Oxygen targets in critical care: The HOT-ICU trial. St Emlyn's • St Emlyn's

Oxygen targets in critically ill/injured patients has been something we’ve talked about on the blog before. In general the evidence to date has suggested that the routine prescription of oxygen to patients is unhelpful and may well be harmful. Back in 2018, our own Dan Horner published in the BMJ on this, together with updated guidelines that advocated the prevention of ‘superoxia’ in patients. However, there was still much uncertainty as to what an appropriate target for oxygen levels are in the critically ill (and in ward patients too). This week we have new evidence presented at the Critical Care Reviews conference in Belfast, and published in the NEJM. The HOT-ICU trial is a target based randomised controlled trial of ICU patients receiving oxygen.

The Critical Care Reviews conference was held online this year. It’s arguably the best conference in the world for anyone interested in evidence based critical care and I’d strongly recommend finding time to attend in person, or virtually. Personally I am really hoping that it will be face to face 2022 when it will be on my bucket list of places to go. The video presentation of the trial is available on the CCR website. If you go to day 1 and start at 2:45:00 you should be able to see the presentation followed by an editorial by Paul Young.

The abstract is below, but as always we strongly recommend you read the full paper for yourself.

What kind of trial is this?

It’s a randomised controlled trial which is exactly what we want to see when testing an intervention.

Tell me about the patients.

HOT-ICU enrolled adult patients within 12 hours of admission to the ICU with hypoxic respiratory failure. This was defined as requiring more than 10L/min of O2 by face mask or requiring an FiO2 of more than 50% .

Patients were recruited across 35 European ICUs between 2017 and 2020.

What about the interventions?

The study randomised patients to two oxygen targets. The high-O2 group aimed for a PaO2 of 90mmHg, wherease the low-O2 group aimed for 60mmHg, Clinicians tried to keep these targets +/- 7.5mmHg unless the patient was on maximal or no O2 supplementation. For those of us in the UK it might be easier to think in kPa and so that’s 8kPa for the low-O2 group and 12kPa for the high-O2 group.

What about the outcomes?

The primary outcome was all cause mortality at 90-days from randomisation. They also included a whole range of secondary outcomes that were typical of those we see in many ICU trials (e.g. renal support).

What are the main results?

HOT-ICU managed to recruit 2928 patients of whom 1462 were allocated to the low-O2 group. Follow up was excellent with data available for over 98% of patients).

One question that’s really important in trials that aim for targets is whether or not the patients actually do get to the target level. This is especially important when the interventions are responding to what might be a rapidly changing clinical picture (as in this trial). It was therefore reassuring to see that there was clear separation between oxygen target groups when the patients had their PaO2s measured. However, the target range for low-O2 strategies did not consistently reach 60mmHg and sat at about 10mmHg higher. The principal investigator, Dr Rasmussen makes the point that it’s easier and arguably safer to keep a patient at 90mmHg than 60mmHg, partly because lower levels might make us nervous but also because 60mmHg is closer to the steep part of the oxyhaemoglobin dissociation curve.

HOT-ICU results by Dr Rasmussen 🇩🇰 presented at #eCCR21 – despite achieving clear separation in PaO2 and SaO2, no difference between the two groups
Eagerly await pediatric results from OXY-PICU trial lead by @pus27 #PedsICU pic.twitter.com/lx0EI699ly
— Jan Hau (@leejanhau) January 20, 2021

For the primary outcome (death at 90 days) there was no difference between the two groups. 42.9% died in the low-O2 group and 42.4% in the high-O2 group. There were no significant differences in adverse events between the two groups.

Where does this evidence fit with other studies?

The discussion is remarkably balanced and the accompanying editorial by @ICUdogma is similarly objective. There are a number of trials published in the area as shown on the tweet below.

@DogICUma 🇳🇿 #eCCR21: Summarising current data on oxygen therapy in adult #ICU in his editorial for HOT-ICU trial hot-off-the-press @NEJM #PedsICU pic.twitter.com/ASV4FWrE9O
— Jan Hau (@leejanhau) January 20, 2021

These trials suggest that we really don’t know about oxygen targets as yet. There are a number of well know trials such as LOCO2, ICU-ROX and OXYGEN-ICU which produced contradictory evidence, perhaps due to different case mixes amongst the patients. HOT-ICU certainly adds to the mix but again contains a quite different patient population from previous studies and so we must be cautious in how this work is used and whom it is applied to.

LOCO2 said low oxy targets were killing people. https://t.co/1Ygav35beL

Oxygen-ICU said high oxygen as killing people. https://t.co/zLYEacCT0f

HOT-ICU (biggest and best to date) says the truth lies smack between them. https://t.co/jYnkvDG4bf

MEGA ROX still to come#eCCR21
— Andy Neill (@AndyNeill) January 20, 2021

If you have time it’s worth looking through the secondary outcomes in the supplementary materials. Here there are some interesting subgroup findings such as an increased survival in the high-02 group post cardiac arrest, but these findings should be considered exploratory/hypothesis generating in nature. However, they are important. Respiratory failure is quite a broad group of patients and it may well be that there are important subgroups of patients who may be affected by oxygen targets.

Dan Horner’s take on the trial is that it seems as though lower oxygen target saturations in a closely monitored environment appear to be safe (and this is in keeping with practice – we are aiming for 90% in Covid for example, in keeping with national guidance), critical care is a heterogenous patient group. We have reasonable evidence now to suggest this is safe in ARDS, STEMI, COPD, post cardiac arrest and pneumonitis. We are still very worried about the idea of using these targets in brain injury (where PaO2 is a vital driver of secondary brain injury), subarachnoid haemorrhage and other neuroscience conditions. Hopefully MEGAROX will help with this, but it will depend on the inclusion/exclusion criteria.

In a trial of this size we won’t be able to show this, but I am hopeful that the MEGA-ROX trial will be able to pull out this kind of subgroup analysis. We should also highlight ICNARC who are launching their own MEGAROX type trial, called UK ROX.

As Paul Young states in the CCR editorial, a trial of this size is unable to detect heterogenous effects between subgroups and they may be important. Previous trials have ranged from 2.9% reduction to 4.2% increase in mortality when low vs. high targets are studied. Could that be due to case mix? Quite possibly.

Makes me glad we have some protocolisation of the high O2 group in #MegaROX. Usual care may not be as liberal as it once was…
— Paul Young (@DogICUma) January 21, 2021

Final thoughts

This is yet another study that demonstrates that high levels of oxygen are not the automatic target for critically unwell patients. The evidence for this seems to be increasingly clear. What is still uncertain is whether this applies to all patients or if there are subgroups who benefit from high or low oxygen targets. We are also not entirely clear whether an intermediate target between the levels in this paper may hold an advantage.

Hopefully the MEGA-ROX and UK ROX trials will answer some of these questions, but in the interim this paper suggests that we can be more conservative with oxygen targets than we have been in the past.

Simon Carley @EMManchester & Dan Horner @RCEMProf