Closing in on the best supraglottic airway for paediatric anaesthesia?

In this month's issue of the journal, Mihara et al. provide an overview of the efficacy of the numerous supraglottic airway devices (SAD) for paediatric anaesthesia currently available 1. Using a network meta-analysis for comparing randomised studies of 16 different devices, the authors suggest that the LMA-Proseal®1 may be the best device for children, based on its low risk of failure and high leak pressure. This result aligns with a recent narrative review 2. This editorial will deal both with the statistical and clinical aspects of this claim. There are several reasons why meta-analysis is beneficial in this context. When there are multiple trials assessing two SADs, there can be differences in the results of these trials. This could be because each trial was only carried out on a subset of the patient population (e.g. particularly light or heavy patients), who each react to the devices in slightly different ways, or because of differences in the way devices are used in each trial. Most of these trials will give us some information about the efficacy of each device, but no one trial will give us perfect information. The hope is that if we analyse the results of all these trials in one meta-analysis we get the best possible estimate of the true effect of each device. Another reason why meta-analysis is so important is because smaller trials on their own may not have the power to detect differences between devices. A meta-analysis of multiple trials has larger power to detect differences and ensure that the best possible SAD is identified. Network meta-analysis is an extension of a meta-analysis. There has been much work into the potential benefits and drawbacks of network meta-analysis 3-9. Due to the large number of SADs available, it is not practical to assess all options in one trial. Network meta-analysis provides the opportunity to compare and rank all 16 of the devices. If we compare two devices in one trial this is called a direct comparison. However, some devices will never have been studied ‘head to head’ in a trial. For example, the LMA-Proseal has never been compared directly to LMA-Unique®, but they have both been compared to LMA-Classic®. Therefore, we can compare the efficacy of LMA-Proseal and LMA-Unique by an indirect comparison through the LMA-Classic. A network meta-analysis takes all possible direct and indirect comparisons into account. This can improve the precision associated with each estimate due to the extra evidence incorporated. However, there is potential for indirect comparisons to be less reliable than direct comparisons. Characteristics in patients that affect efficacy are referred to as effect modifiers, in this case body weight. If the body weights of patients in two trials differ, this can lead to a difference in the relative effect of devices from trial to trial. This can be assessed by examining potential effect-modifiers, and by looking at the consistency of the network, which assesses the agreement between direct and indirect comparisons. In the case of a Bayesian network meta-analysis, the synthesis can also be sensitive to the specification of prior distributions. It is notable that Mihara et al. have followed the guidance of the National Institute for Health and Care Excellence (NICE) Decision Support Unit in carrying out their analysis and associated consistency checks and sensitivity analyses 10, as well as the PRISMA extension statement in their reporting of the network meta-analysis (http://www.prisma-statement.org/Extensions/NetworkMetaAnalysis.aspx). Network meta-analysis has become increasingly popular in recent years; it is now commonly used in health technology assessments and drug re-imbursement decisions. There are a number of guidelines available for network meta-analysis 10-13, including guidelines from the World Health Organization (WHO), which has recently advocated its use. In a recent WHO bulletin 11, it was noted that 10% of Cochrane reviews published since 2015 have used network meta-analysis. It is important to acknowledge that performing a network meta-analysis – even on large numbers of trials – does not automatically provide the required statistical power to perform unlimited numbers of comparisons and subgroup analyses 13-15. In particular, multiple comparisons of many outcomes and subgroups are presented, unadjusted for multiple testing, and it is thus incorrect to interpret findings (e.g. null effects being outside a given confidence interval) as if a single test of statistical significance had been carried out 15, 16. Equally, it is important to note that one cannot make claims of superiority, benefit or harm from the results of a network meta-analysis in the same way as one might in a prospective trial 16, 17. As there are many different SADs on the market, a ‘patchwork’ of different comparative trials has emerged, making it extremely difficult to assess publication bias. Mihara et al. have been able to provide funnel plots (a graphic showing potential publication bias and asymmetry testing 18) on a few of the comparisons between the many SADs; however, for the majority of products, the risk that there are data from unpublished trials results remains impossible to assess. Bearing the above-mentioned aspects in mind, Mihara et al. found that oropharyngeal leak pressure was higher with the i-gel®, LMA-Proseal and the Cobra® perilaryngeal airway, and that the risk of bloodstaining was lower with the i-gel. Additionally, the authors defined a new exploratory outcome, which indicated that device failure may be lowest with the LMA-Proseal, LMA-Classic and LMA-Unique. Further, the risk of insertion failure at first attempt was higher in a subgroup analysis on patients with body weight less than 10 kg. Besides the methodological considerations mentioned in the above, how should we then interpret these results in a clinical context? The supraglottic airway has, since its introduction, undergone significant development and devices are now classified as either first or second generation SADs, with important functional differences such that some clinicians even argue for abandoning first generation devices completely 19. Therefore, some important scientific questions are still left unanswered; should the clinician choose a first or a second generation SAD?; what is the right SAD for a specific paediatric weight group; and which SAD should you use for a specific clinical purpose e.g. procedures with preservation of spontaneous breathing, difficult airway management, resuscitation, and prehospital airway management? Certainly, not all of these questions can be answered using a randomised design, and consequently answers may only be found through observational studies and from trials on non-emergency patients. This is another area where network meta-analysis can be useful, as it can allow inclusion of different study designs 20. Second generation SADs are recommended in difficult airway management guidelines from 2015 21, 22 as they: provide a better seal; can serve as a conduit for tracheal intubation; have a channel for an orogastric tube (thus reducing risk of aspiration); and provide an option for continuous oxygenation during attempts to intubate the trachea 23. The first generation devices, on the other hand, are generally much cheaper. In keeping with the findings of Mihara et al., the improved seal of some second generation SADs also supports their use in patients requiring higher ventilation pressures. Despite these obvious advantages, second generation SADs are unsuitable for certain surgical procedures such as tonsillectomy and adenoidectomy 24. It is interesting that two out of three devices with the lowest failure rate in the analysis were first generation SADs. Does this imply that first generation devices would be the first choice in a difficult airway emergency situation (thereby contradicting recommendations in the Difficult Airway Society 2015 guidelines) 21? On the contrary, not having the combined possibility of gastric drainage, a tight seal and the ability to perform tracheal intubation through the device would discount the use of first generation devices in this context. That leaves only second generation SADs as suitable devices in emergencies, consistent with these devices’ place in difficult airway management algorithms. Another result of the network meta-analysis which should be noted is that the risk of failure at the first attempt was 11.7% with the LMA-Classic, with almost all devices performing within a similar range. In an emergency situation, anaesthetists should bear this failure rate in mind in order to make the first SAD attempt optimal. The authors address the difference in body weight as the cause of heterogeneity between studies. Children undergo rapid growth, with large anatomical changes, and this makes them a heterogeneous population and means that it is difficult to draw conclusions on airway management, although airway and ventilation incidents are common 25. As such, choosing a cut-off point of 10 kg mean body weight may over-simplify the comparison for the purposes of evaluating the performance of a certain SAD; further stratification by weight may be justified in order to properly establish which SAD might be appropriate for a specific paediatric weight group. At the moment, the data seem too sparse to draw firm conclusions. Nevertheless, the paper reports an increased first attempt failure for small children, with an odds ratio (95%CI) of 5.1 (1.6–20.1), which could indicate that we should be even more prepared for unsuccessful SAD placement in the very youngest group of children. So, although the authors propose the LMA-Proseal as the best SAD option for children, there is probably no such one ‘best device’ covering all situations and all age groups of children. Adoption of new devices into clinical practice is a slow process and is influenced, among other things, by personal factors, departmental preference and practice guidelines. Scientific evidence of safety and best performance should be the most pertinent factor when choosing devices for clinical practice. Much of this burden lies on the shoulders of the industry developing new equipment. What has become clear is that regulations on new devices are often not as strict as when new medication is introduced into the market. That led the Difficult Airway Society to assemble the Airway Device Evaluation Project Team (ADEPT) in 2011, with the aim of strengthening the evidence by which new airway devices were introduced into clinical practice 26. Last year, a multicentre trial aiming to test second generation SAD was instigated under this initiative (https://www.das.uk.com/node/424). We can only support any effort for improving the evidence on which clinicians can base their choices when implementing all new equipment – not just airway devices. Seen in this light, the analysis of Mihara et al. offers some new knowledge on the efficacy of SAD. However, it seems that there may still be some way to go before we can draw the full picture of the advantages and disadvantages of the many devices available in today's market.