Timing of tracheostomy in the ICU

The debate of early vs. late tracheostomy is far from settled. Two recent (2015) meta-analysis reviews have come to slightly different conclusions about the effect of early tracheostomy on mortality. It seems to have either little effect or no effect on mortality, no effect on the length of ICu stay, and no effect on the duration of mechanical ventilation. There is marked heterogeneity in the trials, suggesting that there is probably a population who benefit from early tracheostomy (eg. burns, trauma, stroke) and a population who can wait for a chance of extubation.

Question 13 from the first paper of 2013 and the near-identical Question 13 from the second paper of 2017 asked the candidates to "critically evaluate the timing of elective tracheostomy in adult Intensive Care patients". This SAQ was quite unique among the CICM papers, because the model answer makes reference to TracMan, a trial which had not yet been published at the time of the exam. This question may never appear again because this issue is far from settled, and because the evidence is conflicting. It is unlikely that the examiners will ever ask about this again because it will be difficult to mark. If this question were repeated in future papers, modern data expected of the candidates would probably come from the 2015 meta-analysis article by  Andriolo et al and Szakmany et al, which arrived at essentially the same conclusions (in three words, early vs late tracheostomy makes no difference to any parameters of interest when you homogenise the ICU population),

The definition of early and late tracheostomy

The college answer to Question 13 reports that tracheostomy "may be considered as “early” at <10 days or “late” >10 days", which is a definition probably adopted from the 2012 Cochrane review by Gomes et al. Unfortunately, there is no "gold standard" beyond this one: essentially, trial authors have been free to use any arbitrary definitions. In the modern era one might expect that all future studies of this topic will expect to be compared to TracMan, and will therefore use the same 10 day threshold.

Rationale for the debate about tracheostomy timing

One might ask, why do we even bother? In the absence of strong evidence,  intensivists will do whatever the hell they want, guided by concerns of patient comfort, futility versus beneficence, local hospital policy or even financial incentive (see the bit about "up-coding" in the PulmCCM article). And what's wrong with that? Surely it is better to the individualise the decisionmaking, rather than to broadly extend tracheostomy to all patients on the basis of generic trial evidence?

In fact, there are multiple reasons to investigate the timing of tracheostomy. For the purpose of exam revision, these may be summarised as points:

• Tracheostomy is not a benign procedure
• There is a proportion of patients who will require prolonged ventilation, and who will nevertheless escape a tracheostomy by being successfully extubated.
• Being intubated is also not a benign procedure.
• Both tracheostomy and prolonged intubation have risks which include death.
• It would therefore be advantageous to compare those risks, to determine which strategy has the lowest risk of death or permanent disability
• On a lower tier of interest lie surrogate outcome measures such as ICU stay, duration of mechanical ventilation, incidence of VAP and so on.
• There may also be groups which derive some sort of unique advantage from early tracheostomy, for instance because they can be expected to spend a long time on the ventilator: classically, one example is the group of patients with profound weakness (in whom prolonged intubation is unlikely to culminate in extubation).

Arguments for early tracheostomy

• Avoidance of the perils of prolonged intubation: having an endotracheal tube for a prolonged period exposes patients to a host of complications from which the tracheostomy should theoretically be free. Realistically, the main improvement is in terms of patient comfort, which permits the sedation to be weaned - with follow-on effects for haemodynamics and ventilator support.
• Maximise the advantages of tracheostomy - there are numerous positive effects from tracheostomy which the patient could benefit from, so why deny them these benefits for an artificially prolonged period? May as well get all the benefits early, when it might still make a difference to the course of their illness.
• Decrease the length ICU stay:  with one of the supposed benefits of tracheostomy being a more rapid ventilator wean, the earlier you do the tracheostomy the sooner the patient can be weaned and discharged from the ICU. This has implications for resource utilisation in ICU.

Arguments for delaying tracheostomy

• Some patients may end up being extubated after a prolonged intubation:  we just don't know which ones. Presently, there is no satisfactory method to identify them, but broadly speaking patients with reversible (and severe) cardiac and respiratory pathology can be expected to eventually get better, and they may be better served by waiting for a "late-term" extubation.
• A long period of intubation allows these patients to be more easily identified.  The more we spend observing the progress of somebody's illness, the more confidently we can discuss their prognosis. Delaying a tracheostomy therefore improves the quality of decisionmaking. The patients who receive a tracheostomy because they are not progressing at day 10-14 of ventilation can be described as more "deserving" - the indications for tracheostomy are clearer the longer you wait.
• Risks of prolonged intubation are overstated: it is not clear that prolonged intubation is as harmful as it is made out to be, as much of the data regarding protracted mechanical ventilation comes from a bygone era when high-pressure tube cuffs were used. In this day of low-pressure high-volume cuffs, subglottic suction ports, selective digestive tract decontamination, aggressive physiotherapy and posture therapies, as well as rational sedation - these days the risks of staying ventilated for a long time are relatively minor.
• Tracheostomy exposes patients to certain procedural risks, and the practice of routine early tracheostomy exposes more patients to these risks (many of whom would have been extubated after two or three weeks).
• Futile care may be perpetuated by tracheostomy. A tracheostomy may commit a severely disabled or persistently unconscious patient to the ongoing unpleasantness of their survival. The fact that a patient might end up in this category may not be clear in the initial stages of their admission. If the prognosis becomes obvious only after the tracheostomy, the tracheostomy then becomes a part of their unnecessary aggressive management.
• Length of ICU stay is largely a financial rather than medical consideration, and some might argue that it is inappropriate to let such matters influence decisions regarding patient care. If early tracheostomy decreases the length of ICU stay at the cost of numerous inappropriate tracheostomies, then this is not how we should practice medicine.

A summary of the evidence regarding tracheostomy timing

Pre-TracMan data

A Cochrane review from 2012 (Gomes et al) is a good representation of the data which existed at this stage, and which would have been familiar to the exam candidates writing an answer to Question 13 from the first paper of 2013. In fact it might have been the stimulus for the examiner who wrote that specific SAQ. In brief, the systematic review had found four studies worth analysing. In total, 673 patients from four trials were scraped together. The findings of the analysis were as follows:

• Improved mortality with early tracheostomy (which did not reach statistical significance)
• More time off the ventilator (and more success weaning; NNT = 11)
• Slightly shorter ICU stay (by about 1 day)
• No change in duration of hospital stay
• A decreased incidence of pneumonia (12.3% vs 21.9%) with early tracheostomy
• A greater incidence of tracheal stenosis at 10 weeks

However, the reviewers excluded a study where 8 days (not 10) was used as the definition threshold for early vs. late, as well as excluding six studies which specifically compared early tracheostomy with prolonged intubation. There was such heterogeneity of findings among the four remaining trials that Gomes et al were unable to pool the data. "Evidence is of low quality" they complained; "potential differences between early and late tracheostomy need to be better investigated by means of randomized controlled trials".

TracMan trial (Young et al, 2013)

In response to the desperate need for high quality trials, TracMan enrolled 909 patients from 72 British ICUs. The above-linked PulmCCM article dissected it in detail, and there are some useful comments by the audience at the end. In brief, the study had the following features:

• Patients were randomised to undergo tracheostomy on day 10, or immediately (as they were randomised before day 5)
• In the late group, 55% of the patients did not need to undergo tracheostomy- they ended up being extubated before day 10.
• There was no mortality difference, but the early group had more ventilator-free days (1.7 days less), supporting the idea that tracheostomy improves weaning.
• 7% of the tracheostomised patients had significant bleeding

However:

• The study was underpowered: 80% power to find an 8% mortality difference. This was described by the authors as "recruitment fatigue", which sounds as if they got sick of the process - but in fact they also ran out of money, so the trial had to end prematurely.
• Each center only contributed 2-3 patients every year.
• Of the early group, only 92% underwent tracheostomy.
• Neurological and neurosurgical patients were excluded (and that group accounts for most of the tracheostomies performed annually).

In the wake of TracMan

Later, the abovementioned Cochrane review has been revisited by Andriolo et al (2015). In the three years since the last review, more RCTs were performed and eight studies were included (n=1977, of which 909 were from TracMan). The analysis results were as follows:

• There was a statistically significant mortality benefit associated with early tracheostomy (47% vs 53%); the NNT was 11. This mortality benefit was seen at the longest reported follow-up, rather than at 30 days.
• The early group had a higher chance of being discharged from the ICU on day 28.
• There was no statistically significant effect on the duration of mechanical ventilation.
• The early group had decreased duration of sedation.
• Data regarding risk of pneumonia could not be subjected to meta-analysis due to heterogeneity.

Another (non-Cochrane) meta-analysis was performed by Szakmany et al (2015), with n= 2406. The only difference in their findings was an absence of mortality difference; in most other respect this analysis replicated the findings by Andriolo et al.

SETPOINT2

Bösel et al (2022) came back to the themes of TracMan to see whether there is any benefit of early treacheostomy in a population whose need for tracheostomy is  already predictable from an early stage - severe stroke patients. Their findings and conclusions are scrutinised at thebottomline.org.uk with characteristic rigor. In brief summary:

• They randomised 199 to early and 194 to late, and 130 ended up actually getting the late tracheostomy
• Mortality difference was not statistically significant (43.5% in early group vs 47.1%)
• The patients were on average more sick then the general stroke population (50% of their ischaemic strokes went for a decompressive craniectomy, for example)
• A fairly large proportion in each group did not get the treatment to which they were randomised

One take home message here is the implications of the last point. About 33% of the late tracheostomy patients never ended up getting their late tracheostomy. This means that for them an early tracheostomy would have been an unnecessary procedure,  a potentially avoidable source of complications and morbidity.

Timing of tracheostomy in specialised groups

For the modern era, the review article on this topic by Longworth et al (2016) is an excellent example of the evolution of this discussion. We no longer ask "when", but rather "who" when it comes to tracheostomy timing. Cleary some groups derive greater advantage than others. This is an issue which may be neglected by the large all-comer trials. In summary:

• Neurological and neurosurgical patients may benefit from early tracheostomy, as it reduces the incidence of VAP and reduces their length of ICU stay. However, there is no effect on mortality or neurological outcome. If one were arguing against a tracheostomy for a neurocritical ICU patient with likely poor neurological prognosis, one might bring up this point. The SETscore (Schönenberger et al, 2016) is a validated tool to help predict long ICU length of stay, where a score of 8 or above seems to be the optimal cut-off (the full scoring table can be seen here). 
• Cardiothoracic ICU patients do not seem to develop deep-seated mediastinal infections or sternal wound infections as a result of their tracheostomies, and therefore in this group there is probably no reason to delay tracheostomy (i.e. treat this group the same as any other patient group).
• Burns patients with > 60% BSA burns or airway burns  require either multiple trips to theatre or prolonged airway rest, or both. That said, there will be no influence on mortality or length of ICU stay (these will be determined by numerous other factors).
• Trauma patients with severe injuries will require prolonged ventilation for their multiple rib fractures. One may be able to identify these unextubatable patients by their admission CT report. In these people, earlier tracheostomy may be of benefit.