Virchester (the home of St Emlyn’s) is a large conurbation that suffers from many of the issues that affect large urban populations in the western world. One part of that is the large consumption of drugs and alcohol; intentionally, unintentionally, and accidently. Such patients often end up in the emergency department with a low GCS and consequent concerns about the safety of their airway and ventilation. In my early training there was a ‘rule’ that patients with a GCS of less than 8 require intubation, and whilst that has largely been debunked it is still used in practice by those who prefer protocols to thinking. A low GCS may be associated with poor airway protection, but it does not account for the patient’s pathology and more importantly their likely clinical course. There is an excellent explanation of this in a paper from EMA in 2023 by Pellatt et al, who rightly point out that our tox patients are (in general) going to get better reasonably quickly if they can be supported during the low GCS phase, but patients with traumatic brain injury will not. The pathology in TBI thus pushes us towards early definitive airway management (and for lots of other reasons too), whereas the case for definitive airway intervention in tox cases is less clear cut.
Here at St Emlyn’s it is not uncommon for us to RSI patients in the ED as a result of low GCS in the context of poisoning. These patients often then end up with a night on the ICU and a discharge from the ICU in the morning when they are successfully extubated. That’s a great service to patients, and it delivers a high level of care, but it is not without risk/harm and it’s not cheap. Cheekily a colleague did comment that this makes our ICU one of the most expensive and best staffed B&Bs in England, although he declined to comment on the quality of the breakfast.
This month we are looking at a paper from JAMA at the end of 2023 that asked the question regarding intubation in low GCS tox patients. The abstract is below, but as always please read the full paper and come to your own conclusions. There is also an excellent editorial from St Emlyn’s Zaf Qasim here.
What kind of paper is this?
This is a randomised controlled trial, which is great to see. A lot of the toxicological literature is observational and it would be tempting to simply conduct an observational, likely retrospective, study of this question, but those designs are especially prone to bias here. The decision to intubate is always going to be quite subjective and as such observational retrospective studies are always biased and it’s therefore great to see an RCT design here.
Patients were randomised in 1:1 blocks, stratified by the hospitals involved using a sealed envelope system. The nature of the intervention means that the clinical aspects of this study cannot be masked (blinded) to participants or clinicians.
Tell me about the patients
Patients were recruited across 20 emergency departments and 1 ICU in France between May 21 and April 23. They managed to recruit 237 patients into the study, so not the largest study in the world [Ed – or even in France!], but a decent number (though whether there are enough we will come to later). In 16 units patients could be recruited in the prehospital phase of treatment (remembering that the French system often projects physicians into the prehospital setting).
Patients were eligible if they were adults (age≥18 years) with a clinical suspicion of acute poisoning and a GCS of less than 9 as determined by the treating clinician. It’s worth noting the exclusions though. As well as the usual exclusions for pregnancy, prisoners etc…, patients with an easily reversible cause (e.g. opiates) or those with what I would interpret as high risk cardiovascular poisonings (e.g. Ca channel blockers, ACE inhibitors, B-blockers) were also excluded.
What did they do?
It’s important to note that this was not an intubate vs. don’t intubate trial. The standard arm of the trial was that the eligible patients with a low GCS had an RSI a the discretion of the treating clinician and usual management. Those in the intervention arm only got intubated if they then met the following criteria.
- Resp distress ( SaO2 <90%) on nasal O2
- SBP <90mmHg despite 1L crystalloid
The trial period was defined as 4 hours from randomisation.
Tell me about the outcomes?
The principal outcome was a composite hierarchical outcome at 28 days of death, ICU length of stay (LOS) and hospital LOS. The authors also recorded a large number of secondary outcomes linked to ICU stay, ventilation, complications etc.
The statistical analysis uses the Finkelstein-Schoenfeld method which is something I had to read up on! It’s actually pretty cool for composite endpoints as we have here, especially when the endpoints are not really of equal weighting. So death is worse than a longer ICU stay, which is worse than a longer hospital stay. I think we would all concur. In the Finkelstein-Schoenfeld method patients in each group are paired. Every possible combination is created (so in effect one patient in one group has multiple pairs as they are paired with every patient in the other group. This means you end up with a lot of pairs! Within each of those pairs, whichever patient gets the most important event first wins (or loses), if neither have that event they then compare the second most important event and so on.
What did they find?
225 patients were recruited but after exclusions for data permissions and legal protection measures 213 made it through to analysis: 109 in the intervention group and 116 in the control group.
- 18/112 (16.1%) intubated in intervention group vs. 63/109 (58.9%) in control group
- Patients in both groups had taken a range of drugs, but alcohol was predominant in both groups (66.6% of patients)
- The primary composite hierarchical outcome was better in the intervention group (watch and wait). This is expressed as a win ratio for the composite end group. So overall (combined) there were 64.6% winners for intervention (watch and wait) vs. 34.8% for control (and a very small number of ties.
- In terms of the individual components of the composite hierarchical outcome we have the following data.
- No data for death as no outcomes.
- For ICU stay 55.5% of the intervention group spent less time on ICU, 23.7% spent more time and 20.8% were the same
- For hospital stay 43.8% of intervention group spent less time on ICU, 53.4 % spent more time and 2.8% were the same (remember that for hospital LOS, this analysis does not include those that have already won or lost for ICU stay. That’s the hierarchical bit of the method)
- 7/112 (5.4%) adverse events in intervention group vs. 16/109 (15%). although 14 of these were difficult intubation issues despite a very high first pass success rate. Difficult intubation was defined as an IDS of >5 (which is technically moderate to difficult intubation)
So looking at the data here it looks as though a watch and wait approach seems reasonable, with a shorter ICU stay, although if you don’t make it to ICU or there was no difference then you will stay in hospital for longer.
In terms of adverse events then most of the difference is due to differences in intubation difficulties.
So should we stop intubating these patients?
The data is reassuring that a more conservative approach is possible, but as always there are caveats.
The study is quite small when we are considering rare events. Although 237 patients were recruited, only 119 were recruited to the restricted intubation group. Although there were no deaths in this group we need to be mindful that there are confidence intervals around that estimate of a 0% event rate. Confidence intervals for a zero event rate are a little more complex to calculate, but my estimate is that the 95% confidence interval is 0-2.5%, which when considering the potential adverse outcomes are severe, may not be high enough to change my practice entirely. It is also interesting to note that in the sample size calculation the authors estimated a 3% mortality, which did not occur in either group. That figure was based on past data and so it makes me wonder whether this cohort is reflective of usual practice.
Secondly, these patients will still require very close medical and nursing supervision irrespective of whether they are intubated, and whilst the duration of that supervision may change as a result of not intubating the patient, it will still need to be done. This is most likely to be done in the resus room or on the ICU, and in the UK I strongly suspect that the non-intubated and observed patient will not get a bed on the ICU. They will remain a significant task for the EM team to manage.
Thirdly, the Finkelstein-Schoenfeld analysis has no events to work with for death, and so when we look at the data we are really looking at ICU and hospital LOS as the measures analysed for the primary outcome.
Fourthly, patients were intubated in both groups and that need will remain, as will clinical equipoise. In the control group the intubation rate was 58% and so we know that many of these patients are managed without the need for intubation at the moment.
It’s also worth noting the preponderance of alcohol as a causative agent in these patients. This is important as I think many of us manage these patients in a more observational way (assuming the history is good), as we know that the clinical course is reasonably predictable. Other patients in the study were less common, but my approach to the tricyclic patient with a low GCS will be very different to alcohol alone.
In practice I deal with intubation/RSI decisions on deeply intoxicated patients on a very frequent basis. It’s complex, it requires judgement and it requires a bit of time. Time is really important, and one of the key points I teach all the time is to consider the trajectory of the intoxicated patient. It is often possible to get information from bystanders, or just from some observed time in the resus room to determine if the patient is getting better, getting worse, or staying the same. That information, combined with knowledge of the likely causative agent are the major determinants of the decision to proceed to RSI/intubation. In other words, GCS is an important finding, but the decision is far more complex than that.
I think this paper supports my thoughts and practice above. If your patient is stable, safe, observed and supported then there is almost always time to pause and make a sensible decision. That decision may well be to watch and wait, it may be to intervene, but strict rules based on GCS are unlikely to be in anyone’s best interest.
Another great paper from Yonathan Freud and colleagues. If you don’t follow him already, then I’d suggest you start now.
- Freund Y, Viglino D, Cachanado M, et al. Effect of Noninvasive Airway Management of Comatose Patients With Acute Poisoning: A Randomized Clinical Trial. JAMA. 2023;330(23):2267–2274. doi:10.1001/jama.2023.24391
- Qasim Z, Perrone J, Delgado MK. The Value of Not Intubating Comatose Patients With Acute Poisoning. JAMA. 2023;330(23):2253–2254. doi:10.1001/jama.2023.22462
- Cochrane handbook: Confidence intervals when no events are observed. https://handbook-5-1.cochrane.org/chapter_16/16_9_4_confidence_intervals_when_no_events_are_observed.htm
- Frederic Adnet, Stephen W. Borron, Stephane X. Racine, Jean-Luc Clemessy, Jean-Luc Fournier, Patrick Plaisance, Claude Lapandry; The Intubation Difficulty Scale (IDS) : Proposal and Evaluation of a New Score Characterizing the Complexity of Endotracheal Intubation. Anesthesiology 1997; 87:1290–1297 doi: https://doi.org/10.1097/00000542-199712000-00005
- Björn Redfors, John Gregson, Aaron Crowley, Thomas McAndrew, Ori Ben-Yehuda, Gregg W Stone, Stuart J Pocock, The win ratio approach for composite endpoints: practical guidance based on previous experience, European Heart Journal, Volume 41, Issue 46, 7 December 2020, Pages 4391–4399, https://doi.org/10.1093/eurheartj/ehaa665
- Pellatt et al. Intubation for patients with overdose: Time to move on from the Glasgow Coma Scale. EMA. 2023 https://onlinelibrary.wiley.com/doi/10.1111/1742-6723.14254