Time for consensus on rapid sequence intubation?

Association vs. significance

A striking feature of the survey by Zdravkovic et al. is the vast number of included participants. This is, to our knowledge, one of the largest international clinical surveys of anaesthetists. To give this remarkable achievement context, which was accomplished without external funding, the fifth and sixth national audit project baseline surveys included 7125 and 11,104 respondents, respectively 6, 7. The authors targeted respondents using LinkedIn (http://www.linkedin.com), which is novel and innovative. Just as social media is emerging as a useful way in which disseminate medical publications 8 and educational material 9, it seems it may also have a future role to play for those wishing to conduct large‐scale clinical surveys.

The authors looked for associations between the preferences of respondents and a range of factors, such as the national income bracket of respondents’ countries. There are several ways in which to report these associations, and there is no single correct choice of statistical test. Taking other large surveys as an example, the sixth national audit project reported data as number (proportion) throughout, with no data on associations between variables. This seems logical, as respondents were not reporting clinical ‘events’, which were instead a mixture of experiences, perceptions and preferences. The recent survey on suicide among anaesthetists 10 included data on 1916 cases from 1397 respondents. Again, the authors of this survey are clear that the actual number of cases remains unknown, due to the chance of multiple reporting. We argue the data reported by Zdravkovic et al. do not allow us to make inferences about what happens to patients in certain countries, cities, departments or for certain presentations, such as intestinal obstruction. They simply tell us about the preferences of respondents for work ‘as imagined’ for certain scenarios, and not work ‘as done’ for individual patients 11. Put another way, preferences expressed by survey respondents are not the same as clinical interventions or outcomes.

The authors present their results using ORs, which, put simply, indicates the degree of association between two variables 12. Odds ratios do not imply causation, and, in principle, their use is an accepted way in which to describe the results of survey‐based research 13–15. What matters most, however, is the language used before and after the numbers, which must be precise to avoid any misinterpretation. As suggested above, a preference is not the same as a clinician undertaking a procedure in a patient. More importantly, interactions between categories of respondents should, ideally, be acknowledged and accounted for. Each respondent is a member of six interacting categories, including: those trained in gastric ultrasound; those working in a setting with local or national RSI guidelines; consultants vs. trainees; the country of practice; and the city of practice. For each association between a practice preference and a category, the OR should be adjusted for the five other categories. As this has not been reported, the degree to which we can draw firm conclusions about these associations is limited.

Recently, the New England Journal of Medicine issued new guidelines for statistical reporting 16, and other journals seem to be following suit. There is now a requirement for trials and observational studies to prioritise the reporting of estimates of effect or association, along with a 95%CI. Moreover, abandoning p values together with the accompanying black and white cut‐off for ‘significance’ seems to be the direction of travel in medical science at large. This makes sense, as very little of what we do in clinical practice follows the same rules, and the uncertainty we embrace as doctors is not accounted for by hypothesis testing using clear cut‐offs. For a large survey, the use of an estimate of degree of association, in the form of an OR, seems reasonable. Using blanket p values for all comparisons when the number of participants is over 10,000 may nearly always result in statistical ‘significance’.

There are two further limitations worthy of comment. Firstly, it was not possible to determine the consistency of responses from those within the same department. This would allow us to quantify uncertainty, as the number of respondents from the same department answering ‘yes’ or ‘no’ about the existence of local guidelines would be known. Uncertainty could then be factored into the results. Secondly, we were surprised to read that 978 (10.2%) respondents were trained in gastric ultrasound. This is because we (MC and KE) are two of a small number of clinicians in the UK who use gastric ultrasound regularly in our own clinical practice. There are two possible explanations. One is that the use of pre‐operative gastric ultrasound is far more widespread in other parts of the world. The second is multiple responses (more than one response per individual) were logged from individuals with gastric ultrasound expertise. In the case of the latter, this would suggest the results may be prone to bias.

Consensus

In the past, the recipe for RSI would consist of pre‐oxygenation, pre‐defined doses of thiopentone and suxamethonium, direct laryngoscopy and intubation with a tracheal tube with the application of cricoid force, which might be released, should difficulty be encountered 17. There now exists a range of other options, with variations in pharmacological agents and airway equipment to name but two. The large sample size included by Zdravkovic et al. reinforces this variation but tells us little on which pharmacological or airway choices are preferred, but variation on these and other practices are likely increasing too. Nevertheless, we argue the old questions on the use of thiopentone vs. propofol, suxamethonium vs. rocuronium and opioid vs. no opioid are becoming increasingly irrelevant. Some may argue that consensus on RSI, or even cricoid force, are aims that are themselves outdated. Instead, we might wish to consider whether RSI has a place in contemporary practice, and what it actually means when say we are going to perform it as part of our anaesthetic technique.

Future directions

This apparent inconsistency in clinical practice has emerged in part due to a poor evidence base supporting individual interventions. Tried and tested strategies for RSI, along with pragmatic challenges in conducting research in this area, have left high‐quality data wanting. This represents fertile ground for research to elucidate the role of gastric ultrasound in RSI, optimal patient positioning, and perhaps most debatable, the true impact of cricoid force on patient outcomes following RSI in different patient settings 18. We might, however, never be able to generate this evidence, or even define a range of pragmatic end‐points that should be measured. Instead, we might wish to develop consensus on strategies to reduce pulmonary aspiration risk, as well as other RSI‐related complications.

Conclusion

The ‘classical’ RSI seems to have evolved to the extent that there seems to be widespread confusion on what exactly a ‘modified’ or contemporary RSI entails. We argue this confusion can only be overcome by recognising the process of an RSI not as a prescriptive approach to induction of anaesthesia and tracheal intubation, and more about tailoring practices to avoid complications in high‐risk patients. These complications include: haemodynamic compromise; pulmonary aspiration and other major airway complications; awareness during general anaesthesia; and clinically significant end‐organ damage. Any future consensus should perhaps focus less on questions that might now be seen by many as increasingly irrelevant, and more on how we identify patients at risk of RSI‐related complications, particularly pulmonary aspiration, and how to safely mitigate against these. For reminding us of the importance of pulmonary aspiration risk stratification and the need for consensus on strategies to reduce its incidence, Zdravkovic et al. should be congratulated.