Study characteristics and quality assessment
According to the search strategy, a total of 996 related articles were retrieved. A total of 292 articles were excluded after duplicates were removed and 673 articles were initially included. After reading titles, abstracts and full texts, 667 articles were excluded, of which 35 were reviews. A total of 14 reports, 532 non-pre-hospital emergency studies, 40 that did not use laryngeal masks, 36 with non-randomized data and 7 with insufficient data were excluded. Finally, 9 randomized studies [8-16] on manikins and 31 randomized studies [17-47] on humans were included. In the manikin studies, the participants included medical students, paramedics, anesthesiologists, EMT, nurses, intensive care and emergency physicians. The human studies included cases of cardiac arrest, respiratory failure, coma and other critical illnesses. Flow diagrams for the literature selection are shown in Figure 1. The study characteristics are shown in Tables 1-2.
Meta-Analysis of laryngeal masks vs endotracheal intubation in manikin studies
Overall intubation success rate
A total of 8 randomized studies [8, 10-16] reported overall intubation success rates. No statistical heterogeneity was observed (P=0.939, I2 = 0.0%) so a fixed effect model was employed. The results showed that the overall intubation success rates of the laryngeal mask group were significantly higher than those of the endotracheal intubation group [RR=1.10, 95% CI (1.02, 1.18), P<0.00 1] (Figure 2).
Success rates of first intubation
A total of 4 randomized studies [8, 11, 13, 16] reported the success rates of the initial intubation. There was statistical heterogeneity (P = 0.000, I2 = 73.2%) so a random effects model was used. The result showed that the success rates of the initial intubation of the laryngeal mask group were significantly higher than those of the endotracheal intubation group [RR=1.25, 95% CI (1.01, 1.55), P<0.001] (Figure 3).
Eight randomized studies [9–16] reported differences in the insertion times. A random effects model was used for statistical heterogeneity (P = 0.000, I2 = 88.7%). The results showed significantly shorter insertion times of the laryngeal mask group [SMD=-1.53, 95% CI (-1.88, -1.17), P < 0.001] (Figure 4).
Meta-Analysis of laryngeal mask vs endotracheal intubation in human studies
Ventilation efficiency rates
Meta-analysis of 17 randomized studies [17-23, 27-29, 32, 34, 36, 38-40, 42] showed that the ventilation efficiency rates of the laryngeal mask groups were higher than those of the endotracheal intubation groups [RR=1.20, 95% CI (1.06, 1.35), P<0.001] without heterogeneity (P = 0.941, I2 = 0%) (Figure 5).
Overall intubation success rates and rates of initial intubation
A total of 19 [17, 20-23, 25, 28-29, 33-37, 39-42, 44-45] and 3 randomized studies [35, 41, 46] reported the success rates of initial intubation and overall intubation success rates, respectively. The results suggested that the success rates of initial intubation for the laryngeal mask were higher than those for endotracheal intubation [RR=1.29, 95% CI (1.18, 1.40), P<0.001] without heterogeneity (P = 0.840, I2 = 0%). However, no statistical differences in overall intubation success rates were observed [RR=1.11, 95% CI (0.88, 1.39), P<0.001] without heterogeneity (P = 0.979, I2 = 0%) (Figure 6).
Insertion time and SpO2 rise time
A total of 27 [17, 20-31, 33-37, 39-47] and 3 studies [17, 20, 33] reported insertion and SpO2 rise times, respectively. The results suggested that both were shorter in the laryngeal mask groups compared to endotracheal intubation [SMD=-3.48, 95% CI (-4.17, -2.80), P < 0.001; -2.19, 95% CI (-3.06, -1.32), P < 0.001] but with significant heterogeneity (P = 0.000, I2 = 96.8%; P = 0.001, I2 = 85.5%) (Figure 7).
Blood gas index
Three studies [19, 32, 37] reported PaCO2, PaO2 and pH, and 8 studies [19, 23, 25-26, 32, 37, 44, 47] reported SpO2. The pooled results showed that laryngeal mask ventilation increased pH [SMD=3.74, 95% CI (0.68, 6.80), P < 0.001, I2 = 97.4%], PaO2 [SMD=2.93 95% CI (1.43, 4.42), P < 0.001, I2 = 91.4%] and SpO2 [SMD=1.07, 95% CI (0.11,2.04), P < 0.001, I2 = 95.8%] and lowered PaCO2 [SMD=-4.03, 95% CI (-5.32, -2.73), P < 0.001, I2 =83.1%] (Figure 8).
Due to the existence of heterogeneity, we performed subgroup analysis on the initial success rates of the first insertion and insertion time. As shown in Table 3, we failed to identify any source of heterogeneity in insertion times across the human studies.
Sensitivity analysis and publication bias
The results of sensitivity analysis and publication bias are shown in Table 4. Funnel plots of manikin and human studies are shown in Figures 9-10, respectively. Sensitive analysis suggested that all the results were stable. Egger's tests showed the presence of publication bias in the ventilation efficiency rates and insertion times of human studies, but metatrim suggested that the bias had no influence on the final results.