Summary of main results

Our systematic review reveals several important findings. Analysis of 34 trials reporting on conditions regarding tracheal intubation suggests that avoidance of neuromuscular blocking agents (NMBA) is associated with increased risk of a difficult tracheal intubation (DTI) (summary of findings Table for the main comparison). We confirmed this in our assessment of the four trials categorized as having low risk of bias (Analysis 1.1). Here, avoidance of NMBA was associated with DTI with a risk ratio (RR) of 13.27 (95% confidence interval (CI) 8.19 to 21.49; P < 0.00001; number needed to treat for an additional harmful outcome (NNTH) =1.9, I² = 0%). Results showed a significant subgroup difference between trials with low and high or uncertain risk of bias, as avoiding NMBA exceeded the risk ratio of a difficult intubation among trials with low risk of bias compared with trials with high or uncertain risk of bias. Fixed‐effect meta‐analyses did not noticeably change the conclusions. Our trial sequential analysis (TSA) provided firm evidence for the least possible harmful effect on the proportion of DTI in the traditional meta‐analysis of avoiding NMBA compared with using NMBA, even when the significance level was adjusted for repetitive testing and sparse data in a cumulative meta‐analysis.

In our subgroup analyses on DTI, use of remifentanil, alfentanil, or local anaesthesia did not change our primary finding, as avoidance of NMBA was significantly associated with a DTI. However, our subgroup analysis suggests that local anaesthetics may have a protective effect, as results showed a statistically significant subgroup difference between trials using and trials avoiding local anaesthesia in relation to tracheal intubation. Likewise, in the subgroup analysis of trials excluding participants with expected difficult intubation, avoiding NMBA increased the risk of a difficult intubation. Fixed‐effect meta‐analyses did not noticeably change results of the subgroup analyses.

Our analysis of seven trials suggests that avoidance of NMBA is associated with increased risk of upper airway discomfort or injury (RR = 1.37, 95% CI 1.09 to 1.74; P = 0.008; NNTH = 9.1, I² = 13%). Only one trial reporting this outcome had low risk of bias, but results revealed a significant association between avoidance of NMBA and upper airway discomfort or injury (Mencke 2003). Our TSA provided firm evidence for a harmful effect on the proportion of upper airway discomfort and injury in the traditional meta‐analysis when the significance level is adjusted for repetitive testing and sparse data in a cumulative meta‐analysis.

For 13 trials describing conditions for laryngoscopy, our meta‐analysis showed that avoiding NMBA was associated with a difficult laryngoscopy, with RR of 2.54 (95% CI 1.53 to 4.21; P = 0.0003; NNTH = 14.6) (Beck 1993; Bouvet 2008; Combes 2007; Dominici 1990; González Obregón 2010; Harsten 1997; McNeil 2000; Mencke 2003; Mencke 2014; Pang 2014; Scheller 1992; Stevens 1997; Striebel 1995). Only one trial categorized as having low risk of bias reported this outcome (Mencke 2003). Results show no significant association between avoidance of NMBA and difficult laryngoscopy. Because information size was inadequate, our TSA of all 13 trials provided no firm evidence for a harmful effect of avoiding NMBA on upper airway discomfort or injury.

No trials reported mortality, and only one trial reported one episode of a serious adverse event (SAE) related to the use of NMBA (Kahwaji 1997).

Overall completeness and applicability of evidence

The plan to analyse effects of neuromuscular blocking agents for improving conditions during tracheal intubation or direct laryngoscopy in randomized clinical trials in adults followed our published protocol to a great extent. We included all eligible randomized clinical trials up to January 2017. Most of these trials were conducted in high‐income countries, and most participants were undergoing elective surgery. We included participants of both genders, and most participants were American Society of Anesthesiologists (ASA) class I or II and non‐obese, without expected airway management difficulties. Studies showed a high degree of clinical diversity as included combinations of different adjuvant drugs and NMBA were substantial.

About half of the included trials used a (modified) intubation score according to "Good clinical research practice (GCRP) in pharmacodynamic studies of neuromuscular blocking agents" (Fuchs‐Buder 2007; Viby‐Mogensen 1996). Further, a major portion of the remaining trials used intubation scores that incorporated many of the elements described by the GCRP score. Thus, to some extent, included trials showed some agreement on how DTI should be defined. Although trials used a defined outcome in accordance with recommendations for research regarding NMBA, definitions used differ from those traditionally used in research dealing with difficult airway management (Adnet 1997; Lundstrøm 2011b). Although the literature lacks consensus, these definitions often describe multiple (unsuccessful) attempts at intubation and/or shift to more advanced intubation techniques and/or other personnel performing airway management. One must therefore bear in mind that our primary outcome may not reflect a situation that most clinicians consider a truly difficult intubation that threatens the patient's life or health. Thus, one may observe a substantial degree of indirectness concerning our primary outcome describing DTI. Thirteen trials used the difficult laryngoscopy score of Cormack and Lehane (Cormack 1984) as a surrogate measure of a difficult intubation. This outcome measure is often used in studies on difficult airway management; thus it seems reasonable to interpret our findings on difficult laryngoscopy in a traditionally difficult airway management context. Most trials excluded patients expected to have difficult airway management. This may have a significant impact on the applicability of evidence, but subgroup analysis of 'Exclusion of patients with expected DTI versus no exclusion of expected DTI' (Analysis 1.6) showed no significant subgroup differences. Further, several studies have demonstrated that prediction of a DTI is difficult (Lundstrøm 2011b; Nørskov 2015; Nørskov 2016) and involves low sensitivity and specificity; thus the impact of these exclusions is uncertain.

Selected outcome measures varied among trials dealing with sore throat, upper airway discomfort, and injury. Further, by whom, how, or when the outcome measure was evaluated differed or was not clearly defined by some of these trials ‐ a fact that must be taken into account when our results are interpreted in a clinical context.

Of specific notice is the fact that evaluation of the effect on occurrence of serious adverse events (SAE) of avoidance of NMBA compared with use of NMBA is virtually absent, as none of the included trials evaluated effects on mortality (short‐term or long‐term), and only one trial evaluated effects on other types of SAEs, reporting one SAE.

Quality of the evidence

We present our main results on all investigated outcomes in summary of findings Table for the main comparison, which shows grading of the quality of evidence, including imprecision, according to the GRADE approach (Guyatt 2013; Guyatt 2011) and stresses results of meta‐analyses originating from trials with overall low risk of bias. Our review follows the overall plan of a published, peer‐reviewed Cochrane protocol (Lundstrøm 2011a). It represents a comprehensive review of the topic, including meta‐analyses of results from 34 randomized trials with a total of 3565 participants. To our knowledge, no previous meta‐analyses have included trials comparing use or avoidance of NMBA for tracheal intubation. We conducted a thorough review in accordance with methods of the Cochrane Collaboration (Higgins 2011). Between‐trial heterogeneity varied from absent to substantial among the various meta‐analyses in terms of our different outcomes. This may emphasize the diversity of the dose‐finding regimens used in different trials that used substantially different combinations of adjuvant drugs and NMBA. Further, minor differences among definitions of a DTI in some degree may contribute to the heterogeneity. In contrast, we noted no between‐trial heterogeneity regarding our secondary outcome describing conditions needed for direct laryngoscopy, which may be explained by a well‐defined outcome measure (Cormack 1984) that was retrievable from all 13 trials.

We performed trial sequential analyses and calculated TSA‐adjusted confidence intervals reflecting sparse data and multiple testing due to potentially repeated updates of cumulative meta‐analyses to control risk of random errors and to prevent premature statements of superiority of one intervention over another (Brok 2008; Brok 2009; Thorlund 2009;Thorlund 2011; Wetterslev 2008; Wetterslev 2009). TSA revealed that all of our meta‐analyses including the primary outcomes have a very low degree of imprecision, indicating low risk of random error. The finding of an insignificant TSA‐adjusted association between avoidance of NMBA and difficult laryngoscopy may be due to a type II error, that is, a false‐negative finding, exemplified by the cumulative z‐curve not crossing the trial sequential monitoring boundary (TSMB) for futility. However, such an analysis cannot remove risks of bias ‐ detected or undetected. It is worth discussing how much evidence is required when one is dealing with potential benefit or harm. On the other hand, beneficial or harmful effects can occur as the result of random errors; therefore, sufficient information must be assessed to demonstrate benefit or harm beyond reasonable doubt.

Because actual airway management was evaluated just after induction of anaesthesia, no cases were lost to follow‐up. However, cases with upper airway discomfort and/or injury were lost to follow‐up in three trials (Combes 2007; Mencke 2003; Mencke 2014). We therefore performed post hoc "best‐ and worst‐case" sensitivity analyses. We included participants who were lost to follow‐up with or without upper airway discomfort and/or injury. These explorative analyses did not alter our primary findings derived both from all trials and from trials with low risk of bias.

Potential biases in the review process

We strived to reduce bias by identifying all relevant trials through a comprehensive systematic search of the literature. We contacted study authors to retrieve unpublished data, when possible. However, it was difficult to obtain updated contact information for authors of the oldest published trials. This may have introduced bias, as it was generally easier to retrieve additional data from more recently published trials, which also were more likely to be categorized as having low risk of bias. Two review authors assessed trials for inclusion or eligibility and extracted all data in duplicate, thereby reaching a high level of agreement. We did not conduct quality assessments or data extractions while blinded to review authors and bias risks.

Most included trials clearly stated cut‐off values defining a DTI. However, five trials (Beck 1993; Gulhas 2013; McNeil 2000; Scheller 1992; Stevens 1997) used different composite scores without defined cut‐offs. We used underlying references or comparable scores to define cut‐off values, enabling us to include these trials in our meta‐analyses. The Characteristics of included studies section provides detailed explanations. In these cases, some degree of interpretation of outcome measures was necessary, and our assessments may have introduced bias as a consequence. Nevertheless, we believe that this categorization was evident when compared with the existing literature, and we therefore consider risk of introducing bias as low.

The authors of six trials (Harsten 1997; Jiao 2014; Mencke 2003; Mencke 2014; Nimmo 1995; Sivalingam 2001) excluded a few participants from their assessments when tracheal intubation failed. We consider these patients as truly difficult to intubate and thus chose to include them in our meta‐analyses. This may have introduced bias; however, a sensitivity analysis excluding these participants from our meta‐analyses did not change the conclusion.

Agreements and disagreements with other studies or reviews

To our knowledge, no previous systematic review has included meta‐analyses of trials comparing use or avoidance of NMBA for tracheal intubation. However, a narrative review (Woods 2005) concluded that the literature describes successful techniques to intubate the trachea without the use of neuromuscular blocking agents with the patient under general anaesthesia. Further, those review authors concluded that these techniques offer a useful alternative when drugs are contraindicated or undesirable. Another narrative review (Fotopouloua 2012) concluded that induction of anaesthesia without the use of NMBA but in combination with remifentanil provides acceptable conditions for tracheal intubation. Both reviews include several dose‐finding trials in which none of the participants in the different intervention groups were administered NMBA. Thus, it was impossible to compare avoiding versus using NMBA for tracheal intubation. On the basis of findings related to specific drug combinations and concentrations, the review authors concluded that induction without the use of NMBA may offer (almost) perfect or good conditions for intubation.

In large observational studies and reviews on difficult airway management, fractions of DTI range from 2% to 7% (Lundstrøm 2011b; Nørskov 2015; Shiga 2005), which is consistent with our findings on the overall proportion of difficult intubation in patients who are intubated with use of NMBA. The crude proportion was 4.7% in trials with low risk of bias and 8.1% in all trials. However, among participants undergoing tracheal intubation without the use of NMBA, corresponding crude proportions were tremendously higher, at 56.3% in trials with low risk of bias and 24.2% in all trials. These latter proportions are not consistent with those reported in experiences from everyday clinical practice, and several trials may therefore include suboptimal dosing regimens. Thus, we speculate that some of the dose‐finding trials may include control groups with clinically unacceptable dosing regimens that may contribute to an unrealistically increased risk of DTI. As an example, one trial concluded that the optimal remifentanil dose used for intubation with propofol and without NMBA is 4 µg/kg administered in 60 seconds (Bouvet 2009). However in our review, only two of the included trials actually used equivalent or larger doses of remifentanil (Hanna 2010; McNeil 2000). Therefore, we performed a post hoc sensitivity analysis of a best‐case scenario (Analysis 1.7). Dose‐finding trials were represented only by control and intervention groups with the lowest incidence of difficult intubation. Results showed a crude proportion of 25.0% DTI among participants anaesthetized without the use of NMBA, and this again significantly increased the risk of DTI, with a risk ratio of 5.99. Finally, in our attempt to exclude comparisons of suboptimal anaesthetic dosing regimens, we performed a post hoc sensitivity analysis while excluding all dose‐finding trials (Analysis 1.8). We included in our meta‐analysis 16 trials with a single intervention group and a single control group. The crude proportion of DTI among participants anaesthetized without the use of NMBA decreased to 15.4%. However, avoiding NMBA still significantly increased the risk of DTI, with a risk ratio of 3.40. Another possible explanation for the increased crude proportion of DTI among participants induced without NMBA may be that most included trials used a (modified) intubation score according to "Good clinical research practice (GCRP) in pharmacodynamic studies of neuromuscular blocking agents" (Fuchs‐Buder 2007; Viby‐Mogensen 1996). Assessment of intubation conditions included information on jaw mobility and/or reactions to insertion of the tracheal tube and cuff inflation (diaphragmatic movement/coughing). These are reasonable elements for describing ideal conditions for intubation, although they may be affected even when everyday clinical airway management is successful without "true" difficulties. Thus, the GCRP score may be highly sensitive and may be used to enforce the difference between the ideal procedure and a clinically acceptable intubation. Trials reporting conditions for laryngoscopy (described by Cormack 1984) report an outcome measure traditionally used in studies dealing with difficult airway management. In our assessment, the crude proportions of difficult laryngoscopy reported among intervention groups may reflect a more clinically acceptable situation, as 4.2% among all trials and 7.2% in trials with low risk of bias who were intubated without NMBA had a difficult laryngoscopy.

In the literature, one review reported on various causes for postoperative sore throat (El‐Boghdadly 2016). These review authors state that tracheal intubation without the use of NMBA is a potential risk factor for sore throat. Although this study is presented as a systematic review, review authors included only one study (Combes 2007) reporting on upper airway discomfort or injury, review authors performed no bias assessment, and the review included no meta‐analyses. Other conditions such as tube size (Jaensson 2010; Jaensson 2012; Stout 1987; Xu 2012) and use of corticosteroids (Sumathi 2007; Thomas 2007) may play an important role in the postoperative sore throat. In our review, we did not evaluate the impact of these factors in subgroup analyses. The review of EL‐Boghdadly reports postoperative sore throat with a prevalence of up to 62% following general anaesthesia. In a Cochrane review on lidocaine for preventing postoperative sore throat (Tanaka 2015), the crude prevalence of sore throat was 20% to 30% pending the use of lidocaine. In our assessment, the crude prevalence was 33%.