Video versus direct laryngoscopy in critically ill patients: an updated systematic review and meta-analysis of randomized controlled trials

We systematically searched PubMed, Embase, Cochrane Library, and ClinicalTrials.gov from inception to June 23, 2023. The search terms used included ‘video’, ‘intubation’, and ‘laryngoscope’. The complete search strategy is provided in Additional 1: Supplemental Methods 3. Two authors (B.A. and S.L.) independently screened titles and abstracts and evaluated the articles in full for eligibility based on prespecified criteria. Discrepancies were resolved in a panel discussion with a third author (A.R.). Moreover, we used backward snowballing (i.e., review of references) to identify relevant texts from articles identified in the original search.

We considered studies eligible for inclusion if they (1) were RCTs; (2) directly compared VL versus DL; (3) enrolled critically ill patients (admitted to ED or ICU); (4) included adult patients; and (5) presented data regarding any of the prespecified efficacy and safety endpoints. The exclusion criteria were non-randomized studies, quasi-RCTs, cluster RCTs, studies that included patients younger than 16 years old or pregnant patients, studies centered on surgical scenarios, or conference abstracts.

Four authors (B.A., S.M., P.C., and M.S.) independently extracted the data for each study using a standardized study form to determine: authors, clinical trial registration, enrollment period, study publication year, main exclusion criteria (Additional file 1: Supplemental Methods 4), sample size, follow-up period, endpoint definition, baseline patient characteristics, and operator’s characteristics. Any discrepancies were settled through a panel discussion with a fifth author (A.R.).

The definition of operators' experiences slightly varied among studies. To allow subgroup analysis based on this characteristic, we classified operators into two groups, experienced and inexperienced, following specific criteria outlined in Additional file 1: Supplemental Methods 5. Moreover, the Additional file 1: Supplemental Methods 6 highlights how each study selected the device for the second intubation attempt. The classification of a difficult airway was made in accordance with either the study’s definition or the Mallampati 3/4 classification.

Our primary efficacy endpoint was (1) successful intubation on the first attempt, as defined by each individual study. Other efficacy endpoints were (2) successful intubation on the second attempt, (3) Cormack Lehane (CL) laryngeal view grade I, and (4) CL laryngeal view grade I/II. Safety endpoints were (5) incidence of aspiration, (6) esophageal intubation, (7) cardiac arrest, (8) severe hypoxemia, (9) dental injury, and (10) all-cause mortality. Additional file 1: Supplemental Methods 7 describes the endpoint definition of some outcomes.

We conducted prespecified subgroup analyses for the primary outcome. Studies were grouped based on the (1) VL brands and (2) operators’ experience. A sensitivity analysis of the subgroup analysis evaluating the operator’s experience was performed changing the threshold of from 50 to 100 prior intubations to be considered experienced. Subgroup analyses were performed if two or more studies were available in the group.

Two independent authors (B.A. and R.A.) assessed the risk of bias in the included RCTs using Cochrane’s Collaboration tool for assessing the risk of bias in randomized trials (RoB 2) [18]. Any disagreements were resolved through consensus between authors. We explored the potential for publication bias by visual inspection of the comparison-adjusted funnel plots and Egger’s test for the primary endpoint.

We used the random-effects model for all outcomes. We employed risk ratios (RRs) and 95% confidence intervals (CIs) as the measure of effect size for binary endpoints. For continuous endpoints, we utilized weighted mean differences (MDs). Restricted maximum likelihood estimator was used to calculate heterogeneity variance t². We assessed heterogeneity with Cochrane’s Q statistic and Higgins and Thompson’s I² statistic, with p ≤ 0.10 indicating statistical significance. We determined the consistency of the studies based on I² values of 0%, ≤ 25%, ≤ 50%, and > 50%, indicating no observed, low, moderate, and substantial heterogeneity, respectively. All tests were two-tailed, and a p value of < 0.05 was considered statistically significant. If necessary, means and standard deviations were estimated [19]. We conducted a trial sequential analysis (TSA) using random-effects model for the primary outcome, we used a statistical significance level of 5% and a beta of 80%. We used TSA version 0.9.5.10 beta (Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet, Copenhagen, Denmark). We used R version 4.3.1 and the extension packages "meta," "metafor", "dmetar", "ggplot2", and "forestplot" for all calculations and graphics [20‐23]. An in-depth description of the statistical analyses is available in Additional 1: Supplemental Methods 8.

Our systematic search yielded 4278 potential articles (Fig. 1). After removing duplicates, 72 articles were retrieved and reviewed in full for possible inclusion. Of these, 14 RCTs met all inclusion criteria and were included in the primary analysis [11, 12, 14, 15, 24‐32]. We included a total of 3981 patients, of whom 2002 (50.3%) patients were assigned to VL and 1979 (49.7%) were assigned to DL. The mean age of patients in individual studies ranged from 37 to 69 years, and the proportion of males was 63.7%. Table 1 summarizes the main characteristics of the included studies.

Compared with DL, VL significantly increased the number of successful intubations on the first attempt (RR 1.12; 95% CI 1.04–1.20; p < 0.01; I² = 82%; Fig. 2A), the proportion of CL grade I (RR 1.73; 95% CI 1.41–2.12; p < 0.01; I² = 71%; Fig. 2B), and grade I/II (RR 1.12; 95% CI 1.04–1.19; p < 0.01; I ² = 91%; Additional 1: Supplemental Fig. 1). However, no statistically significant difference was observed between the groups in terms of success on the second attempt (RR 1.04; 95% CI 0.94–1.15; p = 0.49; I² = 83%; Supplemental Fig. 2). Regarding TSA, the cumulative Z-curve crossed the required information size obtained with the 3032 subjects, indicating a low chance of type 1 error for successful intubation on the first attempt (Fig. 3).

VL substantially reduced the number of esophageal intubations (RR 0.44; 95% CI 0.24–0.80; p < 0.01; I² = 0%; Fig. 4A) and aspirations (RR 0.63; 95% CI 0.41–0.96; p = 0.03; I² = 0%; Fig. 4B) compared to DL. However, there were similar incidences of dental injury (RR 0.67; 95% CI 0.20–2.24; p = 0.51; I² = 0%; Additional 1: Supplemental Fig. 3A), cardiac arrest (RR 1.66; 95% CI 0.52–5.30; p = 0.39; I² = 0%; Additional 1: Supplemental Fig. 3B), all-cause mortality (RR 1.00; 95% CI 0.87–1.16; p = 0.95; I² = 0%; Additional 1: Supplemental Fig. 3C), and severe hypoxemia (RR 0.98; 95% CI 0.74–1.29; p = 0.87; I² = 22%; Additional 1: Supplemental Fig. 3D).

There was a significant subgroup interaction among the brands of VL employed (p = 0.03; Additional 1: Supplemental Fig. 4A). C-MAC and GlideScope performed similarly, but with McGrath MAC, there was no significant difference between VL and DL (RR 0.99; 95% CI 0.96–1.02; p = 0.43; I² = 11%). Furthermore, there were no significant subgroup interactions when analyzing subgroups stratified by settings (ICU versus ED) (p = 0.48; Additional 1: Supplemental Fig. 4B) or operators’ experience (p = 0.42; Fig. 5). In sensitivity analysis changing the threshold from 50 to 100 prior intubations to be considered experienced, there was no significant subgroup interaction (p = 0.53; Additional 1: Supplemental Fig. 5).

We conducted a Graphic Display of Heterogeneity (GOSH) analysis to investigate the moderate to high heterogeneity in our findings. Our results were consistent across multiple simulations and remained stable after random exclusion of studies. We identified one study as the main outlier [25]. A comprehensive explanation of statistical protocols used to explore heterogeneity is available in Additional 1: Supplemental Results 1 and Additional 1: Supplemental Figs. 6–8, 11, and 12.

Individual RCT appraisal can be found in Additional 1: Supplemental Fig. 9. Regarding the primary outcome, thirteen studies carried high risk of bias due to unblinding of outcome adjudicators due to the nature of intervention, however the DEVICE trial was scored at a low risk of bias due to the presence of an independent observer keeping track of the number of intubation attempts [12]. Moreover, seven studies had some concerns of bias due to the inexistence of protocols [14, 15, 24‐27, 33]. Funnel plot and Egger’s test (p = 0.048) suggested publication bias in the primary outcome, as represented in Additional 1: Supplemental Fig. 10.

This meta-analysis has some limitations. First, there was a substantial heterogeneity in the primary outcome. However, we meticulously addressed this heterogeneity by exploring the potential study-level characteristics, as reported in the Additional 1: Supplementary Appendix. Second, our analysis indicated the presence of publication bias concerning the primary outcome. Third, we identified an elevated risk of bias due to the outcome adjudication of the primary outcome, primarily because blinding was impossible due to its inherent nature. Fourth, aspiration relied on operator-reported data, which may be subject to reporting bias. Fifth, only one of included studies reported the presence of secretion as a reason of intubation failure. Therefore, it was not possible to analyze this important variable. Sixth, the subgroup analysis on different VL brands used by individual studies should be interpreted cautiously, as different manufacturers could provide both standard geometry and hyperangulated blades, which impact could not be analyzed. Finally, the absence of patient-level data precluded a more granular assessment of factors potentially related to the relative efficacy of VL vs. DL, such as the operators’ experience and proportion of patients with difficult airways.