Only a few years ago, back when I was in medical school, I was taught that the goal of oxygen therapy should be to push saturation above 95%, and that anyone who comes in to hospital with a saturation below that should receive oxygen therapy (with the exception of people with chronic obstructive pulmonary disease, whose bodies have adjusted to lower oxygen levels). For those who aren't used to the terminology, the definition of oxygen saturation is the proportion of haemoglobin molecules in the arteries that are "saturated" with oxygen. A normal level for a healthy person is usually 97% or higher.

In the last few years, however, there's been a bit of a shift in thinking. It started with the realization that people with heart attacks who were treated oxygen didn't do any better than those who weren't. This is a good example of a medical reversal - it seemed logical to give oxygen to people with heart attacks, because a heart attack is a blockage in one of the arteries that supply the heart, which means that the heart muscle isn't getting enough oxygen. By increasing the oxygen level in the blood stream, even if only a small amount of that oxygen is able to get past the blockage, it should do some good. At least that was the thinking.

Unfortunately, logic often turns out be wrong, at least in medicine. As it turns out, oxygen isn't the utterly benign substance it's often made out to be. First, oxygen causes blood vessels to constrict, so by pushing up the oxygen saturation you could actually be worsening the blockage. Second, oxygen has an unfortunate tendency to form reactive oxygen species (ROS), which can wreak all kinds of havoc in our cells. We've evolved mechanisms to deal with these, but when a part of the body is depleted of oxygen for a while, many of our defences to deal with ROS are diminished. If you then push up the oxygen arriving in an area enormously, say an area of heart muscle that's been deprived of oxygen for a while but that now once again has good flow thanks to an intervention to remove the blockage, you might actually increase the damage to that area.

Long story short, it turned out that oxygen wasn't good to give to people with heart attacks. It might even be bad. Thankfully, that practice has now stopped in most places. The realization that oxygen might actually be harmful in certain situations has led to a change in thinking about oxygen therapy. No longer is it viewed uncritically as a universal panacea that can be doled out generously to everyone who comes in through the doors of the hospital. At least that's the case in theory. In practice, lots of patients still get oxygen who don't need it.

When hospital staff see a patient who is out of breath, they will often will shove a mask on their face and crank up the oxygen, regardless of what the patient's saturation level is. On a superficial level, I guess this makes some sense. If someone's breathless, they must need oxygen, right?

Wrong. There are many things that can cause a sensation of breathlessness, and a lack of oxygen is just one of them. Our bodies are actually not very good at detecting changes to the oxygen level in the blood stream. If a patient has a saturation of 90% and is out of breath, then the breathlessness is not caused by the somewhat low oxygen. I'll repeat that sentence, just to be sure you didn't miss it. If a patient has a saturation of 90% and is out of breath, then the breathlessness is not caused by the somewhat low oxygen. Our bodies are in fact amazingly poor at noticing the oxygen level in the blood stream, and don't really start to pay attention until the oxygen saturation drops well below 80%.

For the most part, our breathing rate is determined by the level of carbon dioxide in the blood stream (or to be more technically correct, the pH), not by the level of oxygen. It is much more central to our continued life on this planet to keep carbon dioxide within strict limits than it is to make sure that oxygen is always kept at a very high level. That's why people who engage in free diving will often intentionally hyperventilate before going under the water. The purpose isn't to increase the oxygen content in the bloodstream (which is anyway already at 100% or thereabouts). It's to decrease the carbon dioxide content, which will allow them to hold their breath longer, since it will take longer for the carbon dioxide levels to reach the point where their bodies force them to take a breath.

Ok, that was a rather long preamble, but I think I've set the stage sufficiently now. A study was published recently in the New England Journal of Medicine that sought to answer the question of what a reasonable level of oxygen supplementation is in patients who are having trouble oxygenating themselves.

This was a randomized trial carried out at 35 intensive care units (ICU's) in seven different European countries. In order to be included in the trial, participants had to be over the age of 18 and have respiratory failure for which they were receiving at least ten liters of oxygen. Unusual causes of respiratory failure which require special treatment, such as carbon monoxide poisoning and cyanide poisoning, were excluded from the study.

Participants were randomized to an oxygen saturation target of either 90% or 96% (technically they were randomized to a PaO2 of 60 mmHg or 90 mmHg, but it's roughly the same thing). The study wasn't blinded, since treating a patient to a specific target requires knowing what their saturation is.

2,928 patients were included in the trial, a nice big number that should show a difference in mortality if there is one, especially considering that the mortality rate in patients treated in ICU's is high.

The median actual saturation level in the low oxygen group ended up being 93%, while the median in the high oxygen group ended up being 96%. In other words, not a huge difference. Maybe the nurses felt squeamish about letting the oxygen sit at 90%, as my experience tells me they are wont to be. Obviously, this will make the study a little less useful than it could have been.

Let's get to the results.

At the 90-day point after recruitment in to the study, 42,9% of patients in the low oxygen group had died, as compared with 42,4% in the high oxygen group. This absolute difference of 0,5% was nowhere close to being statistically significant (p-value 0,64).

When it comes to serious adverse events, 36,1% of participants in the low oxygen group suffered a serious adverse event, as compared with 38,1% in the high oxygen group. Again the difference was nowhere close to being statistically significant (p-value 0,24).

In other words, there was no meaningful difference between the groups in terms of either mortality or serious adverse events. This was a big, high quality study, and it is telling us that we don't need to push patients' oxygen saturation up over 95%. If the saturation happens to be at 93% then that's just fine, no extra oxygen required.

If every nurse, nursing assistant, and doctor became aware of this, hospitals could probably cut down on oxygen use enormously, and since the perceived need for oxygen therapy is one of the main reasons for keeping patients in hospital rather than sending them home, it would frequently also allow for earlier discharge. Considering that hospitals in many western countries are constantly operating at 100% capacity (which was the case long before the advent of covid-19), this could free up a lot of hospital beds.