood facts in the field, because an approval is not a universal certificate of quality. It is a decision made by one regulator, under one set of rules, on one body of evidence, for one specified use, at one moment in time.
This essay is about that map. It describes how approval, restriction and exceptional access differ, what the pattern for BHOC products actually looks like, and how a patient community should read a claim of approval elsewhere. I will try to be careful about what I can and cannot confirm, because the details vary with the source and this is a subject on which loose language does real harm.
An approval is a regulator’s judgment that a product’s benefits outweigh its risks for a defined population under defined conditions. Each part of that sentence hides a variable.
The judgment depends on the evidence submitted. A regulator in a small market may accept data from trials done elsewhere, may require local studies, or may weigh a company’s dossier differently from a regulator in a large market with a long history of assessing similar products.
The benefit-risk balance depends on what alternatives exist. A product that would be marginal where safe blood is plentiful may be valuable where it is scarce. Regulators in different countries face different alternatives, and they legitimately reach different conclusions.
The defined population and conditions matter enormously. An approval to treat anemia in adult surgical patients does not authorize use in trauma, in children, or in pregnancy. The label is a fence.
And an approval is a snapshot. It can be revised, restricted or withdrawn. It says nothing about how the product has fared since.
When a reader sees the phrase approved in a country, each of those questions is still open.
Here is what I can state with reasonable confidence, drawing on a company’s own description and on published histories, and flagging where these sources differ.
The veterinary product, Oxyglobin, was approved in the United States and Europe in the late 1990s, and its current owner describes it as approved for over twenty years, with more than 240,000 units sold and off-label use in about thirty species. That is a company statement, not an independent audit, but it is consistent with the general history.
The human product, Hemopure, was approved in South Africa in 2001 for the treatment of acute anemia in adult surgical patients. That approval is reported consistently across sources. It was also approved in the Russian Federation. The company’s website gives the year as 2010 and describes the indication as acute, all-cause anemia. A published historical review states the same approval without giving a year, and one encyclopedic source reports the approval in a slightly different way. I mention the discrepancy because it illustrates a point: even the basic facts of approval are recounted with variations, and a careful writer should cite the regulator’s own record, which I have not seen.
In the United States, Hemopure has never been approved for human use. The Food and Drug Administration placed a hold on a proposed trauma trial in 2003 and did not approve an application for orthopedic surgery that year. Since then the product has been available under the agency’s expanded access mechanism, used to treat patients with life-threatening anemia who cannot receive blood. A registered expanded access protocol exists for that purpose.
I am not aware of any approval of a bovine hemoglobin carrier for human use in the European Union, though I would not be surprised to learn of specific national authorizations or named-patient arrangements, and I have not searched for them exhaustively.
Two features of this map stand out. It is small: two countries, one indication. And it was not built through the largest, most demanding regulators. That does not make the approvals meaningless. It means they carry a particular kind of evidence.
There are several reasons a company might pursue approval in a smaller market first, and none of them is sinister.
The regulatory path can be shorter and cheaper. A company running short of money has a strong incentive to reach any market where it can start to earn revenue and gather data.
The need can be greater. In a country with a limited blood supply and a high burden of trauma or of diseases that cause anemia, the argument for an alternative is stronger, and a regulator can reasonably weigh it that way. South Africa, with both a high rate of trauma and constraints on blood availability, is a plausible example, although I would want to see the regulator’s own reasoning before saying so.
Some regulators have procedures for products that meet an unmet need, allowing conditional or provisional authorization on less complete evidence than the largest agencies require, with post-marketing obligations to gather more.
A first approval can provide a base for further trials. The product can be used in a real health system, adverse events can be recorded, and a supply chain can be tested. In principle, that is valuable. In practice, the value depends on whether the data are actually collected and published.
A frequent question from patient groups is why the United States does not simply accept an approval from another country. The answer is that a regulator’s obligations are to its own population and its own standards.
An approval abroad is a piece of evidence. It shows that another agency reviewed the product and found the benefits to outweigh the risks in its context. If the reviewing agency is rigorous, and the data are complete, and the trial population resembles the domestic one, the approval carries weight. American law permits foreign clinical data to support an application under certain conditions, principally that the data come from well-conducted studies and apply to the American population.
But the approval is not a substitute for the application. The regulator here will ask about the same issues it would ask about any product: the quality of the manufacturing, the completeness of the safety database, the adequacy of the trials, and above all the comparison with the standard of care. If the standard of care in the reviewing country differs from the standard here, the comparison may not transfer.
There is also history. The class of products has a record that includes trials with excess deaths and a pooled analysis that raised alarms. A regulator looking at a new application inherits that memory and will look harder at the data than for a product from a class with a cleaner history. This is prudent, not vindictive.
Between approval and prohibition lies a range of exceptional access mechanisms, and understanding them is essential for a patient community, because most people who receive an unapproved oxygen carrier do so through one of them.
In the United States the agency’s expanded access framework, sometimes called compassionate use, permits a patient with a serious or life-threatening condition, and no satisfactory alternative, to receive an investigational product outside a clinical trial. There are several forms. An individual patient request is made by a treating physician, who obtains the manufacturer’s agreement to supply the product and submits a request to the agency. In a true emergency the agency can authorize use by telephone and the paperwork follows. Larger forms cover intermediate-size groups of patients and broader treatment protocols.
Each form requires the agreement of several parties: the manufacturer must be willing to supply, the physician must be willing to take on the responsibility of oversight, the hospital’s ethics committee generally must review, and the patient or a surrogate must consent. Any of them can say no, and the product is not owed to anyone.
Other countries have their own arrangements, with names like named-patient supply, special access and authorization for unlicensed medicines. Their rules differ in what they require and how fast they move. In some, a physician can import a product for a single patient with relatively light paperwork. In others, the process is closer to a formal application.
From a patient’s point of view, the key features of these routes are speed, cost and certainty. They can be fast, but only if everyone involved knows what to do. They can be expensive, because the manufacturer may charge and the hospital may bear administrative costs. And they are uncertain, because the answer is discretionary.
Here is the difficulty that connects the map back to the science. The more a product is used through exceptional routes, the less it is studied.
Expanded access is not designed to generate evidence. Patients are not randomized, outcomes are not standardized, follow-up is uneven, and reporting is voluntary. A collection of case reports, however moving, cannot show that a product works. Patients who receive the product are, by definition, in desperate circumstances, and some will die whatever is given. Others will survive for reasons that have nothing to do with the treatment.
A registry that recorded each case in a common format would make the record more useful. It would specify what was collected, in whom, and how outcomes were defined, and it would be reported in aggregate. The FDA has encouraged data collection from expanded access, and companies sometimes publish series of cases, but the result is a scatter rather than a dataset.
I raise this as a reason for patient groups to think of exceptional access as a responsibility as well as a right. A group that helps a patient obtain a product under expanded access might also help ensure that the case is recorded, with consent, in a way that can be pooled with others.
One event in the recent record shows how the map can be used constructively. A press release announced that a trauma trial in the prehospital setting would be sponsored by the United States Department of Defense and coordinated by Stellenbosch University in South Africa, evaluating a bovine hemoglobin carrier. The arrangement is instructive on its face: an American funder interested in military applications, a South African academic center with experience in trauma care and a product already approved there, and a trial design that could produce data relevant to both.
I have not found a completed report of that trial in the sources I could reach, and I will not describe results I cannot cite. What I can say is that the structure of it is sensible. An approved product in one jurisdiction can be tested under the local approval, in a setting with a genuine need, generating data that might eventually support an application elsewhere. It is a model that other programs could follow, and one a patient organization could encourage.
Words on this map are slippery, and patient communications should be exact.
Approved should be followed by a country, an indication, and a year, and by the source of the statement. Available under expanded access should not be shortened to available. Used in South Africa should not be turned into proven safe. Studied in a Department of Defense trial should not be translated into endorsed by the military.
When a company says a product has been used in many patients, ask how many, in what setting, under what protocol and with what follow-up. Numbers without denominators and without outcomes are not evidence.
And when a sentence begins with the phrase the FDA has not approved, it is worth completing it with what the FDA has actually done: held a trial, declined an application, or permitted use under expanded access. Each of those is a different fact.
If I were drawing the ideal map for the community, it would have six columns for each product and each country: the regulator’s name, the year, the indication, the legal status, the type of evidence relied on, and a link to the regulator’s own document. It would be maintained by someone with no stake in the outcome. It would state what is unknown.
Such a map does not exist in one place. Building it would be a service to patients, clinicians and journalists alike, and it is well within the capacity of an advocacy organization with the patience to read regulatory records. It would also protect the community against a common failure of the field, which is repeating a favorable sentence until it becomes a belief.
It helps to picture what a regulator actually reads, because the word approval compresses a mountain of paper into one syllable. A submission for a biologic product typically has four large parts.
The first covers quality and manufacturing. It describes the source material, the herd or the donor pool, the steps of the process, the tests done on every batch, the limits set for impurities, and the evidence that the process is consistent from one batch to the next. For a bovine product this is where the animal-health documentation lives, and where the viral and prion clearance studies described in another post are submitted. Many products that are scientifically interesting founder here, because a process that works in a laboratory must be shown to work the same way every time at scale.
The second part is nonclinical: studies in animals and in the laboratory that show what the product does, how long it lasts, which organs it reaches, and what happens at doses well above the intended one. Regulators look especially at the kidney, the heart and the vessels for a hemoglobin product, given the history.
The third is the clinical section, the trials in people. For a product aimed at emergencies it is rarely as tidy as for a drug taken daily for a chronic condition. Trials may be small, may enroll a mixture of patients, and may use comparators that reflect local practice.
The fourth is the proposed labeling and the plan for monitoring after approval: what the leaflet will say, who may use the product, and what data the company promises to collect once it is on the market.
Each part can pass or fail on its own. A company can have persuasive clinical data and be refused because its manufacturing was inconsistent. It can have clean manufacturing and be refused because the trials could not distinguish benefit from chance. When a regulator says no, the reason can be found in the response letter, and a patient advocate who reads it learns more about the state of the product than from any press release.
Suppose you receive a message, a brochure or a news story that says a product is approved somewhere. Here are the questions I would ask, in order.
Who approved it? The name of the regulator, not just the country. Some agencies issue approvals that other authorities regard as thorough. Others act more quickly on less.
For what? The indication as written on the label, in the label’s own words. Anemia in surgery, anemia from any cause and trauma are different claims.
For whom? Adults or children, hospital or field, with or without blood available.
When? A decade-old approval may predate newer information, and an approval can be quietly allowed to lapse when a company stops supplying a market.
On what evidence? Randomized trials, uncontrolled series, or a decision based on prior approval elsewhere.
With what conditions? Some approvals come with obligations to run further studies. Did the studies happen?
Is it actually available? A product may be approved and yet not be manufactured, not be stocked, or not be affordable. Approved on paper and available in a hospital are two separate states.
What has happened since? Have there been safety communications, restrictions or withdrawals?
None of these questions requires expertise. Each one converts a comforting phrase into a set of specifics, and the specifics either hold up or they do not.
I want to end the substantive part of this essay on a practical worry. A fragmented map has a cost that falls on patients and that no single regulator is responsible for.
When each country decides independently, developers must repeat work. Trials are designed to satisfy different agencies, files are prepared in different formats, and small companies exhaust their resources on parallel submissions. In a field already short of capital, duplication is lethal.
Regulators in different countries do have mechanisms to cooperate, including shared inspections, mutual reliance on each other’s reviews in some circumstances, and international guidelines for the content of dossiers. These tend to work best among large, well-resourced agencies, and less well for the small markets where a niche product might first be approved.
A patient organization cannot rewrite these systems. It can help in two ways. It can push for common data standards, so that the outcome of an emergency use in one country can be read in another. And it can support trials whose designs are acceptable to more than one regulator, so that a single study can serve several applications. Neither of these is glamorous. Both would reduce the number of patients who wait while paperwork repeats itself.
Approval is a door, not a verdict. A product that has passed through one door has not thereby passed through all of them, and a product that has not passed through the largest one has not thereby failed. The fair way to read the map is as a set of partial answers to the same underlying question, asked by different people with different stakes.
The company’s description of its approvals and expanded access program appears on the website of HbO2 Therapeutics. A narrative review of the history of hemoglobin-based oxygen carriers, published in Annals of Blood, states the approval of Hemopure in South Africa in 2001 and in Russia and describes its use in the United States under expanded access. Details of the Department of Defense sponsored prehospital trial were announced in a press release from the sponsor and coordinating center. Readers should consult the regulators’ own databases for the current status of any product, since approvals change and secondary sources vary.
A foundation that campaigns for a medical technology holds a strange kind of power. It does not make the product, it does not regulate it, and it does not prescribe it. What it does is shape the sentences that patients, families, clinicians and legislators repeat to one another. Those sentences travel further than any trial report. They decide who becomes curious, who becomes frightened, and who decides the topic is not worth a phone call.
This essay is about the craft of those sentences. It is aimed at anyone writing about BHOC for a general audience: advocates, volunteers, journalists and the patients who post about it. It is practical, and it makes an argument that some readers will find uncomfortable: that the most effective way to advance a promising but unproven technology is to be visibly, boringly, consistently honest about what is not yet known.
Three features of the BHOC field raise the stakes of language.
First, the history is bad. Products in this class have been associated with harm in trials, and a pooled analysis in 2008 combined sixteen trials of several products and found a higher risk of death and of heart attack in treated patients. Any reader who has heard about that history arrives with suspicion. Careless enthusiasm confirms the suspicion.
Second, the audience is vulnerable. People who search for an oxygen carrier are often people or families in trouble: someone who cannot receive blood, someone in a long illness, someone frightened. They are inclined to hear hope in any sentence that offers it. That inclination puts a moral duty on the writer.
Third, the science is subtle. Terms like oxygen affinity, cell-free hemoglobin and vasoconstriction have precise meanings, and their casual use creates false impressions. A word like artificial blood, for instance, suggests a complete replacement for something that a bovine hemoglobin solution is not.
The simplest tool I know for honest communication is to sort every claim into one of three tiers and to mark the tier in the sentence.
Tier one is what is established. Things that have been measured, in humans, reproducibly, and reported in the peer-reviewed literature. For example: a bovine hemoglobin carrier has been approved for anemic dogs, and has been approved for human surgical anemia in South Africa. Those are facts with sources.
Tier two is what is plausible. Claims that follow from mechanism or from animal data or from small studies, and that have not been confirmed in humans. For example: a cell-free carrier might reach tissue that red cells cannot when a vessel is narrowed. It is a reasonable hypothesis, supported by physiology and by experiments, and it is not proven as a clinical benefit.
Tier three is what is hoped. Aspirations about what the technology might one day do. For example: a shelf-stable carrier might someday allow emergency oxygen delivery in remote places. That is a goal, and it should be written as one.
The rule is that no sentence may present a tier-two or tier-three claim in the grammar of a tier-one claim. The words that carry the tier are small: is, may, might, has been shown, is hoped. Their precision is the whole point. A page that swaps them casually, turning may into does, has crossed from advocacy into promotion.
Patients are asked to reason about risk, and the way numbers are presented can push their reasoning in either direction.
Take the 2008 pooled analysis. It is often summarized as showing a thirty percent increase in the risk of death and nearly a threefold increase in the risk of heart attack. Those are relative risks. A relative risk is a ratio, and a ratio alone says nothing about how many people are affected. If the baseline risk of an event is one in a thousand, tripling it gives three in a thousand: a large relative increase and a small absolute one. If the baseline is one in ten, tripling it gives three in ten, which is a catastrophe.
The honest presentation gives both numbers: the relative change and the absolute counts, in the same population, over the same time. Statisticians who study risk communication have found that people understand natural frequencies, such as three out of a thousand, much better than percentages or probabilities. A sentence like out of every thousand patients treated, this many more had a heart attack is clearer than a hazard ratio.
The same care applies to favorable numbers. A company may report that most patients receiving a carrier survived. Most compared with what? A group of patients who could not receive blood, treated under expanded access, will have a survival rate that reflects the severity of their condition and the selection of who was chosen. Without a comparison group, survival numbers are not evidence of benefit. They are descriptions.
I would propose a plain rule: any number about outcomes should be accompanied by its denominator, its comparison, and its source, in the same sentence or the next one. If any of the three is unavailable, say so.
Some words are especially treacherous.
Artificial blood. The product is not blood, and it does not do what blood does. It carries oxygen, and little else. It does not clot, it does not fight infection, and it does not last. Blood substitute has the same problem. Oxygen carrier or oxygen therapeutic is more accurate.
Safe. A word to avoid. Safe compared with what, at what dose, in whom, and for how long? A product can be reasonably safe for a defined use and unsafe for another. Say what was observed.
Approved. As described in another post, this needs a regulator, a country and an indication.
Miracle, breakthrough, revolutionary. Words that belong on a marketing page, not a fact sheet. They cannot be verified, and they attract exactly the reader you are trying not to mislead.
Cure. A carrier treats a condition, at best. It does not cure anemia or a disease of the vessels, and calling it a cure is untrue.
Natural. Bovine hemoglobin comes from an animal. It is purified and chemically modified. The word invites assumptions about safety that the science does not support, and it is irrelevant to whether the product works.
Not a word but a habit: the passive voice that hides who found what. It has been shown that is weaker than a 2021 randomized trial in the New England Journal of Medicine found that. The reader deserves to know who said it.
An advocacy organization that works with a developer, receives support from a company or shares goals with one, has an obligation to say so at the start of every relevant communication, not at the bottom of a page.
The obligation is not about wrongdoing. Collaboration is normal and often essential. It is about giving readers what they need to weigh what they are told. A sentence like this organization is working with the developer on outreach and has not received payment for its statements, or has received support, is short and settles a question that would otherwise linger.
Readers are more forgiving of a disclosed relationship than of a discovered one. In fields with a history of financial misstatement, as this one has, credibility is a scarce resource and disclosure is how it is earned.
Some of your readers will be people in crisis. They need a different register from an analyst.
Lead with what they can do. If you are describing a route to access, put the practical steps first: whom to ask, what documents are needed, what the physician must do. Then give the context.
Be gentle about prognosis. You do not know their case. Avoid statements that predict, and refer to the treating team for anything specific.
Avoid pressure. Never suggest that a patient should demand a specific product. The decision belongs to the patient and the clinician together, and the person on the other end of the page may have a doctor who knows things you do not.
Say plainly that the product is not available everywhere, may not be appropriate, and carries risks. A short paragraph of this kind, placed high on the page, is not a legal disclaimer. It is respect.
Others will be doctors, scientists, reporters and officials who have seen bright promises fail. They read looking for the overreach.
Give them sources. A claim followed by a reference can be checked. A claim without one is a tell.
Give them the bad news yourself. Name the 2008 analysis, the trials that failed, the companies that collapsed. Do it before they do. A page that acknowledges the record earns the right to be heard on the rest.
Give them the gaps. State what is not known. An honest description of the missing evidence is more persuasive to a skeptic than a stack of encouraging anecdotes.
And leave out the adjectives. A skeptical reader treats adjectives as noise.
To make this concrete, here is the structure I would use for a one-page fact sheet, described in words so that it can be adapted.
Begin with a single sentence that says what the product is and what it is not. Follow with a short paragraph on where it stands: which regulators have acted, in which countries, for what use, and its status in the country of the reader. Then a paragraph on what the evidence shows, tier by tier, with sources. A paragraph on the known risks and unknowns. A paragraph on how a patient might access it, if at all, and what to ask a clinician. A line of disclosure about the relationship of the publisher to any developer. And a date, because the facts change.
Keep it to a page. If it is longer, most readers will not finish it, and the ones who do not finish it will remember the first paragraph.
An example makes the difference tangible. Imagine a small study of a carrier given to twenty patients who could not receive blood, of whom fifteen survived to hospital discharge. This is invented for illustration and does not describe any real study.
The promotional telling: Seventy-five percent of patients survived thanks to the carrier, a breakthrough for people who cannot receive blood. Every clause is either unsupported or misleading. Thanks to implies a cause the study cannot show. Breakthrough is a verdict, not a finding.
The cautious but empty telling: A small study reported some survivors and some deaths, and more research is needed. Nothing false, and nothing a reader can use.
The honest telling: In a series of twenty patients who could not receive blood and who were given the carrier under expanded access, fifteen were alive at discharge and five died. The study had no comparison group, so it cannot show whether the carrier changed the outcome. Patients differed widely in the cause and severity of their anemia. The authors reported the following adverse events. The series suggests the product can be given in this setting; it does not show that it saves lives.
The third version is longer, and every sentence in it is one a journalist, a clinician or a regulator can check. It is also the version that survives a hostile question. Writing it takes discipline, but the discipline is mechanical: state the number and the denominator, state what is missing, state what the result does and does not show.
Advocates get asked questions they cannot answer. Is it safe for my father? Will insurance cover it? Should I ask the surgeon for it tomorrow?
The temptation is to say something reassuring. The better response has three parts. Say what you do know, briefly. Say what you do not know and cannot know from where you sit. And say who can help: the treating physician, the hospital ethics service, the manufacturer’s medical information line for an expanded access request.
There is nothing shameful in the second part. People who ask hard questions in hard moments usually know that no one can give them certainty. What they resent is being handled. A plain statement of the limits of your knowledge is a form of respect, and it is remembered.
It helps to prepare a short list of the questions you cannot answer and the referrals for each, and to keep it where every volunteer can see it. Consistency matters: two volunteers giving different answers to the same question is worse than either giving a modest one.
Social media compresses. A post of two hundred and eighty characters or a short video cannot hold a three-tier evidence framework. Yet it is where many readers first meet the topic.
I would suggest a few rules. Do not make a claim in a post that you would not make in a fact sheet. Link the sentence to the sheet, and say in the post itself that the link has more detail. Prefer questions to declarations: What would it take for a cell-free oxygen carrier to be tested in people who cannot receive blood? invites readers in without promising anything. Avoid superlatives and avoid the word safe. And never share a testimonial as evidence. A patient’s story can be told with compassion, but it should be labeled as a story, not a result.
When a post turns out to be wrong, correct it visibly and quickly. A short reply that says we got this wrong, here is the correct figure, and here is the source, costs a little pride and buys a lot of trust. Deleting quietly costs the trust and saves the pride, and it usually leaks.
Every organization that communicates about science makes errors. The measure of quality is not the absence of errors but what happens after. I recommend a few habits.
Keep a public corrections note. Date each entry, state what was wrong and what is right.
Date every page. A statement that was accurate a year ago may not be now, and readers deserve to know how old a claim is.
Assign someone to check citations. A second pair of eyes catches transposed numbers and misremembered years, the two most common errors in this kind of writing, and I have made both in the course of preparing these essays.
Invite outside review. A clinician, a statistician and a patient reading your draft before it goes out will each catch different things.
A last point, not about accuracy. The tone of a page carries information. Anger reads as certainty; a steady, curious voice reads as reliability. Patients have good reasons to be angry about the state of care, and advocacy is often born from that anger. It is a legitimate fuel. But when the goal is to convince a regulator or a clinician, the anger is best kept in the motivation and out of the prose.
This does not mean blandness. Vivid, specific writing, a real patient’s story told with consent, a well-chosen number, a clear picture of a mechanism, is the opposite of bland. It means that force should come from precision, not volume.
I have been arguing for restraint, and I should say what restraint gets you.
It gets credibility, which is the only thing that lets an advocacy organization be invited into rooms where decisions are made. Regulators, clinicians and legislators talk to groups that they trust not to embarrass them.
It gets better participation in trials. Patients who understand what a trial can and cannot offer enroll with realistic expectations, drop out less and report more accurately.
It gets protection. When something goes wrong in a field, as in any field, the organizations that overpromised are the ones that face the anger. The ones that were honest can say we told you this was uncertain, and mean it.
And it gets a longer life for the whole enterprise. The history of oxygen carriers is partly a history of enthusiasm outrunning evidence, followed by disillusionment and withdrawal of support. A field where advocates keep their claims small and their sources visible is a field less likely to boom and bust.
A final practical suggestion is to test your own materials on a reader who has no stake in the cause. Hand a fact sheet to a friend who is a nurse, a teacher or a mechanic, ask them to explain it back in their own words, and listen for the places where their version drifts from yours. Those are the sentences that need rewriting, and no amount of expertise substitutes for the discovery.
The paradox of advocacy is that the strongest statement is often the most modest. A sentence that says here is what we know, here is what we do not, and here is what we are doing to find out is not a weak sentence. It is the one that can survive contact with a skeptical reader on a bad day.
Words move patients. They can move them toward a decision they will regret or toward one they will understand. That responsibility does not go away because the cause is good. It grows.
Guidance on communicating risk with natural frequencies and on presenting absolute and relative changes together is available from the risk communication literature, including work by Gerd Gigerenzer and colleagues. The 2008 pooled analysis of hemoglobin carrier trials was published in the Journal of the American Medical Association. Readers preparing public materials should also consult their regulators’ guidance on communications about investigational products, which restrict claims of safety and effectiveness.
The history of oxygen carriers is usually told as a sequence of products: this one failed, that one was withdrawn, another went bankrupt. That telling is accurate and misses what is most useful about the past, which is the sequence of ideas. Each generation of researchers believed it understood why its predecessors had failed. Each was partly right and partly wrong, and the wrong part turned out to be the one that mattered. This essay follows the ideas rather than the companies, and it pays attention to the moments at which a mistaken explanation held a field in place for a decade.
I will be careful throughout about what the sources establish. The history here draws on a narrative review of hemoglobin-based oxygen carriers and on accounts of the Army’s research program, and where dates or numbers matter I give them as those sources do.
The idea appears early, and it is easy to see why. Hemoglobin is the molecule that carries oxygen in blood, and red cells are the awkward package that holds it. If a patient is bleeding and blood is unavailable, why not extract the hemoglobin and inject it?
The first recorded attempts date to the end of the nineteenth century. A German physician, von Stark, gave hemoglobin under the skin to anemic patients in 1898 and could not obtain a stable preparation. In the years before and during the First World War, researchers refined the preparations. A pair of American physicians, Sellards and Minot, infused hemoglobin solutions into thirty-three subjects, in volumes of a few milliliters up to about thirty. Every recipient passed hemoglobin in the urine, a visible sign that the kidney was clearing it. Yet thirty reported no overt harm. It was an encouraging result for the wrong reason: the doses were tiny, and the kidney’s reaction was mistaken for tolerance.
The first idea was thus simple and incomplete. It treated hemoglobin as a fuel and neglected the vessel, the organ and the immune system through which it would pass.
In 1949 a group led by Amberson published one of the most sobering cases in the whole literature. A woman had suffered a postpartum hemorrhage, compatible blood could not be found, and the physicians gave her a stroma-free hemoglobin solution, hemoglobin purified from red cells with the membrane fragments, the stroma, removed. The review I am following describes what happened: the patient improved initially, then developed oliguria, a sharp fall in urine output, and died of renal failure.
It is a haunting account because it holds both halves of the story. The hemoglobin worked: it raised her oxygen-carrying capacity while nothing else was available. It also destroyed her kidneys. The same observation would be made again, in animals and in people, for the next fifty years, and the explanation would change three times.
After the war, researchers noticed that the injury to the kidney seemed worse when the hemoglobin preparation was crude. The obvious suspect was the debris from broken red cells. Fragments of membrane are known to activate clotting and cause damage to small vessels, and a solution containing them could plausibly injure the kidney.
The idea produced a program of purification. Researchers in the 1960s, including Rabiner, showed in dogs that a well-purified, stroma-free hemoglobin could serve as an oxygen-carrying replacement in hemorrhagic shock and could sustain animals through exchange transfusions. The review describes his results as showing the solution to be lifesaving in dogs undergoing exchange transfusion. It looked as though the problem had been identified and solved.
Then came the human data. In the late 1970s, studies by Savitsky and colleagues gave purified stroma-free hemoglobin to healthy volunteers. The review is direct about the results: all developed oliguria with reduced creatinine clearance; two reported abdominal pain; seven developed a slow heart rate and high blood pressure. The kidney was affected even though the stroma were gone. The second idea was wrong, or at least incomplete.
The next explanation came from chemistry. A hemoglobin molecule outside a red cell is not stable as a four-chain unit. In dilute solution it splits into two-chain halves, called dimers, and the dimers are small enough to pass through the filter of the kidney. Once in the tubules they can precipitate, and they were held responsible for the kidney injury.
That explanation led directly to the chemical modifications that defined the next generation. If the tetramer could be prevented from splitting, the dimers would never form. Chemists therefore developed ways to cross-link the four chains together, and to polymerize several tetramers into larger assemblies. The diaspirin cross-link, which bridges the two alpha chains, was one of these. Glutaraldehyde polymerization, used in the bovine products, was another.
The third idea was much closer to the truth, and it produced the products that reached the clinic. But it too had an incomplete part. Cross-linking prevented the kidney injury from dimers, and yet the new products still raised blood pressure. The hypertension would need a fourth explanation.
At this point in the history a particular institution enters. The United States Army had a strong interest in a blood substitute that could be carried to a battlefield, stored without refrigeration, and given without typing. It supported research at the Letterman Army Institute of Research and elsewhere, and its scientists helped develop the diaspirin cross-linked hemoglobin. The review describes that program as a collaborative effort between the pharmaceutical company Baxter and the Army, and says that Baxter later continued alone.
A published account from the Letterman program bears a title that summarizes the state of the field as it stood: a review of modified hemoglobin research at Letterman, attempts to delineate the toxicity of cell-free tetrameric hemoglobin. The word delineate is honest. The researchers were trying to separate the effects of the molecule into parts, to say what came from the kidney, what from the vessels and what from contamination. They were doing what a good research program does, and it is instructive that after so many years they still framed the task as delineating toxicity.
The explanation for hypertension arrived in the 1990s and was, in retrospect, elegant. The lining of blood vessels makes nitric oxide, a gas that relaxes the smooth muscle in the vessel wall and keeps the vessel open. Free hemoglobin binds nitric oxide with extraordinary avidity, several orders of magnitude tighter than oxygen binds. When hemoglobin is loose in plasma, it can slip through gaps in the vessel lining, capture the nitric oxide near its source, and remove it. Without the gas, the vessels constrict, the pressure rises and blood flow to organs falls.
This account explained the hypertension, the reduced flow to some organs, the abdominal pain from constriction of the gut vessels, and much else. It also suggested a design principle: make the hemoglobin too big to leave the vessel lumen, and it will not reach the source of the gas. The bovine glutamers and the PEGylated products can be read as attempts to apply that principle.
Yet the fourth idea also turned out to be an incomplete story. Even large molecules seemed to produce some vasoconstriction, and it became clear that the gas is present in the lumen and not just in the wall. The hypothesis remains a leading explanation, and it is discussed elsewhere in this series. What I want to note here is the pattern: each generation’s insight was real, and each was mistaken in believing it complete.
Meanwhile, human trials proceeded, and their results were disturbing. The review reports the outcome of the trauma trial of the diaspirin product published in 1999: within twenty-eight days, twenty-four of the fifty-two patients who received the product had died, forty-six percent, against eight of forty-six, seventeen percent, who received saline. A trial of a polymerized human hemoglobin in trauma, published years later, reported thirty-day mortality of thirteen percent in the treated group against ten percent in the control group, a difference that did not reach conventional significance but pointed in an unfavorable direction. A pooled analysis in 2008, combining sixteen trials, reported increased risks of death and of heart attack.
These results have been analyzed to exhaustion. I want only to observe how they interacted with the ideas. Each product had been designed on the basis of the explanation then current. Each failed for reasons that the explanation had not anticipated. The pattern suggests that the theory was always a step behind the biology.
The field today holds a set of hypotheses rather than a single explanation. Nitric oxide scavenging is one. Oxidation of the iron, which produces free heme and reactive species, is another. The interaction of free hemoglobin with the endothelium and with platelets, the effect on the immune system, the role of oxygen affinity in triggering the arteriolar constriction described elsewhere on this site, the assay interference that clouds measurement, and the heterogeneity of the polymer mixtures are all in play.
What distinguishes the present from the past is the tools. Researchers can now measure nitric oxide directly in tissue, they can characterize polymer distributions precisely, they can image microvessels, and they can compare molecules head to head in standardized animal models. None of these guarantees success, but they raise the odds that the next explanation will be tested before it is bet on.
I draw four lessons from the history, and I offer them as instruction to anyone entering the field now.
Distrust the complete explanation. Every theory in the sequence was partial. The moment an explanation is presented as sufficient, it stops being examined, and the unexplained residue becomes the next disaster.
Human data corrects animal data. The dog results of the 1960s were sound and did not predict the human response. Species differ in kidney handling, in vascular reactivity and in immune response. A compound that behaves well in animals has passed the first exam, not the last.
Small doses mislead. The early human studies used doses that were too small to show harm and gave a false comfort. Toxicity that appears at high doses was not visible at low doses, and the doses that matter clinically are the high ones.
Institutions matter. The Army’s interest was a driver of the research, and its willingness to pursue the problem over decades produced knowledge that a commercial firm alone might not have generated. It also shows that a strong sponsor is not a guarantee of success. Both statements are true.
The title of this essay refers to trenches because the impulse behind the field has always come from war. Sellards and Minot worked in the era of the First World War, when the question of what to give a soldier who had lost blood at the front was pressing. The Army’s laboratory programs of the late twentieth century were driven by the same question. The current Department of Defense interest in prehospital trials is its descendant.
That lineage carries a moral. The people who wanted a battlefield oxygen carrier were not chasing a curiosity. They were trying to keep young people alive who otherwise would die. A century of failures has not reduced the strength of that motive, and it explains why the effort has continued through so many discouraging results. The history should discourage credulity and should not discourage persistence.
If the pattern holds, the next generation of researchers will discover that today’s explanations were incomplete. The question is how to be ready for that. I would suggest three practices drawn from the sequence above.
Design studies that can detect harm early, with adequate sample sizes, independent monitoring and pre-specified stopping rules.
Keep the mechanisms open. Do not treat any single explanation as the answer, and measure more than one candidate cause in every study.
Preserve the record. Much of the knowledge in this essay comes from a handful of review articles, and it is fragile. A patient community that supports an archive of the field’s history, with the original reports accessible, protects the next generation from repeating the last one.
It is worth pausing on the arithmetic of caution. A few milliliters of hemoglobin solution is a trivial fraction of what a hemorrhaging adult needs. The blood volume of an adult is roughly five liters, and a resuscitation that hopes to replace a meaningful share of lost oxygen capacity must deliver quantities that are orders of magnitude larger than the early experiments used. When the dose rises by a factor of a hundred, effects that were invisible at the low end can become dominant. The kidney can clear a small load without difficulty and may be overwhelmed by a large one. A vessel that shrugs off a trace of nitric oxide scavenging may clamp down when the scavenging is sustained.
This is a general feature of pharmacology, and it is easy to forget when a field is young and eager. The lesson for readers of the historical record is to ask, of any early safety claim, what dose it applied to and what dose the eventual use would require. A finding of no harm at a tenth of the intended dose is a finding about the tenth.
A second difficulty in reading this history is terminological. The same substance has been called stroma-free hemoglobin, cell-free hemoglobin, modified hemoglobin, hemoglobin solution, blood substitute and oxygen therapeutic, and the field has argued at length over which label is honest. A blood substitute implies that the product does everything blood does, which none of them do. An oxygen carrier implies a narrower and more accurate claim. In this series the term is BHOC when the source is bovine and the discussion is of the class, and the older names are used only where they belong to a particular historical report.
The naming matters because names shape expectations. Patients, physicians and regulators who hear substitute expect a replacement and are disappointed by a bridge. Those who hear carrier expect a limited tool and judge it accordingly. Much of the disillusionment recorded in the literature came from expectations set too high by the vocabulary.
Reading the review literature, one is struck by how much of the history depends on a few sentences repeated from paper to paper. The account of the 1949 case is short. The account of the volunteer studies is shorter still. It is entirely plausible that details have been lost or simplified in retelling, and a careful reader should treat every secondhand summary, including the ones in this essay, as an invitation to consult the original.
There is also a survivorship problem in the record. Programs that ended quietly, with a few disappointing animal experiments and no publication, left almost no trace. What we know is weighted toward the programs that reached humans or generated controversy. The true history includes many unpublished dead ends, and the reader should assume that the ideas visible in the literature are only the ones that were pursued far enough to be written down.
It is easy, in a narrative of ideas, to forget that each entry in the record was a person. The woman in 1949 had a baby and a hemorrhage and a physician who reached for the only thing available. The volunteers in the 1970s agreed to an experiment and lay in a laboratory while their urine output fell. The trauma patients in the 1990s were enrolled under an exception to informed consent, because they were too injured to agree, and the community consultation that accompanied such trials is a subject of its own.
Keeping those people in view changes the tone of the analysis. It becomes harder to speak of a failed generation as if it were a set of errors on a chart. The investigators were, on the whole, serious people trying to save lives, and the participants gave something real. The best way to honor both is to extract from the record everything it can teach, and to say so plainly.
For readers who want the sequence in one place, here is the outline, as the sources give it. In 1898 hemoglobin was first tried in anemic patients. In the era of the First World War, Sellards and Minot infused small volumes into a group of subjects. In 1949 the postpartum case ended in renal failure. In the 1960s dogs survived exchange transfusion with purified hemoglobin. In the late 1970s human volunteers developed oliguria with the stroma-free preparation. Through the 1980s and 1990s, chemical modification produced cross-linked and polymerized products, with Army and industry collaboration. In 1999 the diaspirin trauma trial reported its mortality difference. In 2008 the pooled analysis appeared. Bovine products meanwhile reached veterinary use and limited human approval in two countries. Each of these dates is drawn from the reviews cited below, and the reader is encouraged to verify them.
The history of hemoglobin-based oxygen carriers is reviewed in a narrative article published in Annals of Blood, on which the dates and case details above are based. The Army program is described in a report titled Review of Modified Hemoglobin Research at Letterman, published in Artificial Cells, Blood Substitutes, and Biotechnology. The 1999 trauma trial of the diaspirin cross-linked product appeared in the Journal of the American Medical Association, and the pooled analysis in the same journal in 2008. Figures quoted are from the review and should be checked against the original reports.
Nobody who works on bovine hemoglobin can avoid the question eventually. A skeptic at a conference, a journalist, a hospital administrator, or a patient’s relative will ask it plainly: how much cattle does this take? It sounds like a joke and is not one. The question contains three separate inquiries, which are technical (how much hemoglobin is in a cow, and how much of it can be recovered), economic (what does the raw material cost and who controls it), and moral (is it right to use an animal this way, and does the answer change if the use saves human lives). This essay works through the first honestly, gestures at the second, and treats the third with the seriousness it deserves.
A warning about method. I am going to build an illustrative calculation from round numbers. The numbers are assumptions chosen to be plausible and are labeled as such, and none should be read as a measurement of any real manufacturing process. The purpose is to teach the shape of the reasoning, so that a reader can substitute better numbers and see how the answer moves.
Start with the animal. An adult beef animal weighs several hundred kilograms. Blood makes up a small percentage of body weight, and I will assume, for illustration, that a full-grown animal carries on the order of tens of liters of blood. I will also assume that cattle blood contains hemoglobin at a concentration broadly similar to human blood, in the neighborhood of a dozen grams per hundred milliliters. Real values vary with breed, age, health and the way blood is collected, and a reader with access to a veterinary physiology text should replace these with better ones.
Not all of that blood can be collected. At slaughter, a substantial fraction is retained in the carcass and organs, and only part flows out during bleeding. So the collectable volume per animal is less than the total, perhaps by a large factor. Then the blood must be processed: red cells separated from plasma, washed, lysed, the hemoglobin purified from the debris, cross-linked and polymerized, filtered and formulated. Each step loses material. A realistic yield, from collected hemoglobin to finished product, might be well under one hundred percent, and I will treat it here as an unknown fraction, call it y.
Now the demand side. A product dose is measured in grams of hemoglobin. The one bovine product with human approval in two countries has historically been supplied in bags of a couple of hundred milliliters at a concentration of roughly thirteen grams per deciliter, which is a few tens of grams per bag, according to the company’s published descriptions. A patient who needs the product over a course of treatment may receive one bag or several. A veterinary dose is scaled to body weight and is typically a fraction of a bag.
Put the pieces together and the shape emerges. If one animal yields, after all losses, a quantity of finished hemoglobin equal to some number of bags, and a treatment course uses a few bags, then each animal supports a handful of treatments. That is the honest order of magnitude: not one animal per patient, and not a thousand patients per animal, but somewhere in a range that depends on the yield and the dose, and that a reader can compute for themselves once the assumptions are fixed.
The calculation above is sensitive to two quantities more than any others. The first is the collection fraction. If blood is collected under conditions designed for the purpose, with anticoagulant, clean handling and controlled temperature, more can be recovered per animal than in ordinary slaughter, where blood is a byproduct of little value and is handled accordingly. The second is the process yield, which depends on the chemistry and on how much of the purified hemoglobin ends up in the target polymer size range. A process that discards the undesired fractions wastes more raw material than one that recycles them.
Both quantities are improvable, and both have been the subject of engineering effort. A manufacturer that doubles the yield halves the number of animals per patient. That is why the supply question is not a fixed feature of the technology. It is a function of process engineering, and the engineers have every incentive to improve it.
There is a further complication that a reader should keep in mind: the raw material must meet safety standards that ordinary slaughter blood does not. Earlier essays in this series discussed the sourcing regime for cattle, the herd controls and the regulatory limits set in response to prion disease. Those requirements restrict which animals can be used and therefore restrict supply independently of the arithmetic of hemoglobin content. A herd that meets the standard may be small, and a product that depends on it inherits the size.
The natural comparison is with blood donation. A donor gives a unit of whole blood at a session, and a unit of red cells contains a quantity of hemoglobin that is in the same general range as a few bags of the oxygen carrier. One might therefore conclude that an animal is a rough equivalent to a number of donors. That comparison is misleading in two ways.
First, a donor is renewable. A healthy adult can give again after a couple of months, and a population of donors supplies a continuous stream. An animal is used once. The comparison should therefore be between the donor population’s annual output and the herd’s annual turnover, not between one donor and one animal.
Second, the product and the donation are not equivalent clinically. A unit of red cells carries oxygen inside cells that last weeks and interact with the vasculature in the way nature designed. The carrier is a temporary bridge with a short circulating life. One does not simply exchange the units one for one. The relevant question is what the patient needs during the window in which blood is not available, and that is usually a much smaller quantity of hemoglobin than a full replacement.
To get a sense of proportion, consider the scale at which cattle are raised and slaughtered. Beef production worldwide is measured in many tens of millions of animals per year. In any single country with a large beef industry, the annual slaughter runs into the millions. Compared with those figures, the number of animals that would be needed to supply an oxygen carrier for a specialized medical market is a very small fraction, unless the market becomes enormous.
I say this cautiously, because the fraction depends entirely on which market. A niche supply for trauma patients who cannot receive blood, and for veterinary practice, is one thing. A general substitute for transfusion across a health system would be another, and would require a supply chain on a different scale. The honest position is that the technology could supply the first from existing beef infrastructure, and that the second remains speculative until the product’s clinical role is established. I do not have a defensible number for the second, and I would distrust anyone who offers one with confidence.
The strongest ethical argument in favor of bovine sourcing is that the blood is already there. Cattle are raised and slaughtered for food, and their blood is largely a waste stream or a low-value input to animal feed and industrial uses. If it can be converted into a medical product that saves human lives, the argument goes, no additional animal is killed on account of the product. The medical use adds value to something that would otherwise be discarded.
There is force in this. It converts what looks like an extra harm into a use of an existing one. Someone who accepts the existing practice of raising cattle for food can reasonably accept the use of the blood.
The argument is also incomplete. A byproduct that becomes valuable changes the economics of the primary product. If blood becomes worth a substantial amount per animal, it slightly raises the value of each animal and thereby slightly supports the industry that produces it. At small scale the effect is negligible. At large scale it is not. And the argument does nothing for a reader who objects to the industry itself. For that reader the question is whether a life-saving use is sufficient to justify participation in a practice they reject, and the byproduct framing does not answer it.
Some readers will object from a very different direction, not that using cattle is wrong, but that it is unnecessary, because human donors and, eventually, engineered sources could do the job. Recombinant hemoglobin, produced in bacteria or yeast, avoids the animal entirely. Hemoglobin from outdated human blood avoids the herd. Hemoglobin from other sources, such as marine worms, has been proposed.
These alternatives are real, and several are discussed elsewhere in this series. What can be said here is that each has its own difficulties. Recombinant production at the scale of grams per patient is expensive with current methods, and the molecules require engineering to overcome the same toxicities that plague other cell-free hemoglobins. Human-derived hemoglobin depends on the donor supply that the technology is meant to supplement. The bovine molecule has the advantage that it has been studied for decades, that its oxygen release does not depend on the cofactor that human hemoglobin needs, and that the raw material is available at scale. It is not the only possible answer. It is the answer with the most history.
I want to say something about the moral status of the animal that is neither dismissive nor performatively grieving. Cattle are sentient, and the conditions in which they are raised and killed vary enormously. A responsible program would draw its blood from herds raised to high welfare standards, slaughtered by methods designed to minimize suffering, and audited by independent inspectors. Those requirements happen to align with the biosecurity requirements, since healthy, traceable animals are also the ones that meet the safety standards.
That alignment is fortunate, and it should not be mistaken for the whole of the ethical case. The deeper question is whether the interests of an animal can be outweighed by the interests of a person who would otherwise die. Most moral traditions answer yes in some form, and some answer no. The reader who says no is not being irrational, and a serious advocate of the technology should say so, and should respect the position rather than trying to talk the person out of it.
There is a practical dimension of the ethical question that mainstream discussions rarely mention. Patients differ in their religious and cultural attitudes to animal-derived products. Some avoid products derived from particular species, and in a country where cattle are held sacred, the use of a bovine product in medicine may be contentious. In others, the objection may concern the manner of slaughter or the status of the blood itself.
Hospitals that serve diverse populations already deal with these concerns for other medicines, including gelatin capsules, heparin from pigs and heart valves from cattle and pigs. The established practice is disclosure and choice: patients are told the source and can decline. For a product intended for emergency use, when a patient may be unable to speak, the difficulty is sharper. The consent essays in this series discuss it. The relevant point here is that supply arithmetic and cultural acceptability interact. A product that a significant share of potential patients would refuse has a smaller effective market than the herd could supply.
A last practical consideration is fragility. A product that depends on a specific herd, a specific plant, and a specific regulatory certification is exposed to shocks. A disease outbreak in the herd, a contamination event at the plant or a change in the regulation of cattle-derived material could interrupt supply for months. The history of the field includes companies whose entire operation rested on a single facility, and the closure of that facility ended the business.
A resilient design would diversify: several source herds in different regions, more than one manufacturing site, and a stockpile sufficient to bridge a temporary interruption. Each of those adds cost. In a low-margin market, the temptation is to skip them. The result is a technology that is theoretically abundant and practically brittle.
To make the reasoning concrete, let me walk through a hypothetical, using invented round numbers that a reader can replace. Suppose that the blood collected from one animal, under controlled conditions, contains hemoglobin equal to a hundred bags of finished product at the strength described earlier, before losses. Suppose that the overall yield from collected hemoglobin to finished, releasable product is one in three. Then one animal yields about thirty-three bags. If a typical course uses three bags, one animal supports about eleven courses.
Now vary the assumptions. If collection is half as efficient, the figure falls to five or six courses. If the yield doubles, it rises to twenty-odd. If a course uses only one bag, as in some veterinary cases and perhaps in some human bridge uses, the figure triples. The range from a handful to several dozen is not a scandal and not a miracle. It is a plain feature of the numbers.
The reader should keep two things in mind about this example. First, every input is hypothetical. I have not reported the actual yield of any manufacturing process, and a published figure from a company would supersede the assumptions above. Second, the interesting question is not the central estimate but the sensitivity. Whatever the true numbers are, the answer is likely to move by a factor of several with plausible changes in process, and therefore the honest response to the question in the title is a range with reasons.
Suppose the skeptic at the conference asks the question again. A decent answer would run as follows. Each animal supports a small number of treatments, the exact number depends on process yield and dose, and I would rather show you the calculation than quote a figure. The blood comes from animals raised for food, which changes the moral picture without erasing it. The supply is adequate for a specialized market and untested for a general one. And the alternatives, recombinant and human-derived, are worth pursuing in parallel, because a field that depends on one source is fragile.
That answer contains no false precision and no evasion. It also leaves the moral question where it belongs, with the questioner. The arithmetic can tell you how many animals. It cannot tell you whether a life is worth them, and pretending otherwise is a way of avoiding the argument.
There is a final reframing. The title asks how many cows is a life, and the question could be inverted. How many lives is a cow? If one animal supports, on the assumptions above, several patients whose alternative was death because compatible blood was unavailable, the ratio looks favorable to many people. If the patients would have survived anyway, or if the product does not reduce mortality, the ratio is meaningless.
That is why the clinical evidence matters more than the supply arithmetic. Supply is a constraint on scale, but benefit is what justifies the effort. A product that saves few lives per animal is still worth making if the lives are real and the alternative is nothing. A product that saves none is not worth a single animal. Everything else, including the yield calculations, is secondary to the question of whether the carrier works when it is needed. The essays elsewhere on this site examine that question, and this one has tried to make sure that the supply question is asked in the right order: after the evidence, not before it.
A reader evaluating a company’s claims about supply can put a short list of questions to it. How many animals are in the qualified herd, and how is that herd audited? What fraction of collected hemoglobin ends up in releasable product, and what happens to the rest? How many days of stock does the company hold, and what is the plan if the single plant goes offline? What is the cost of raw material per finished bag, and how much of the price is the herd and how much the process? Which independent body has inspected the welfare and biosecurity arrangements, and are its reports public?
None of these questions is hostile. A serious manufacturer will have answers, and the willingness to give them is itself informative. Vague answers about proprietary processes are a warning that the arithmetic has not been done, or has been done and is unflattering. In a field with a long record of promises that outran evidence, transparency about supply is one of the few things a patient community can reasonably demand.
The manufacturing and sourcing of bovine-derived oxygen carriers is described in the published literature on the approved bovine product and in the company’s own technical descriptions, which state the concentration and bag size assumed above. The regulatory framework for cattle-derived materials is set out in the United States Food and Drug Administration’s rules on bovine spongiform encephalopathy, and comparable rules apply elsewhere. Beef production statistics are published by national agricultural agencies and by the United Nations Food and Agriculture Organization. All figures in the worked example are illustrative assumptions, not reported data.
When a clot blocks an artery in the brain, the tissue it fed does not die all at once. That fact, which took decades to establish, is the whole reason anyone dreams of an oxygen carrier for stroke. In the region nearest the blockage, blood flow falls so low that cells die within minutes. Around that core lies a rim of tissue whose flow is reduced but not abolished, tissue that is silent but alive, held in suspension by a trickle of supply. That rim is called the ischemic penumbra, from the Latin for almost a shadow. Its fate depends on time, on collateral blood vessels, and on whether the blockage is cleared. If oxygen could be delivered to the penumbra by some route that does not depend on the blocked artery, the rim might be kept alive until the artery reopens.
This essay examines that hope. It is one of the most seductive applications of an oxygen carrier and one of the least proven, and the reasons it is both are instructive.
The concept came from experiments in animals, and the term was introduced in the late twentieth century by investigators who noticed that in the tissue surrounding a stroke core, electrical activity could be lost while the cell membranes remained intact. Silent cells that could still be rescued were a different category from dead ones. Later imaging studies in patients showed regions where perfusion was reduced and diffusion appeared normal, which was read as tissue at risk but not yet lost. The mismatch between those two images is the practical working definition of the penumbra in many stroke centers today.
Two points deserve stress. The penumbra is dynamic. It shrinks as the core expands, and in the absence of reperfusion it can be consumed in hours, in some patients much faster and in others over a day or more. And it is not a single sharp boundary. It is a gradient of flow and of injury, with regions closer to the core more likely to die and regions farther out more likely to survive. Any therapy aimed at the penumbra is aimed at a moving target with blurred edges.
The reasoning is straightforward. Blood carries oxygen in red cells, and a red cell is about the width of the smallest capillaries. In a narrowed or partially blocked vessel, cells may move slowly or not at all, and the oxygen they carry may not reach the tissue beyond. A solution of dissolved hemoglobin has no such constraint. It travels in plasma, and plasma can continue to move through channels too narrow for cells. Molecules of hemoglobin, being far smaller than cells, can slip past partial obstructions, deliver oxygen at the far side, and thereby supply the tissue in the shadow.
The idea gets extra support from the way the bovine carrier releases oxygen. Its oxygen affinity, as discussed earlier in this series, is set by chloride rather than by the cofactor that human red cells use, and its release profile in tissue differs from that of a red cell. In some experiments the carrier’s ability to offload oxygen readily has been proposed as an advantage where oxygen tension is low. Whether that translates into benefit in the brain is a question for data, not for argument.
Studies in rodents and other animals have examined hemoglobin-based carriers in models of focal cerebral ischemia. I will describe the pattern in general terms rather than quote specific effect sizes, because I would not want to misreport them. Several groups have reported reductions in infarct volume when a carrier was given during or soon after the occlusion, and some have reported preservation of function. Others have found that the carriers worsen outcomes in certain conditions, especially when the blood pressure rises sharply or when the molecule is small enough to cross a leaky barrier.
That variation matters. The results depend on the carrier used, on the dose, on the timing, on the species, and on the type of ischemia, whether a permanent blockage, a temporary one, or a hemorrhagic injury. Any claim that oxygen carriers help in stroke should be read in that light. The supporting evidence is a set of positive findings under specific conditions, not a general demonstration.
The brain is a poor place to tolerate a pressor effect. As earlier essays described, cell-free hemoglobin scavenges nitric oxide and can raise blood pressure. In a healthy circulation this is a nuisance, and in a hypovolemic trauma patient it may even be useful. In a stroke patient it is a hazard in two ways.
First, the vessels feeding the penumbra are already maximally dilated in an attempt to draw in more blood. A drug that constricts vessels can defeat that effort, reducing flow exactly where it is most needed. Second, a sudden rise in systemic pressure raises the risk of bleeding into damaged brain tissue, especially after clot-dissolving therapy. Stroke teams routinely manage blood pressure closely, and a product that pushes it up unpredictably would complicate that management.
Designers have responded by favoring larger molecules, which are less able to reach the vessel wall, and by testing lower doses. Those strategies may narrow the problem without eliminating it. The most careful position is that the pressor effect is a serious obstacle to use in stroke that would have to be shown to be manageable, not assumed away.
The single largest change in the treatment of stroke in recent years has been mechanical thrombectomy: a catheter is threaded to the blocked artery and the clot is pulled out. For patients with large-vessel blockages, this procedure has been shown in randomized trials to improve outcomes, and it has extended the window during which treatment can help, in selected patients identified by imaging. It has changed what an adjunct therapy has to prove.
An oxygen carrier for stroke has two possible roles in this landscape. It could serve as a bridge, given at the first point of contact, in an ambulance or a small hospital, to keep the penumbra alive during the transfer to a center that can perform thrombectomy. Or it could serve as an adjunct after reperfusion, to protect tissue during the vulnerable period after flow returns. Both roles are plausible, both are untested for carriers, and both must show benefit over a standard of care that has become very good.
The bridge role is the more compelling, because delay to treatment is the largest determinant of outcome for patients in remote locations. If a carrier could extend the salvage window by even an hour or two, it would be valuable. But the same delays that make the bridge appealing make the trial difficult, because it must be run in ambulances and small hospitals, where consent, imaging and monitoring are hardest.
Stroke physicians say that time is brain, and there is a widely quoted estimate that in a typical large-vessel stroke a great number of neurons are lost each minute that treatment is delayed. I will not repeat the figure, because it is an average across patients, calculated from assumptions, and individual patients vary enormously. What the phrase captures correctly is that the salvageable tissue diminishes as time passes.
What it hides is heterogeneity. Some patients have excellent collateral circulation, and their penumbra persists for many hours. Others have poor collaterals, and their window is short. A therapy that buys time will help the second group most, and the first group least. That suggests that a trial should select patients by imaging, enrolling those with a large mismatch and a poor collateral pattern, rather than treating stroke as a uniform condition. Trials that enroll everyone dilute a real effect in a crowd of patients who do not need the therapy or cannot benefit from it.
How would we know whether a carrier helps? The classic outcome in stroke trials is the functional score at ninety days, a coarse scale from independence to death, assessed by a clinician. It is a meaningful outcome, and it is noisy. Small differences in function are hard to detect without large numbers of patients.
Imaging offers an intermediate measure. The final infarct volume, measured on a scan several days after the event, indicates how much tissue was lost, and a therapy that reduces infarct volume would be a candidate for further trials. But infarct volume does not always track function, and regulators typically require a clinical outcome for approval.
A cleverer approach would use imaging of oxygenation itself: techniques that estimate tissue oxygen extraction in the penumbra before and after a carrier is given. If the tissue in the shadow really receives more oxygen, that should be visible. Such measurements exist in research settings, and are not routine. A phase two study built on them could give a direct answer to the mechanistic question, which is whether the carrier delivers oxygen where it is needed, before a large outcome trial is attempted.
The history of stroke research is littered with agents that protected the brain in animals and did nothing in humans. Dozens of neuroprotective drugs have failed in clinical trials, and the reasons have been analyzed at length: animal models that do not match human disease, treatment given earlier in animals than is achievable in patients, poor trial design, and underpowered studies. A serious proposal for an oxygen carrier in stroke has to reckon with that record.
The lessons apply directly. Test in models that resemble human stroke, including older animals and animals with other illnesses. Treat at times that are achievable in practice, not at the moment of occlusion. Use imaging to select patients. Pre-specify the outcome. Replicate in more than one laboratory. Each of those steps is unglamorous, and skipping them is how the field ended up with so many disappointments.
Not all strokes are blockages. In hemorrhagic stroke, a vessel bursts, and blood damages the tissue around it. The role of an oxygen carrier here is different and more doubtful. Adding a fluid that can raise blood pressure to a patient who is bleeding into the brain is a dangerous idea, and no reasonable protocol would do it without an imaging diagnosis to rule out hemorrhage. The practical consequence is that any stroke trial of a carrier requires a scan first, which is a further barrier to prehospital use. Mobile stroke units, ambulances with a scanner, exist in some cities, and they are the most natural setting for a bridge trial. They are also few.
Putting these pieces together, the state of the evidence is as follows. The biological rationale is sound: dissolved hemoglobin can reach places cells cannot, and the penumbra is a place where that might matter. Some animal experiments show benefit. Others show harm. The pressor effect is a real hazard in a condition where blood pressure control is central. The standard of care has improved sharply, raising the bar. And human trials of carriers in stroke, to the best of my knowledge, have not established benefit. I hold that last sentence with some humility, because the literature is large and I cannot claim to have read all of it, and I encourage the reader to check the current registries.
The right posture is neither enthusiasm nor dismissal. The idea deserves a proper test, designed with the lessons of the past in hand. It does not deserve premature claims.
If I were designing the first study, I would begin small and mechanistic. Enroll patients with a large-vessel blockage awaiting thrombectomy, selected by perfusion imaging for a substantial penumbra. Randomize them to the carrier or to standard care, under blinding where possible. Measure blood pressure continuously, with strict stopping rules if it crosses a threshold. Measure tissue oxygenation by imaging before and after. Compare the penumbra that survives at follow-up. The primary endpoint would be the amount of at-risk tissue that is saved, and the secondary endpoints would be function at ninety days and safety.
Such a study would be modest in size and would answer the question of whether the mechanism operates in humans. If it does, a larger trial could test clinical benefit. If it does not, the field would have avoided spending years on a hope. Either outcome is progress.
For a patient’s family, the message is cautious. No oxygen carrier is an approved treatment for stroke, and anyone who says otherwise is ahead of the evidence. Standard care, clot removal and dissolution where appropriate, remains what works, and speed to a capable center is what matters most.
For a researcher, the message is an invitation. The penumbra problem is one of the cleanest tests of the basic idea behind oxygen carriers, because the tissue in question is alive and starving and the oxygen has to travel a route that cells cannot. A carrier that fails there has a limitation worth knowing. A carrier that succeeds there has found its purpose. The experiment is worth doing carefully, and worth doing soon.
Stroke arrives without warning, and the patient often cannot speak. Language is among the functions most commonly lost, and the very act of consenting to a trial requires the faculty the stroke has taken. The usual solution is to seek consent from a relative, and a relative who has just watched a loved one collapse is in no state to weigh a research protocol. Investigators have long struggled with this, and the consent essays elsewhere in this series discuss the ethics of emergency exceptions and community consultation.
In stroke there is an extra wrinkle. Because the treatments that work best are the ones given fastest, every minute spent explaining a trial is a minute of brain lost. Trial designers have tried short consent scripts, video explanations, and deferred consent, in which treatment starts under an exception and the patient or family is informed afterward. Each has its critics. The key point for oxygen carriers is that a trial in stroke inherits all these difficulties and adds one more: the product has a history of harm in earlier trials, and a family told that a product has an uncertain safety record is entitled to weigh it carefully. Honest disclosure, in plain words, is part of the design and not an obstacle to it.
It is worth remembering that stroke outcomes are not decided in the first hours alone. Patients who survive spend months in rehabilitation, and their recovery depends on how much tissue was saved, but also on therapy, on age, and on whether they have other illnesses. A trial that reports a modest reduction in infarct volume must ask whether that reduction translates into a meaningful difference in a patient’s life: the ability to dress, to speak, to return to work. Patient advocates in stroke have pushed researchers to include measures that matter to survivors, and any oxygen carrier program should listen.
A paramedic asked to carry a stroke bridge would want answers to practical questions before anything else. Can the product be stored in the vehicle, and at what temperature? Room-temperature stability was one of the original attractions of the bovine carrier, and it is what makes the ambulance a plausible venue at all. Can it be given without knowing the blood type? Yes, in principle, because there are no red cell surface markers, which removes the delay of typing. And what must be checked before it is given? Here the answer is uncomfortable: at minimum, a rough assessment that the event is a blockage rather than a bleed, and a blood pressure that is not already high. Each requirement narrows the population who could receive it and lengthens the checklist, and each is a place where a well-meant protocol can fail in the field.
The solution, if there is one, is to build the checklist into the trial itself and to test it on simulated calls before real patients are enrolled. Emergency medicine has a long tradition of doing so, and stroke teams have refined their own pathways for years. An oxygen carrier trial should borrow that discipline and not reinvent it.
The concept of the ischemic penumbra is reviewed in the stroke literature, including articles in Stroke and The Lancet Neurology. Mechanical thrombectomy trials appeared in the New England Journal of Medicine and other journals from the middle of the past decade. Reviews of neuroprotective drug failures include the work of the STAIR committee on preclinical recommendations. Studies of hemoglobin-based oxygen carriers in animal models of cerebral ischemia are found in the journals of cerebral blood flow and of artificial cells and blood substitutes. Readers should consult the primary sources and current trial registries for the latest findings.
There is a peculiar fact at the center of radiation oncology that most patients never hear. The same dose of radiation kills cells far more effectively when oxygen is present than when it is absent. A tumor that has outgrown its blood supply, and contains regions of low oxygen, is therefore harder to cure with radiation than the same tumor would be if it were well oxygenated. The effect has been known for well over half a century, it has been measured repeatedly, and it has shaped the design of clinical trials from the middle of the twentieth century onward. It also creates an opening for an oxygen carrier, one that is quite different from the emergency uses that usually come to mind.
This essay explains the fact, describes the attempts to exploit it, and asks what a hemoglobin-based carrier could add. I will be careful in reporting numbers, and where I do not have a figure I trust I will say so.
When ionizing radiation passes through tissue, it damages DNA, partly by direct hits and partly through the chemistry of the reactive species it creates. The critical step is what happens to the damaged sites. In the presence of oxygen, the lesions are chemically fixed and become permanent. Without oxygen, the damage can be chemically repaired. The result is that hypoxic cells survive doses that would kill well-oxygenated ones.
The ratio between the dose needed in the absence of oxygen and the dose needed in its presence is called the oxygen enhancement ratio. For the kinds of radiation used in most clinics, it is commonly quoted as somewhere between two and three. In plain terms, a fully hypoxic cell may need two to three times the dose. I give the range as commonly cited in textbooks, and a reader who needs a specific value for a specific radiation type should consult a radiobiology reference. The practical consequence is that a small pocket of hypoxic cells in a large tumor can be the part that survives treatment and regrows.
Tumors are notoriously badly plumbed. They grow quickly, and their blood vessels are disorganized, leaky, and often too few. Two kinds of hypoxia are described. Chronic hypoxia arises from distance: cells located beyond the diffusion range of oxygen from the nearest vessel. A classic study in the middle of the last century estimated that range at roughly the order of a hundred to two hundred micrometers, and the number has been repeated in textbooks ever since. Cells beyond it are starved.
Acute hypoxia is different. It arises when a vessel, perhaps a poorly formed one, closes temporarily, and the tissue it fed loses oxygen for minutes to hours. Then it may reopen, and the cycle repeats. The two kinds behave differently, and this matters for any therapy that aims to reach the starved cells. A carrier that helps with chronic hypoxia must reach cells far from vessels. A carrier that helps with acute hypoxia must penetrate the network when flow is intermittent.
Because the problem is old, so are the attempts to solve it. The history is worth knowing, because a new carrier will be judged against it.
One approach is to make the patient breathe high concentrations of oxygen, in some protocols at increased pressure in a chamber. The rationale is that raising the oxygen carried in the blood will drive more into the tumor. Trials of this strategy in the 1950s and 1960s, using hyperbaric chambers, gave suggestive results in some cancers. The technique was cumbersome, radiation had to be delivered while the patient was in the chamber, and it did not become routine.
Another approach is to give a drug that mimics oxygen in the chemistry of radiation damage. These radiosensitizers, such as the nitroimidazoles, act in hypoxic cells. One of them, nimorazole, has been used in head and neck cancer in a Danish program, and a trial reported an improvement in local tumor control. I mention it from general knowledge of the literature and would encourage checking the primary report for the details.
A third approach combines breathing a carbon-dioxide-enriched oxygen mixture with the vitamin nicotinamide, which is intended to reduce the acute closing of vessels, in a regimen abbreviated in the literature by an acronym. Results have been mixed, with benefit in some tumor types and not others.
A fourth approach is to raise the hemoglobin level in anemic patients, because anemia is associated with worse outcomes after radiation. Transfusion and erythropoietin were tried, and the erythropoietin trials in particular raised safety concerns about tumor progression and thrombosis, which effectively closed that path in several cancers.
Each approach addresses part of the problem, none has solved it, and that persistent gap is what invites a cell-free carrier.
There is one episode that bears directly on the question, and it involves not an oxygen carrier but a molecule that changes how the patient’s own hemoglobin behaves. Efaproxiral, developed under the name RSR13, was an allosteric modifier: it bound to hemoglobin inside red cells and lowered its affinity for oxygen, so that the cells released more of their load in tissue. The idea was to increase the oxygen reaching tumors during radiation, without adding any new oxygen-carrying material to the blood.
It was tested in a large randomized trial of whole-brain radiation in patients with brain metastases. My recollection is that the trial did not meet its primary endpoint across all patients, and that a subgroup analysis suggested a benefit in patients with metastases from breast cancer, which led to a follow-up trial. I hold these details loosely, and a reader should consult the published reports rather than rely on my summary. What the episode showed, whatever its precise results, was that changing oxygen delivery to tumors in a controlled clinical setting is testable, that patients can be enrolled, and that the outcome measures exist.
It also showed that a subgroup signal is not a result. The field has a long record of disappointment after such signals, and the story is a reminder to design confirmatory trials rather than to celebrate a subset.
Here the logic returns to the carrier. Suppose a patient is given a hemoglobin solution shortly before a radiation session. The carrier circulates in plasma, and its small size relative to red cells allows it to enter the irregular, narrow channels that tumors often contain. In principle it could increase oxygen delivery to regions that cells reach poorly. Because it releases oxygen in a chloride-dependent way and does not depend on the cofactor inside red cells, it may offload readily at the low oxygen tensions found in tumor tissue.
A number of laboratory studies in the past few decades have explored this idea, with hemoglobin-based carriers and with perfluorocarbon emulsions, in animal tumor models. The pattern I understand from reviews is that tumor oxygenation could be increased measurably in some models, and that radiation response improved in some of them. I cannot quote effect sizes with confidence, and in the interest of not inventing numbers I will not try. I will say that the results were promising enough to sustain interest, and mixed enough to prevent conclusions.
Radiation therapy is delivered in daily sessions, typically over several weeks, and each session lasts minutes. Oxygen enhancement matters at the moment the beam passes. A carrier given for this purpose would therefore need to be present in the tumor, at an effective level, during each treatment, and its circulating life would have to be matched to the schedule.
That constraint is severe. Cell-free hemoglobin products have circulating half-lives measured in hours to a day or so, depending on the product and dose. A daily infusion, timed before each session, might be feasible. Thirty infusions over six weeks of a product with known side effects is another matter, particularly when the side effects include vasoconstriction and interference with laboratory tests.
An alternative would be to reserve the carrier for a few key sessions, or to use it in hypofractionated regimens, in which a small number of large doses are given. Modern radiotherapy increasingly includes such regimens, especially in stereotactic treatments, where a high dose is delivered in one to five sessions. In principle a carrier is better suited to a schedule of five sessions than thirty.
A cancer patient is not a healthy volunteer. Many have been treated with drugs that affect the kidneys, the heart or the blood counts. The known effects of cell-free hemoglobin, including its pressor action and its effect on the kidney, fall on organs that may already be under stress. The trial designers would have to screen carefully and monitor closely.
A second concern is more specific. Oxygen supports tumor growth, and a therapy that improves oxygenation might in theory help the tumor as well as the treatment. In practice, the evidence from anemia correction suggests that the link between oxygenation and outcome runs in the helpful direction when radiation is used, but the erythropoietin experience shows that interventions that raise hemoglobin can carry separate risks. A carrier is not the same as a growth factor, and it acts briefly, but the possibility has to be examined.
A third concern is the interference with imaging and laboratory measurement that these solutions cause. A patient on radiotherapy is monitored with blood tests and scans, and a colored, oxygen-binding plasma constituent can distort some of them. Clinicians would need to know.
The evidence needed is graded. The first step would be to show that a carrier increases tumor oxygenation in patients, using a measurement that can be done during care. Several exist: probes inserted into tumors in research settings, imaging methods sensitive to oxygen, and tracers that accumulate in hypoxic regions and can be seen on scans. A small study that showed a consistent rise in tumor oxygen after the carrier, in a tumor type with known hypoxia such as some head and neck and cervical cancers, would establish the mechanism.
The second step would be to show that the rise translates into biological effect: a greater fall in tumor size during treatment or a higher rate of complete response. These outcomes can be measured within months.
The third step is the outcome that matters, local control and survival, which take years and require large trials. Investors and regulators know this ladder well, and the field has stumbled on it before. The reason to lay it out is that each rung is achievable, and the first two are within the reach of a small academic group.
One point in favor of trying the idea in oncology is that the patients are planned. Radiotherapy is scheduled days ahead. There is no emergency, no need for prehospital use, no exception to informed consent. The patient can be told the truth, can read the information, can ask questions, and can decide. The consent problems that haunt trauma trials do not arise.
That makes oncology an unusually clean setting in which to learn about the product’s behavior in humans. Dose can be controlled, monitoring can be intensive, and follow-up is routine. If the carrier’s safety profile is going to be established anywhere, a planned setting is better than a chaotic one, and information gathered there would benefit every other use.
A knowledgeable skeptic would raise several points, and they deserve answers rather than dismissal. First, that hypoxia in tumors is often severe and located far from vessels, so that a plasma-borne carrier, which also depends on the vasculature to reach the tissue, might not get there. That is a fair objection to the idea for chronic hypoxia, and the answer must be empirical. Second, that modern radiation techniques deliver higher doses more precisely, and that some of the problem has been overcome by dose escalation. That is true for some tumors, and a residual advantage may remain for others. Third, that the pharmaceutical industry has little appetite for an adjunct with a difficult safety history. This is probably correct, and it means that early work is likely to come from academic groups and small companies.
None of these objections kills the idea. Together they define the conditions under which it might succeed: a tumor type with proven hypoxia, a planned schedule with few sessions, a carrier chosen for its large size and low pressor effect, and a study designed to measure oxygenation first.
This use is not the emergency drama that the public associates with blood substitutes, and that may be exactly why it is worth attention. It calls on a property of the carrier that has been under-exploited, the ability of small particles in plasma to reach places cells cannot, and it puts the property to work in a setting where the surrounding conditions are controlled. It would not save lives on a roadside. It might improve the odds for a patient whose cancer would otherwise be marginally resistant to the beam.
For the Foundation’s purposes, it also illustrates something general about the technology. The demand for oxygen is not confined to hemorrhage and shortage. It appears wherever tissue is starved, whether by injury, by clot, by cold storage or by a tumor that has outrun its vessels. A carrier that proves safe in any one of these settings opens the others. That is the unifying logic of the whole series, and the radiation question is one more place to test it.
To see why a small pocket of resistant cells matters so much, consider a simple illustration with invented round figures. Suppose a tumor contains a very large number of cells, and suppose a course of radiation kills a fixed fraction of the well-oxygenated cells at each session, leaving the rest. After many sessions, the survivors among the well-oxygenated cells shrink toward zero. Now suppose that a small proportion of the cells are hypoxic, and that each session kills a much smaller fraction of these, because of the oxygen effect. After the same number of sessions, the hypoxic survivors far outnumber the oxygenated ones, even though they started as a small minority. The tumor that regrows will be seeded by them.
The illustration teaches two things. It explains why hypoxia can control the outcome out of proportion to its share of the tumor. And it shows why the goal of an oxygen-delivering therapy need not be to oxygenate the whole tumor. Reaching even a fraction of the resistant cells could shift the balance, because the cure depends on the last cells standing. That is a more modest target than it first appears, and it is one reason the idea keeps returning.
A further biological fact helps. During a course of radiation, tumors often reoxygenate on their own. As oxygenated cells die, the tumor shrinks, vessels move closer to surviving cells, and some hypoxic regions become better supplied. This is one reason radiation is delivered in many small fractions rather than a single large dose: it gives time for that reoxygenation to occur between sessions.
Reoxygenation is uneven, and in some tumors it is slow or incomplete. A carrier could be viewed as a way of speeding or completing a process that biology performs unreliably. That framing is gentler than the picture of a carrier as an external supply, and it suggests that even a modest boost at the right moment might be worth having. It also suggests measuring: if reoxygenation can be tracked by imaging during treatment, the carrier could be timed to sessions where it is most needed, and not given routinely.
A patient facing radiation who reads about this should ask their oncologist three questions. Is my tumor type one in which hypoxia is known to matter? Is there a clinical trial of any oxygen-enhancing approach that I could join? And what would the risks be, given my heart, my kidneys and my other treatments? The honest answers today are likely to be that hypoxia matters in some cancers, that such trials are rare, and that the risks would need individual assessment. A good oncologist will say so plainly.
The oxygen effect in radiobiology is covered in standard radiation biology textbooks, including the discussion of the oxygen enhancement ratio and of chronic and acute hypoxia. The early work on the diffusion limit of oxygen in tumors dates to a mid-twentieth-century study by Thomlinson and Gray. Nimorazole in head and neck cancer was studied in the Danish DAHANCA program. Efaproxiral was studied in randomized trials of whole-brain radiation for brain metastases, reported in the oncology literature. Reviews of hemoglobin-based carriers and perfluorocarbons as radiosensitizers appear in the journals of radiation oncology and of artificial cells and blood substitutes. Numbers cited above are general textbook values and should be verified against primary sources.
Almost every study of an oxygen carrier is a study of a first exposure. A patient bleeds, receives the product once or over a few days, and is followed until discharge or for a set number of weeks. The scientific and regulatory record is built on that pattern. Yet the patients who might benefit most from a carrier are not always one-time users. Some have conditions that recur: sickle cell crises, repeated bleeding from a vascular malformation, recurrent anemia in a person whose religious beliefs preclude transfusion. Some will be trauma survivors who, years later, face a second emergency. For all of them the question is what happens on the second exposure to a protein that comes from another species.
That question has received far less attention than it deserves. This essay lays out what is known, what can be reasoned, and what has to be measured, and it tries not to pretend that the answer is settled.
The immune system meets a foreign protein in two ways. On first contact, it is a stranger, and the response takes days to develop. On the second contact, it is a familiar enemy, and the response is faster, larger and often of a different kind. This is the basis of vaccination and also of allergy.
Bovine hemoglobin is a protein from another mammal. It is similar to human hemoglobin in structure and different enough in sequence that the human immune system can, in principle, recognize it as foreign. Two things reduce the likelihood of a strong response. Chemical modification, such as the cross-linking and polymerization used in the bovine products, may alter the surface in ways that hide some sites. And mammalian hemoglobins are closely related, which may render the molecule less provocative than a protein from a distant species. Neither guarantees tolerance.
The other possibility is that the immune system does respond, producing antibodies that bind the carrier. Such antibodies could do several things. They could clear the carrier rapidly from the circulation, reducing its effect and its half-life. They could form complexes that deposit in the kidney or vessels. Or they could cause an allergic reaction of the kind that can be dangerous. Which of these happens, if any, is an empirical question.
The largest body of experience with repeated exposure comes not from human medicine but from veterinary practice, where a bovine carrier has been given to dogs and, off label, to other species for many years. The company that markets it has stated that the product has been used widely across species for over two decades. Some animals must have received it more than once, either in the same illness or in separate ones.
I do not have access to a systematic analysis of the immunological consequences in those animals, and I would be inventing if I claimed otherwise. What can be said is that the experience has not, to my knowledge, produced a widely known pattern of severe reactions on re-exposure, and that the labeling for the veterinary product has historically described it as intended for a limited number of infusions. A serious effort to answer the question would search adverse event reports, ask veterinary specialists about their experience with second doses, and, where possible, measure antibodies in animals that received repeated treatments. Such data may already exist in scattered form, and collecting it would be inexpensive relative to its value.
Human experience is more limited. Trials of bovine and other hemoglobin-based carriers included immunological monitoring in some cases. My understanding is that the general finding across several products was that antibody responses were infrequent and did not seem to produce clinical problems within the trials’ periods. I hold that statement loosely, because I cannot reconstruct the specific studies and because the trials followed patients for short periods after a single course.
The key limitation is the design. To see a memory response, you need a second exposure after a delay long enough for immunity to develop, typically weeks to months. Most trials do not provide that. A patient who was exposed once, and was followed for thirty days, tells us nothing about what happens at day ninety. This is not a criticism of the trials, which were designed to answer other questions. It is an observation that the specific question has not been asked.
Some features of the products matter for immunogenicity. Polymerization increases the molecular size, and larger particles are more likely to be taken up by immune cells that present them to the rest of the system. Cross-linking with glutaraldehyde creates new chemical structures that were not in the original protein, and such structures can act as haptens, small chemical groups that provoke an immune response when attached to a carrier. PEGylation, the attachment of polyethylene glycol chains, is used in some products to shield the protein, but antibodies against polyethylene glycol itself have been described in the general literature on PEGylated drugs, and they can accelerate clearance of such drugs on repeat dosing.
The details vary by product, and the conclusion is that each product must be evaluated on its own. A finding of low immunogenicity for one chemistry does not transfer to another.
There is also the matter of purity. Preparations that contain traces of other bovine proteins, from the plasma or the red cell membrane, present a wider range of targets to the immune system. The purification steps that remove these impurities are therefore also immunological safeguards. A change in manufacturing that lets more impurity through could change the immune profile without changing anything on the label.
Contaminants that are not proteins matter too. Bacterial endotoxin, the fragments of cell walls of gram-negative bacteria, is a potent stimulator of the immune system, and even small amounts in a product can cause fever, low blood pressure and inflammatory signaling. Such contamination can make an otherwise well-tolerated product seem immunogenic, and it can prime the immune system to respond more vigorously on subsequent exposures. Testing for endotoxin is standard in the manufacture of injectable products, and the limits are strict. For a product made from animal blood, collected in an environment that is not sterile in the way a pharmaceutical plant is, the discipline needed is greater.
Physicians who care for patients needing repeated transfusions know the problem of alloimmunization. A patient who receives many units of red cells may develop antibodies to minor antigens on donor cells, which make later transfusions more difficult to match. In certain populations, such as patients with sickle cell disease, this is a serious clinical problem that is managed by extended matching of blood.
The carrier’s promise to be free of such antigens is real: without cell surface markers, there is nothing to match. But that promise does not cover the carrier’s own immunogenicity. A patient who would otherwise become sensitized to red cells might instead become sensitized to a bovine protein. Whether that trade is favorable depends on the size and consequences of each. An honest description of the product would distinguish clearly between the absence of blood-group antigens, which is true, and the absence of immunogenicity, which has not been shown.
A research program that took this issue seriously could proceed in stages, each modest and each informative.
The first is to measure. In existing samples from earlier trials, and in veterinary populations, look for antibodies against the carrier and against the chemical modification. Test their ability to bind the product and whether they change its behavior in plasma.
The second is to model. In animal studies, expose animals twice and three times, at intervals of weeks, and compare the circulating half-life and the clinical response at each dose. A shortening of the half-life on the second dose would be a signal of antibody-mediated clearance.
The third is to observe in people. In a trial with planned repeat dosing, measure antibodies before each dose and at intervals afterward. Include clinical monitoring for allergic reactions and for signs of immune complex disease, such as rash, joint pain, and changes in kidney function.
The fourth is to build a registry. For any product used under expanded access, where the physician is treating a single patient in extremity, systematic data on repeated exposure is rarely collected. A registry that logs each dose, the interval and the outcome would accumulate knowledge from cases that would otherwise be lost.
The most challenging scenario is not the second dose but the tenth. Suppose an oxygen carrier were considered for use in a chronic condition, such as a severe anemia with no other option. Then a patient would receive many exposures over months or years. Immune tolerance to a foreign protein can develop under some circumstances, but sensitization is equally possible, and the balance depends on dose, route and frequency. The history of protein therapeutics, including enzymes and antibody drugs produced in animal cells, is full of cases where antibodies developed in a fraction of patients and reduced efficacy over time.
The lesson is that chronic use cannot be extrapolated from acute use. It requires its own evidence, with long follow-up and immune monitoring. A responsible development plan would treat acute and chronic indications as distinct programs and would not allow success in the first to imply readiness for the second.
A patient offered a carrier under expanded access should understand a few things. It is made from a foreign protein. The chance of a serious allergic reaction on first exposure is believed to be low, though not zero. The risk on later exposure is not well known. If a second dose is contemplated, the physician should be told about the first, and the patient should be observed carefully during and after the infusion, with resuscitation equipment at hand.
These points are unglamorous, and they belong in the consent conversation. The consent essays in this series argued that informed consent for an emergency product should include what is known and what is not known. Repeat dosing sits squarely in the second category, and saying so is part of respecting the patient.
One might ask why a question so basic remains open. The answer is a mixture of economics and incentives. A trial that follows patients for months after a single exposure is expensive, and a company with limited funds will choose the endpoints its regulator requires. Veterinary sales generate revenue without the obligation to collect immunological data. Academic laboratories rarely have access to the product. And the assumption that a bovine protein, chemically modified, will behave benignly has been reinforced by the absence of reports of trouble, an absence that may reflect the absence of anyone looking.
That last point is worth stressing. A silence in the record is not evidence of safety. It is evidence that no one was asked to write anything down. The remedy is deliberate: ask, record, and publish.
I would propose that the community concerned with oxygen carriers agree on a minimal dataset for every exposure, whether in a trial, in expanded access, or in a veterinary clinic that is willing to share. The dataset would record the product and lot, the dose, the date, the species and the patient’s identifier, any prior exposure, any reaction within a defined period, and where possible a stored serum sample. The cost is small, the value is large, and the data would answer the second-dose question within a few years of accumulating.
Such a shared record would also serve other purposes. It would support pharmacovigilance, would help detect manufacturing drift, and would give regulators a factual basis for decisions. The Foundation could host or sponsor the coordination, since the work is neutral, unglamorous and beneficial to every party.
Suppose a study does measure antibodies. What would the result mean? A positive test means that the patient’s serum contains something that binds the carrier. It does not by itself mean harm. Many people carry antibodies to many foreign proteins without symptoms, and binding in a laboratory dish is not the same as clearance in the body. The clinically relevant questions are graded. Do the antibodies neutralize the function of the carrier? Do they speed its removal? Do they form complexes that lodge in tissue? Do they accompany symptoms?
A careful analysis therefore pairs the antibody test with a functional one, such as the circulating half-life after a repeat dose, and with a clinical one, such as the incidence of allergic reactions. A result that shows antibodies with no functional or clinical consequence is reassuring. One that shows a shortened half-life without symptoms is a practical problem, since the product would work less well the second time. One that shows symptoms is a safety signal. Reporting all three together prevents both false alarm and false comfort.
Immune responses follow a pattern in time. Antibodies of the first kind, which appear within about a week or two of exposure, are followed by a more mature response over the next weeks. If a second dose arrives during the transition, when antibody levels are rising, immune complexes are most likely to form. If it arrives long after, when antibody levels have fallen but memory cells remain, the response can be swift and strong. The riskiest interval may therefore not be the shortest or the longest but somewhere between, and it is unlikely to be the same for every patient.
For a clinician, the practical consequence is that the date of the first exposure is a clinical fact that belongs in the record, easily accessible when a second is considered. Emergency care rarely provides for this. A patient treated in one hospital may arrive months later at another, with no link between the records. The registry proposed above would help, as would something as simple as a wallet card given to every recipient stating the product, the date and the lot.
The question of immunogenicity feeds back into the choice of raw material. Bovine hemoglobin is one option. Others include hemoglobin from pigs, from horses, from human donors and from engineered microbes producing human sequences. In principle, a molecule with the human sequence should be less immunogenic than one from a cow. In practice, recombinant proteins produced in microbes can differ from natural ones in modification and folding, and chemical stabilization changes the surface regardless of the source.
One reasonable view is that the immune question is a reason to keep several sources under study rather than to commit early to one. If repeat dosing turns out to be a problem for bovine products, a human-sequence alternative may be the only route for chronic use. If it turns out not to be, the bovine product retains its advantages of scale and experience. The information will only arrive if someone looks, and until it does, diversification is prudent.
A regulator reviewing an oxygen carrier for approval could ask for immunogenicity data proportional to the intended use. For a single-dose emergency indication, the demand for repeat-dose data might reasonably be modest, with a commitment to post-approval surveillance. For an indication that implies repeated exposure, the standard should be higher, and evidence from repeat-dose studies should be required before approval. Splitting the requirement by indication would avoid both extremes: burying an emergency product in paperwork and approving a chronic one on evidence collected for a different purpose.
Post-approval surveillance deserves a special mention. Products approved under conditions of unmet need, with a commitment to further study, have a mixed record on fulfilling the commitment. A public schedule of the promised studies, with the results posted when they are due, would make the arrangement more than a formality.
The first dose of a carrier is a bet on the moment. The second is a bet on memory, the body’s and ours. The moment has been studied for a century, and memory has hardly been studied at all. If the field is to mature from emergency improvisation to reliable medicine, it must learn what its products do to the immune system that meets them twice. The experiments are not exotic, and the money required is small compared with a pivotal trial. What is needed is the decision to look.
The immunogenicity of protein therapeutics and of PEGylated drugs is reviewed in the general pharmaceutical literature, including work on anti-polyethylene glycol antibodies. Immunological monitoring in hemoglobin-based oxygen carrier trials is described in individual trial reports and in reviews of the class published in transfusion and artificial cells journals. Alloimmunization in sickle cell disease is reviewed in hematology literature. The veterinary product’s labeling and the company’s statements on its use across species are available from the manufacturer. Claims above about the absence of reported problems reflect my limited reading and should be checked against adverse event databases.