Background
Securing the airway with tracheal intubation (TI) is a fundamental treatment for emergency and critical care patients with respiratory dysfunction or decreased airway protection. Direct laryngoscopy (DL) is the primary method for TI, but it can be challenging when performed in emergencies because the patient often is in life-threatening condition and has the factors that make TI difficult, such as limited mouth opening, unstable cervical spine, blood or secretions in the airway, and facial trauma, and in addition the expertise of available practitioners varies [
1,
2]. The first-attempt success rate of urgent TI in emergency and critical patients is relatively low [
2‐
5], and unsuccessful or prolonged TI can be life-threatening and may result in severe complications [
6‐
11].
Video laryngoscopy (VL) is a new device that contains a miniaturized camera at the blade tip to visualize the glottis indirectly. This method was developed at the beginning of the 21st century [
12]. It has been shown that VL improves laryngeal visualization compared with DL [
13,
14] and provides some advantages in surgical patients, especially those with difficult airways [
15‐
19]. The use of VL in emergent and critical situations has also been tested in several observational studies, which have shown that VL can lead to better intubation outcomes [
4,
5,
20,
21]. A recent meta-analysis of intensive care unit (ICU) patients demonstrated that, compared with DL, VL reduces difficult intubation and increases the first-attempt success rate [
22]. Of the nine studies included in that meta-analysis, however, only three (
n = 157 subjects) were randomized controlled trials (RCTs) [
23‐
25]. An observational study, whether prospective, nonrandomized, or retrospective in design, does not control for the operators’ experience with each device or for patients’ conditions and thus may bias the determination of the efficacy of different airway devices.
Recently, the performance of VL and DL in patients needing emergency TI was compared in several RCTs, and some of them showed no benefit regarding success rate or intubation time with VL [
24,
26‐
32]. In view of this, we performed a systematic review and meta-analysis that included only RCTs comparing the performance of VL and DL for emergency TI with respect to the intubation outcomes and complications. Our review is registered with PROSPERO (
http://www.crd.york.ac.uk/PROSPERO, CRD42017054804).
Discussion
To our knowledge, this is the first meta-analysis and systematic review of available RCTs comparing VL and DL for TI in emergency and critical care patients, including the quality of evidence. In this analysis, the first-attempt success rate was used as the primary endpoint because multiple intubation attempts performed outside the operating room can significantly increase the risk of life-threatening complications [
6,
43,
44]. Furthermore, improving the first-attempt success rate has been regarded as the main goal of emergency TI [
45]. Our results show that laryngeal visualization was improved by using VL. This is consistent with findings for surgical patients in the operating room [
12]. However, better visualization did not translate into an improved first-attempt success rate or other intubation outcomes or complications, except for a lower rate of esophageal intubation. Prehospital intubation outcomes were even worsened with lower first-attempt and overall success rates with VL when TI was performed by experienced operators.
Evidence derived from surgical patients shows that VL is associated with better intubation outcomes, especially for inexperienced operators and patients with difficult airways [
16,
28,
46]. This is because TI in the operating room is controllable, such as with the common use of RSI and NMBAs, patients’ fasting state, and favorable oxygenation, as well as appropriate light or intubation position. For highly experienced anesthesiologists, it seems unlikely that a single device will show superiority unless a difficult airway is encountered [
47‐
51], whereas for novices who have not yet received long-term DL training, visualization of the airway on a video screen can allow their supervisors to directly assist them in completing an intubation themselves, thus reducing the number of attempts and improving the safety of airway management [
52]. However, emergent TI is quite another thing. Although TI in the emergency department or ICU is frequently performed by paramedics or emergency medicine physicians who do not practice TI with DL on a daily basis [
53], and although the patients often have a higher risk of difficult airways [
9,
54], the operators may not benefit from using VL as novices in the operating room. There are several uncontrollable factors that may explain this difference. First, critically ill patients with a poor oxygen reserve capacity are more subject to hypoxia, which makes it more likely that operators will turn to alternatives such as DL, a flexible or rigid bronchoscope, or at least further mask oxygenation. If TI is not completed within the allowed time, inexperienced operators will be replaced by more experienced operators earlier, making the first-attempt success rate much lower. Second, secretions or blood in the airway might impair laryngeal visualization with VLs [
26,
28]. Third, RSI and NMBAs will be chosen with caution owing to circulation compromise, certain airway problems, operators’ experience, or accessibility of medicine. Prehospital intubation is more challenging, owing to additional risk factors such as ambient light, limited workspace, special positioning, and chest compression during CPR [
55]. Under chest compression, increased intrathoracic pressure can cause reflux of gastric contents, resulting in more attempts and longer intubation time with the VL. Prolonged intubation time and subsequent hypoxemia have been identified as major reasons for increased mortality in patients undergoing prehospital intubation [
56]. In addition, in prehospital care, DL is more accessible, and most operators are experienced in using it.
It must be emphasized that performance of VL is different between devices owing to various designs and shapes [
57,
58]. Even a slight design modification may significantly change the success rate, intubation time, and use of adjunct maneuvers [
59]. Some types of VLs have their own design-related deficiencies that may dwarf their results. For example, the A.P. Advance™ VL (Venner Medical International, St Helier, Jersey, UK), with a large video screen, shows the plastic part of the blade tip instead of the relevant airway, contributing to its poor performance [
58]. Studies included in our analysis used three types of VLs (angulated, Macintosh, or channeled), including five different devices (GlideScope, C-MAC, McGrath MAC, Airwayscope, and Airtraq). In the prehospital setting, two of three included studies used channeled VL. The channeled VL, with its integrated design, might be more portable in the prehospital setting, but it is bulkier and may require other team members to maneuver the tracheal tube [
58]. It should be noted that the poor performance of the VL is due mainly to the prehospital setting itself rather than to the devices chosen. We therefore did a related subgroup analysis only in the in-hospital setting. No difference was identified between VLs and DLs, regardless of the devices used. Although an angled blade design was assumed to facilitate laryngeal visualization and thus to lead to a better intubation outcome, it may afford less room for tracheal tube insertion and increase stylet use in patients with a normal airway, resulting in increased procedural difficulty and prolonged intubation time [
25,
60]. In addition, pooling of results from studies evaluating different VLs may lead to intrinsic inconsistencies. An especially important issue neglected in the design of the five included studies comparing the Macintosh-type VL and DL is that the Macintosh-type VL can provide the two options of DL and VL in one device. When one attempt fails, the operators can immediately switch to another option to successfully complete the TI without having to make a second attempt [
61]. This unique feature of Macintosh-type VLs is significantly different from DLs and angulated VLs, which can provide only one option. Thus, definition of laryngoscopy attempts used in these studies is desirable for DLs but not for Macintosh-type VLs [
62].
The results of some studies indicate that VLs should be used with caution in critical patients because of a prolonged intubation time and subsequent possible higher incidence of severe life-threatening complications [
23,
25,
30]. Our review shows that incidences of aspiration, severe low oxygen saturation, and in-hospital deaths did not differ between VLs and DLs. However, these results remain unreliable owing to the limited number of participants included. Our review shows a lower rate of esophageal intubation using VLs than that in another study [
22]. This might be somewhat meaningful because “even a single episode of recognized esophageal intubation is associated with desaturation, increased risk of aspiration, and cardiac arrest” [
63]. Moreover, an important and promising finding in one of our included studies and another observational study is that the use of a VL has a higher first-attempt success rate with fewer chest compression interruptions in the emergency department [
29,
64].
Our study included only RCTs and quasi-RCTs. Although blinding was not adopted in most studies, we judged “no blinding” as low risk because it seems impossible to blind personnel in urgent situations at times. In the prehospital setting, moreover, there is never time for allocation concealment, and even randomization using a common method such as a random number table is impractical. Risk assessment of bias for the included studies showed that 7 of 12 studies could be classified as low-risk studies. Therefore, in general, this supports the quality of our study. The funnel plot, with its visually symmetrical distribution, qualitatively indicates a low risk of publication bias. Given that the quality of most evidence was low or moderate owing to a moderate or high level of heterogeneity, subgroup analysis and sensitivity analysis based on some potential clinical heterogeneous factors also were performed in our review.
There are some limitations of our review. First, although subgroup analyses were performed, there were still other clinical heterogeneities in subgroups, such as patients having different conditions, use of various intubation strategies, and use of any adjacent tool or maneuver. Whether patients with predicted difficult airways were enrolled was another important heterogeneous factor. However, for emergency or critical care patients, the traditional predictors of difficult airways, such as thyromental distance, Mallampati score, or neck mobility, cannot be recorded, because all intubations are performed so urgently that there is never a chance to make predictions or subsequent grouping before randomization. One observational study showed that VLs significantly increased the intubation success rate in emergency patients with difficult airways [
65]. In the absence of a difficult airway, however, the use of VLs may even bring some disadvantages [
25]. Whether anesthetics were used and the choice of medication can also introduce heterogeneity. RSI with sedatives, narcotics, and NMBAs has been shown to facilitate TI and decrease intubation-related complications in reasonable circumstances [
66,
67]. Because most of the studies included in our review did not have strict protocols regarding medication, subgroup analysis according to medications seemed impossible. Anyway, the study by Silverberg et al. [
24] demonstrated a much higher first-attempt success rate using VLs. Sensitivity analysis excluding this study did not alter the results, but the heterogeneities within the subgroups disappeared, indicating that this study may be the main factor leading to heterogeneity. The effect of the NMBAs on the result was unclear. It may be the negative influence of alternating of devices that use different configurations on the learning curve of operators with the DLs that led to a lower success rate with DLs. Second, owing to ethical considerations, some patients had to be excluded on enrollment, such as patients with low oxygen saturation [
24], those with an immobilized cervical spine, and patients with predicted difficult airways, or those excluded owing to attending physicians’ discretion and unavailability of devices at the time of eligible patient arrival [
25]. It is unclear whether these excluded patients would benefit from one of the interventions. Third, the classification of the operators’ qualifications and the definition of intubation time or overall success rate used in our analysis were based on previous papers or our own judgment, and this might somehow be arbitrary.