Supraglottic Airways Compared With Face Masks for Neonatal Resuscitation: A Systematic Review

This SR and meta-analysis was completed as part of the ILCOR NLS Task Force continuous evidence review process based on knowledge gaps identified in the 2015 ILCOR NLS guidelines.³⁰  The SR and meta-analysis were guided by the Cochrane Handbook for Systematic Reviews of Interventions³¹  and reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement for meta-analysis of health care interventions.³²  The protocol was registered before article selection with the Prospective Register of Systematic Reviews (PROSPERO) [identifier CRD42021230722.]

Randomized controlled trials (RCTs), quasi-randomized controlled trials (quasi-RCTs), and nonrandomized studies (nonrandomized controlled trials, interrupted time series, controlled before-and-after studies, and cohort studies) were eligible forinclusion. Unpublished studies (eg, conference abstracts, trial protocols), review articles, editorials, comments, case reports, animal studies, and manikin studies were excluded. All years were included without language restrictions if an English abstract was available.

The Population, Intervention, Comparator, Outcome question was developed by the ILCOR NLS Task Force and approved by the ILCOR Scientific Advisory Committee.

• Population: newborn infants ≥34 ^0/7 weeks’ gestation receiving intermittent PPV during resuscitation immediately after birth.

• Intervention: supraglottic airway.

• Comparator: face mask.

• Outcomes: the primary outcome was failure to improve with the assigned device as defined by the study authors. Secondary outcomes included intubation during initial resuscitation, time to heart rate (HR) >100 beats per minute from the start of PPV during initial resuscitation, duration of PPV during initial resuscitation or time to cessation of PPV, chest compressions or epinephrine (adrenaline) during initial resuscitation, soft tissue injury as defined by authors, admission to a NICU, air leak (pneumothorax, pneumomediastinum, pulmonary interstitial emphysema, or pneumopericardium) diagnosed during the initial hospital stay, survival to hospital discharge, and neurodevelopmental impairment (abnormal motor, sensory or cognitive function, or low educational achievement using an appropriate, standardized test or examination) at ≥18 months.

The choice of “failure to improve with the assigned device as defined by the study authors” was a pragmatic primary outcome to assess whether primary use of SA to provide positive pressure ventilation leads to improving condition of neonates undergoing resuscitation after birth. The decision to use the criteria “as defined by the study authors” was partially informed by an earlier ILCOR systematic review that showed it was a consistently reported outcome, and any more specific review team definition could lead to excluding relevant studies or dividing studies into multiple small groups. As most investigators were following Neonatal Life Support or Helping Babies Breathe resuscitation guidelines, the practical definition of “failure to improve” was consistent across studies (Table 1).

Outcome ratings using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) classifications of critical or important were based on a consensus for international neonatal resuscitation guidelines (range 1 to 3 low importance for decision-making, 4 to 6 important but not critical for decision-making, 7 to 9 critical for decision-making).^33,34  Potential subgroups (late preterm and SAs without an inflatable cuff) were defined a priori. If necessary, study authors were contacted to request missing data. Missing standard deviations (SD) were imputed or estimated following guidelines outlined in the Cochrane Handbook.³⁵ 

Literature searches in Medline, Embase, the Cochrane Database of Systematic Reviews, the Cochrane Central Register of Controlled Trials, the Cochrane Methodology Register, the Database of Abstracts of Reviews of Effects, and Cumulative Index to Nursing and Allied Health Literature were conducted by an information specialist (H.N.W.) from Stanford University’s Lane Medical Library in close consultation with the review team. The initial search was completed on November 13, 2020, and updated on July 28, 2021 and December 9, 2021 (see Supplemental Information). The subject headings and keywords were adapted for the respective databases. Registries were searched for ongoing clinical trials or unpublished work through July 28, 2021, including the International Clinical Trials Registry Platform (who.int), United States Clinical Trials Registry (ClinicalTrials.gov), Cochrane Central Register of Controlled Trials (cochranelibrary.com), European Union Clinical Trials Register (clinicaltrialsregister.eu), and Australian New Zealand Clinical Trials Registry (anzctr.org.au). Covidence³⁶  (Covidence Systematic Review Software, Veritas Health Innovation, 2020) was used for management of the search results and screening of the studies.

Two reviewers (N.K.Y. and B.H.Q.) independently screened titles and abstracts for potential eligibility. In the event of a disagreement at abstract screening, the full text was reviewed. Both reviewers independently evaluated the full text of potentially eligible studies for inclusion. Disagreements were resolved by consensus of the full review team. The first reason for exclusion was captured according to a predetermined, ordered list of exclusions (manikin or animal studies, SA used as rescue airway, SA used for continuous positive airway pressure, SA compared with intubation, resuscitation not immediately after birth, or device used in other clinical environments [eg, nursery, emergency department, operating room]). Inter-rater agreement for article selection was assessed using Cohen’s κ coefficient at the abstract and full-text stages using Covidence³⁶  software.

Data extraction was completed for each included study by 2 authors independently using a standardized form. Risk of bias (RoB) was assessed for each outcome by 2 team members using the Cochrane RoB 2 tool³⁷  for randomized trials (C.J.D.M. and G.M.W.) and the Risk of Bias in Non-Randomized Studies of Interventions (ROBINS-I) tool³⁸  for observational studies (N.K.Y. and B.H.Q.). Review team members were excluded from assessing inclusion or RoB for any study in which they had participated as an investigator. RoB for each outcome was summarized across studies. Two authors (C.J.D.M. and G.M.W.) independently assessed the certainty of evidence (CoE), or confidence in the estimate of effect, for each outcome using the GRADE framework.³⁹  The optimal information size (OIS) was used to assess imprecision. Imprecision was considered present if the total number of study participants included was less than OIS for the outcome under consideration.⁴⁰  The RoB and GRADE assessments were evaluated by the full review team to reach consensus and ensure consistency. The evidence profile tables were presented and discussed with the ILCOR NLS Task Force and content experts.

GRADEpro⁴¹  (GRADEpro Guideline Development Tool, McMaster University and Evidence Prime, 2021) and Review Manager⁴²  (RevMan version 5.4.1. The Cochrane Collaboration, 2020) were used to summarize and analyze the data. Meta-analyses were performed if ≥2 RCTs or quasi-RCTs were available. Results from RCTs and quasi-RCTs were analyzed together. Observational studies were analyzed and reported if fewer than 2 RCTs or quasi-RCTs provided outcome data or if the CoE from these RCTs or quasi-RCTs was very low. If the study permitted crossover between devices, outcomes were assigned to the device assigned at randomization using the intent- to-treat principle. For dichotomous outcomes, pooled unadjusted risk ratios (RRs) and corresponding 95% confidence intervals (CIs) were reported using the Mantel-Haenszel fixed effect method. The pooled risk difference (RD) and number needed to treat (NNT) were calculated when the RR revealed a statistically significant difference. Pooled continuous variables were reported as weighted mean differences (MDs) and corresponding 95% CIs using the Mantel-Haenszel fixed effect method. Forest plots were used for graphical representation of RRs and MDs. Heterogeneity was measured using the I² statistic.^31,43  Significant heterogeneity was considered present if the I² statistic was >50%. We explored statistical heterogeneity using posthoc sensitivity analyses. Subgroup analyses were planned according to (1) gestational age (term versus late preterm infants) and (2) SA design (inflatable cuff versus no inflatable cuff).

The search strategy identified 1849 records (Fig 1). One additional study⁴⁴  was identified by the authors through hand-searching of citations. One study⁴⁵  reported additional secondary outcomes from a subset of patients enrolled in an included RCT.⁴⁶  After removing 442 duplicates, 1407 titles and abstracts were screened, and 1387 studies were excluded (Fig 1). A total of 21 full-text articles were assessed for eligibility and 9 were included in the final review. Inter-rater agreement was moderate (κ .53) at the title and abstract stage and complete (κ 1.0) at the full-text stage.

The SR included 5 RCTs,^44,46–49  1 quasi-RCT,⁵⁰  and 2 retrospective cohort studies^51,52  enrolling a total of 2075 newborn infants (1857 in RCTs and quasi-RCTs, 218 in cohort studies) (Table 1). All except 1 study⁴⁴  required enrolled newborn infants to be at least 34- or 35-weeks’ gestational age, and all studies required an estimated birth weight greater than 1.5 or 2 kg. All studies used a size-1 SA. Two RCTs^46,49  used an SA without an inflatable cuff.

RoB was a concern for all studies because the nature of the intervention prevented the personnel responsible for administering PPV in the delivery room from being blinded to group assignment (Fig 2). In addition, many outcomes were at increased risk of ascertainment bias because outcome assessors were aware of the group assignment.

Most outcomes selected for GRADE analysis were rated as low or very low certainty of evidence (CoE) because of high RoB and imprecision (Fig 2). Although the CoE for the outcome “failure to improve with the device” was also downgraded for RoB and imprecision, it was upgraded to moderate CoE by the GRADE method because the pooled effect demonstrated a strong association.

The primary outcome, failure to improve with the device, was reported in all 6 RCTs and quasi-RCTs with outcome data available for 1823 newborn infants (Fig 2).^44,46–50  Meta-analysis showed that using an SA compared with a face mask for PPV immediately after birth decreased the probability of failure to improve with the assigned device (RR 0.24; 95% CI 0.17 to 0.36; P <.001; I² = 35%; RD −11%, 95% CI −13% to −8%; NNT = 10; moderate CoE) (Fig 3).

Endotracheal intubation during initial resuscitation was reported in 4 RCTs and quasi-RCTs.^46–48,50  Meta-analysis revealed that PPV delivered through an SA compared with face mask decreased the probability of intubation (RR 0.34, 95% CI 0.20 to 0.56; P <.001; I² = 78%; RD −5%, 95% CI −6% to −3%; NNT = 20; low CoE) (Figs 2 and 4). In sensitivity analysis, the heterogeneity associated with this outcome was not decreased, and the benefit remained (RR 0.19, 95% CI 0.09 to 0.37; P <.001; I² = 63%), after removing the study⁴⁶  where personnel capable of intubating were not always available. When the single quasi-RCT⁵⁰  was removed, heterogeneity was decreased but the RR was no longer statistically significant (RR 0.65, 95% CI 0.36 to 1.19; P = .17; I² = 45%).

The duration of PPV during the initial resuscitation was reported in 4 RCTs and quasi-RCTs.^47–50  Meta-analysis revealed a shorter duration of PPV using an SA (MD −18 seconds; 95% CI −24 seconds to −13 seconds; P <.001, I² = 94%; low CoE) (Fig 2). In sensitivity analysis, the heterogeneity for this outcome was attributed to 1 study.⁴⁸  When this study was removed, the beneficial effect was retained and statistical heterogeneity was fully resolved (MD −30 seconds, 95% CI −36 seconds to −24 seconds; P <.001, I² = 0%). Data reported separately from a subset of patients enrolled in 1 included trial showed that using an SA compared with face mask reduced the time to achieve HR >100 beats per minute from the start of PPV (median 13 seconds, interquartile range 9–15 seconds versus median 61 seconds, interquartile range 33 to 140 seconds).⁴⁵ 

Meta-analysis revealed no difference between groups for the probability of receiving chest compressions^46–48  or epinephrine (adrenaline)^47,48  during resuscitation (Fig 2). The RR of air leaks could not be estimated from the RCTs, because no events occurred in either group in the 2 studies reporting this outcome.^47,48  Meta-analysis of the 2 cohort studies revealed no difference between groups in the probability of air leaks between groups.^51,52  No difference in the incidence of soft tissue injury was found in meta-analysis of 4 RCTs and quasi-RCTs^46–48,50 ; however, injuries (2 events in each group) occurred in only 1 study.⁴⁶ 

No difference was found in the probability of admission to the NICU,^46–49  but meta-analysis revealed significant heterogeneity (Fig 2). In sensitivity analysis, the heterogeneity was attributed to the high NICU admission rate (96% in both groups) in 1 study.⁴⁶  When this study was removed, the treatment effect favored the SA and statistical heterogeneity fully resolved (RR 0.60, 95% CI 0.40 to 0.90, P = .01, I² = 0%). Neither the single RCT⁴⁷  nor meta-analysis of the 2 cohort studies^51,52  reporting survival to hospital discharge showed a difference between groups. No data were reported for neurodevelopmental impairment at ≥18 months of age.

No data were reported for prespecified subgroup analyses by gestational age (term versus late preterm). For the primary outcome of “failure to improve with the device,” analysis based on device design (inflatable cuff versus no inflatable cuff) found no evidence of an interaction (P = .29, I² = 10%).

Ventilation of the newborn infant’s lungs is the most important step in neonatal resuscitation and identifying the most effective interface for PPV is a research priority.^4,5  This systematic review and meta-analysis found that SAs compared with face masks may be more effective in achieving successful resuscitation of late preterm and term newborn infants who receive PPV immediately after birth. The analysis demonstrated that use of an SA decreased the probability of failing to improve with PPV and endotracheal intubation. In addition, there was a reduction in the time required to achieve HR >100 beats per minute and the duration of PPV. There was no significant difference in other important and critical outcomes, including the probability of receiving chest compressions or epinephrine (adrenaline) during the initial resuscitation, admission to the NICU, air leak, soft tissue injury, admission to the NICU, or survival to hospital discharge.

During neonatal resuscitation, inflating the lungs clears liquid, establishes functional residual capacity, increases pulmonary blood flow, and increases and maintains heart rate. Despite the importance of effective PPV, achieving the tidal volume required for gas exchange with a face mask can be difficult. Using a face mask, previous investigators have shown that clinicians have difficulty delivering consistent tidal volumes or inflation pressures in manikin models and achieving chest expansion during actual resuscitations.^8,53  Leak around the mask and obstructions to gas flow limit the effectiveness of face mask PPV. In a simulation study involving experienced clinicians providing face mask PPV with a term manikin, the median leak was 71% and obstruction was observed in 46% of inflations.⁵⁴  Among birth attendants in a resource-limited setting, mask leak was nearly 90%.⁵⁵  Because the SA bypasses upper airway structures and makes a seal around the glottis, administering PPV through an SA may address these limitations, improve the delivered tidal volume, and decrease the chance of PPV failure. By delivering gas flow directly to the glottis, PPV with an SA may decrease dead space ventilation, increase volume delivery to the lower airway, and more rapidly establish a functional residual capacity.

Although this review did not find an interaction between SA design and failure to improve, laryngeal seal may be an important determinant of SA performance and could be affected by device size, shape, and whether an inflatable cuff is used. A simulation study using a neonatal manikin model compared the leak around 7 SA devices with a round and a triangular face mask.⁵⁶  Only the noninflatable cuff SA used in 2 included RCTs showed significantly less leak (5.7%) than face masks (35.7% to 42.7%).^46,49  Because the SA makes a seal against the tissue surrounding the glottis, the reliability of leak measurements in a manikin model are limited by the fidelity of the manikin’s airway anatomy and the materials simulating the upper airway tissue and; therefore, may not accurately represent leak when used in vivo.⁵⁷  In a subset of patients enrolled in one included RCT using a noninflatable cuff SA, there was no difference in the median leak or tidal volume achieved with the noninflatable cuff SA compared with face mask; however, there was a reduction in the proportion of breaths with a very low tidal volume (<2.5 mL/kg) when using the SA and a shorter time to heart rate recovery.⁴⁵  SAs without an inflatable cuff may be less likely to cause soft tissue injury, although this review suggests that soft tissue injury is uncommon regardless of the type of SA.

Our findings are similar to those in a previous Cochrane Systematic Review.⁵⁸  The present SR extends the previous analysis by including final published results from one RCT⁴⁹  and adding the largest RCT published to date.⁴⁶  As a result, the present review includes outcomes from an additional 1163 newborn infants enrolled in RCTs. The previous review also found that use of an SA compared with face mask decreased the probability of failure to improve with PPV and intubation. Although the direction of effect is the same, adding the large recent RCT⁴⁶  to the meta-analysis has slightly decreased the magnitude of effect. In addition, the previous Cochrane review reported that an SA decreased the probability of NICU admission. The updated analysis found no difference between groups but added significant heterogeneity. This appears to be related to a low threshold for NICU admission in both groups of the newly reported and largest RCT.⁴⁶  Although delivery room intubation is expected to increase the probability of NICU admission, it is likely that the threshold for NICU admission varies based on unit-specific criteria and will not be critical for decision making when choosing the interface used for PPV. Neither review found a difference in chest compressions or epinephrine (adrenaline) administration. Because chest compressions and epinephrine (adrenaline) are required infrequently when PPV is effective,^59,60  cross-over to the opposite device was allowed if PPV failed in 4 of the included RCTs,^44,46,47,49  and advanced resuscitation was not consistently available in the largest RCT,⁴⁶  the studies included in both reviews had limited ability to identify a difference in these outcomes.

A persistent gap in knowledge is the training required for successful SA insertion and the potential for skill decay. The results of the present SR suggest that successful SA insertion for neonatal resuscitation can be achieved by clinicians with varying backgrounds, including unsupervised midwives, with limited training. The included studies generally provided brief training with a manikin as a component of existing neonatal resuscitation courses. In some cases, additional training was provided in the form of video conference meetings, didactic videos, or video recordings of actual resuscitations with corrective feedback.⁴⁹  Only 2 RCTs required participants to demonstrate successful insertion in a clinical setting before study participation.^47,48  The largest RCT trained midwives by adding a module to the Helping Babies Breathe resuscitation curriculum and only required participants to demonstrate SA insertion in a manikin before study participation.⁴⁶  With this training, unsupervised midwives achieved successful resuscitation in 96.5% of newborn infants randomized to the SA group. With manikin training alone, Zhu reported that pediatricians successfully inserted the SA on the first attempt in 98.5% of enrolled newborn infants with a mean insertion time 7.8 seconds (range 4.5 seconds to 15 seconds).⁵⁰  The short insertion time was similar to that reported in the RCTs by Feroze⁴⁴  and Singh⁴⁷  who enrolled anesthesiology staff and residents. No studies have compared training methods, assessed whether training requirement vary by SA design, or evaluated whether SA skills decay after training. Additional research is needed to determine the amount and type of training that is needed to deliver optimal ventilation with an SA.

The strengths of this SR include a comprehensive literature search performed by an information specialist, use of standardized methods with assessments of RoB for each outcome, and review by an international group of neonatal resuscitation content experts. The main limitation of this review is the potential for performance and observer bias in the included studies caused by the inability to blind the personnel performing the interventions. For the quasi-RCT where allocation was clustered by day, the lack of blinding may have also introduced proficiency bias if practitioners with greater experience with the SA were more likely to attend births on days allocated to SA. These biases may have led to differential application of cointerventions as 4 of the included studies permitted use of the alternative device before proceeding to intubation. Although this may have decreased the effect size for benefit with the SA, it could also have masked potential harms. Although the nature of the intervention may have precluded blinding of personnel in the delivery room, many of the included studies did not employ strategies to reduce ascertainment bias, such as electronic data capture, video recordings, and review by blinded outcome assessors. Two included RCTs used video recordings to improve the accuracy of data collection in the delivery room; however, the research staff assessing these outcomes were not blinded to group allocation.^46,49  Finally, although the review includes outcomes for nearly 1800 newborn infants, there was the risk of type 2 error for detecting both benefits and harms as the OIS was reached for only one outcome. As a result, the CoE for most outcomes were downgraded for imprecision.

The results of this SR and meta-analysis suggest that in late preterm and term infants who require resuscitation after birth, ventilation may be more effective if delivered by SA rather than face mask and may reduce the need for endotracheal intubation. Moreover, the results suggest that the SA can be successfully inserted by a wide range of clinicians with brief training using manikins.

We thank Hong-nei Wong at Stanford University for her assistance with the development of the search strategy and conduct of the literature search.

The following ILCOR NLS Task Force members provided input on the review protocol, the interpretation of the results, and the article as experts in neonatal resuscitation: Dr Daniela T. Costa-Nobre, Federal University of São Paulo, São Paulo, Brazil; Dr Peter G. Davis, The Royal Women’s Hospital, Victoria, Australia; Dr Maria F. de Almeida, Federal University of São Paulo, São Paulo, Brazil; Dr Walid El Naggar, Dalhousie University, Halifax, Nova Scotia, Canada; Dr Jorge G. Fabres, Universidad Catolica de Chile, Santiago, Chile; Dr Joe Fawke, Leicester Royal Infirmary, Leicester, United Kingdom; Dr Elizabeth E. Foglia, University of Pennsylvania, Philadelphia, Pennsylvania; Dr Ruth Guinsburg, Federal University of São Paulo, São Paulo, Brazil; Dr Shigeharu Hosono, Jichi Medical University, Saitama, Japan; Dr Tetsuya Isayama, National Center for Child Health and Development, Tokyo, Japan; Dr Vishal S. Kapadia, University of Texas Southwestern Medical Center, Dallas, Texas; Dr Mandira D. Kawakami, Federal University of São Paulo, São Paulo, Brazil; Dr Henry C. Lee, Stanford University School of Medicine, Palo Alto, California; Dr Han-Suk Kim, College of Medicine, Seoul National University, Seoul, Korea; Dr R. John Madar, University Hospitals Plymouth NHS Trust, Plymouth, United Kingdom; Dr Firdose L. Nakwa, University of Witwatersrand, Johannesburg, South Africa; Dr Jeffrey M. Perlman, Weill Cornell Medical College, Cornell University, New York, New York; Dr Charles C. Roehr, Oxford University Hospitals, National Health Service Foundation Trust, United Kingdom; Dr Mario Rüdiger, University Hospital Carl Gustav Carus, Dresden, Germany; Dr Takahiro Sugiura, Toyohashi Municipal Hospital, Toyohashi, Aichi, Japan; Dr Daniele Trevisanuto, University of Padua, Padua, Italy; Dr Jonathan Wyllie, James Cook University Hospital, South Tees National Health Service Foundation Trust, Middlesbrough, United Kingdom.

CI

confidence interval

CoE

certainty of evidence

GRADE

Grading of Recommendations, Assessment, Development and Evaluation

ILCOR

International Liaison Committee on Resuscitation

MD

mean difference

NLS

neonatal life support

NNT

number needed to treat

PPV

positive pressure ventilation

Quasi-RCT

quasi-randomized controlled trial

RCT

randomized controlled trial

RD

risk difference

RoB

risk of bias

RR

risk ratio

SA

supraglottic airway device

SR

systematic review