Helmet noninvasive ventilation compared to facemask noninvasive ventilation and high-flow nasal cannula in acute respiratory failure: a systematic review and meta-analysis
- Dipayan Chaudhuri1,2,13,
- Rehman Jinah1,13,
- Karen E.A. Burns2,3,4,
- Federico Angriman3,5,6,
- Bruno L. Ferreyro3,6,7,
- Laveena Munshi3,7,
- Ewan Goligher3,8,9,
- Damon Scales3,5,6,
- Deborah J. Cook1,2,10,
- Tommaso Mauri11,12 and
- Bram Rochwerg1,2⇑
- 1Dept of Medicine, Division of Critical Care, McMaster University, Hamilton, ON, Canada
- 2Dept of Health Research Methods, Evidence, and Impact, McMaster University, Hamilton, ON, Canada
- 3Interdepartmental Division of Critical Care Medicine, University of Toronto, Toronto, ON, Canada
- 4Li Ka Shing Knowledge Institute, Unity Health Toronto – St Michael's Hospital, Toronto, ON, Canada
- 5Dept of Critical Care Medicine, Sunnybrook Health Sciences Centre, Toronto, ON, Canada
- 6Institute of Health Policy, Management and Evaluation, Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada
- 7Dept of Medicine, Sinai Health System and University Health Network, Toronto, ON, Canada
- 8Dept of Medicine, Division of Respirology, University Health Network, Toronto, ON, Canada
- 9Toronto General Hospital Research Institute, Toronto, ON, Canada
- 10Division of Critical Care Medicine, St Joseph's Hospital, Hamilton, ON, Canada
- 11Dept of Pathophysiology and Transplantation, University of Milan, Milan, Italy
- 12Dept of Anaesthesia, Critical Care and Emergency, Fondazione IRCCS Ca Granda Ospedale Maggiore Policlinico, Milan, Italy
- 13These authors share first authorship
- Corresponding author: Bram Rochwerg (rochwerg{at}mcmaster.ca)
Abstract
Background Although small randomised controlled trials (RCTs) and observational studies have examined helmet noninvasive ventilation (NIV), uncertainty remains regarding its role. We conducted a systematic review and meta-analysis to examine the effect of helmet NIV compared to facemask NIV or high-flow nasal cannula (HFNC) in acute respiratory failure.
Methods We searched multiple databases to identify RCTs and observational studies reporting on at least one of mortality, intubation, intensive care unit (ICU) length of stay, NIV duration, complications or comfort with NIV therapy. We assessed study risk of bias using the Cochrane Risk of Bias 2 tool for RCTs and the Ottawa–Newcastle Scale for observational studies, and rated certainty of pooled evidence using the GRADE (Grading of Recommendations, Assessment, Development and Evaluation) framework.
Results We separately pooled data from 16 RCTs (n=949) and eight observational studies (n=396). Compared to facemask NIV, based on low certainty of evidence, helmet NIV may reduce mortality (relative risk 0.56, 95% CI 0.33–0.95) and intubation (relative risk 0.35, 95% CI 0.22–0.56) in both hypoxic and hypercapnic respiratory failure, but may have no effect on duration of NIV. There was an uncertain effect of helmet NIV on ICU length of stay and development of pressure sores. Data from observational studies were consistent with the foregoing findings but of lower certainty. Based on low and very low certainty data, helmet NIV may reduce intubation compared to HFNC, but its effect on mortality is uncertain.
Conclusions Compared to facemask NIV, helmet NIV may reduce mortality and intubation; however, the effect of helmet NIV compared to HFNC remains uncertain.
Abstract
Helmet NIV may reduce mortality and intubation when compared to facemask NIV; however, large, well-designed RCTs are needed on this topic https://bit.ly/3i1rCnS
Introduction
The European Respiratory Society/American Thoracic Society clinical practice guidelines strongly recommend noninvasive ventilation (NIV) use for patients who have acute respiratory failure (ARF) due to cardiogenic pulmonary oedema and exacerbations of chronic obstructive pulmonary disease (COPD), and conditionally recommend its use for patients with ARF due to other causes, including trauma, post-operative respiratory failure and those with immunocompromise [1]. For patients with ARF, NIV is typically applied with a facemask interface [2]. However, at higher airway pressures, the facemask interface may be difficult to tolerate and associated with air leaks, thus impairing oxygenation and limiting the mean airway pressure that can be applied to maintain lung recruitment [3]. Additionally, patients may not tolerate the facemask mask due to claustrophobia or facial pressure ulceration [4].
The helmet interface is a relatively new interface for NIV delivery. A transparent hood is positioned over the patient's head with a seal at the neck using a soft collar. The helmet reduces air leak due to better seal integrity at the neck and improves tolerability because there is no direct contact with the patient's face [5]. In patients with potentially infectious respiratory illness such as coronavirus disease 2019 (COVID-19), the reduced air leak and attendant decrease in droplet dispersion is especially valuable [6]. Furthermore, when compared to the facemask interface or high-flow nasal cannula (HFNC), the helmet reduces inspiratory effort, preserves lung volumes and allows for lower inspiratory support, possibly by mitigating air leak or allowing for more effective provision of positive end-expiratory pressure (PEEP) [3, 7, 8]. A recent JAMA Network meta-analysis comparing all noninvasive oxygenation strategies in patients with purely hypoxaemic respiratory failure demonstrated that helmet NIV may lower mortality and the need for intubation compared to conventional oxygen therapy [9]. However, only a small number of randomised controlled trials (RCTs) were included in the JAMA Network review [3, 5, 10–12] and it did not evaluate other patient-important outcomes such as complications, comfort or duration of NIV. Moreover, with a focus on only hypoxaemic respiratory failure, the effect of helmet NIV on the other forms of ARF remained uncertain. The COVID-19 pandemic has increased helmet NIV use [13]; however, uncertainty regarding the benefits and harms of helmet NIV in clinical practice remains. Given several recently published RCTs and observational studies evaluating helmet NIV, along with the shortfalls of the previous JAMA Network systematic review addressing the topic [9], we conducted a systematic review and meta-analysis to address the following research question: “In adult patients with ARF of all types, does use of helmet NIV reduce mortality, intubation rate, intensive care unit (ICU) length of stay and the risk of complications compared to facemask NIV or HFNC?”.
Methods
We registered the protocol of this systematic review with PROSPERO (CRD42020222942) and report our findings using the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist (supplementary table E1).
Search strategy and selection criteria
We performed a comprehensive search of the following databases from inception until 23 October 2020: MEDLINE, Embase, Web of Science, The Cochrane Library, International HTA Database, EBSCO CINAHL Complete, LILACS and WHO COVID-19 Global literature on coronavirus disease. The search was updated on 31 March 2021. We used keywords “noninvasive ventilation” or “oxygen inhalation therapy” or “oxygen therapy” or “respiratory insufficiency” or “respiratory insufficiency” or “adult respiratory distress syndrome” or “respiratory failure” or “acute respiratory failure” or “adult respiratory distress syndrome” or “continuous positive airway pressure” or “positive end-expiratory pressure” and “head protective devices” or “helmet”. We did not exclude trials based on language or quality. We searched the bibliographies of included articles and prior meta-analyses on the topic. We consulted experts in the field to identify unpublished studies. A copy of our initial search strategy is included in the supplementary material.
Study selection
Two reviewers (D.C. and R.J.) screened citations independently and in duplicate in two stages: first examining the title and abstracts, and then the full text of selected citations. We captured reasons for study exclusion after reviewing the full texts of identified trials. A third reviewer (B.R.) adjudicated disagreements.
We included parallel group and crossover RCTs and observational studies that had an intervention and comparator cohort. We included studies that compared helmet NIV to NIV through another interface or HFNC in adult patients with ARF of any aetiology. Included studies had to report at least one of the following outcomes of interest: mortality, intubation rate, duration of mechanical ventilation, ICU length of stay, hospital length of stay, patient comfort, modality tolerance and NIV-related adverse events. We excluded observational studies without comparative analysis as well as case studies and case reports.
Data extraction and quality assessment
Two independent reviewers (D.C. and R.J.) working in pairs abstracted data in duplicate using a standardised data abstraction form. We collected data on trial characteristics, demographic data, interventional and control details, and outcomes. A third reviewer (B.R.) adjudicated disagreements where needed.
We assessed risk of bias in duplicate using the modified Cochrane Risk of Bias 2 tool for RCTs [14]. We assessed each RCT using the following domains: randomisation sequence generation, allocation concealment, blinding, incomplete data, selective reporting and other bias. For each domain, we rated risk of bias to be “low”, “high” or “some concerns”. The overall risk of bias for each trial was the highest risk attributed to any domain except for blinding (of the caregiver and patient specifically), as blinding is not feasible even with sham devices for these trials. For observational studies, we used the Newcastle–Ottawa Scale [15] and assessed each cohort or case–control study using the following domains: selection, comparability and exposure/outcome. For each domain, we rated the risk of bias by a star system, whereby the greater the number of stars, the lower the risk of bias. We assessed overall certainty of evidence for each outcome using the GRADE (Grading of Recommendations, Assessment, Development and Evaluation) framework [16]. To assess for publication bias, we also created funnel plots for the outcomes of mortality and intubation.
Data analysis
We pooled RCTs and observational studies separately. In keeping with the GRADE methodology, when presenting pooled data from both RCTs and observational data, we focused on the results with the higher certainty. We used the DerSimonnian–Laird random effects model with inverse variance weighting to generate pooled treatment effects across studies. We assessed heterogeneity between trials using a combination of the Chi-squared test, the I2 statistic and visual inspection of the forest plots [17]. We present results of dichotomous outcomes using relative risk (risk ratio) and continuous outcomes as mean difference (MD) with 95% confidence intervals. We also tabulated absolute differences with 95% confidence intervals. We performed all statistical analysis using RevMan 5.3 (Cochrane Collaboration, London, UK) software.
We planned for five a priori subgroup analyses: 1) COPD/hypercapnic respiratory failure versus non-COPD/hypercapnic respiratory failure patients, 2) congestive heart failure (CHF)/pulmonary oedema versus non-CHF/pulmonary oedema patients, 3) COVID-19-related ARF versus non-COVID-19-related ARF patients, 4) immunocompromised versus non-immunocompromised patients and 5) high versus low risk of bias studies. A priori, we hypothesised that COPD patients, CHF patients, COVID-19 patients, immunocompromised patients and trials at high risk of bias would show greater benefit with helmet NIV therapy.
Results
Search strategy and study characteristics
We reviewed 974 citations, and included 16 RCTs (n=949) [3, 7, 18–31] and eight observational studies (n=396) [32–39] (figure 1). We present the characteristics of the included RCTs in table 1 and the observational studies in supplementary table E4. RCTs included between 10 and 188 patients. Of the 16 included RCTs, four were crossover studies [7, 18, 20, 31] and two were only published in abstract form [24, 25]. Overall, 13 studies compared helmet NIV to facemask NIV and three trials compared helmet NIV to HFNC [7, 23, 25]. Three trials applied the helmet NIV in continuous positive airway pressure (CPAP) mode [21, 24, 25] and 13 trials applied bilevel helmet NIV [3, 7, 18–20, 22, 23, 26–31].
Six trials included patients with hypoxic respiratory failure, of which one trial each focused on patients with ARDS [3], pulmonary oedema [24], chest trauma [22] and COVID-19 [23], and two on mixed hypoxaemic respiratory failure [7, 25]. Two trials examined patients with post-extubation respiratory failure [20, 22] and the eight remaining trials enrolled exclusively patients with hypercapnic respiratory failure/COPD [18, 19, 26–31]. We summarise the risk of bias for the included RCTs in supplementary table E2a and c. Six trials were adjudicated to have low or intermediate risk of bias [3, 7, 19, 22, 23, 27, 31], while the remainder were judged to be at high risk of bias.
Of the eight observational studies, four were case–control studies [33, 36, 38, 39] and four were cohort studies [32, 34, 35, 37]. Observational studies included between 20 and 99 patients. Three studies compared helmet NIV to HFNC [32, 35, 37] and five studies compared helmet NIV to facemask NIV. Four studies only used helmet CPAP as their intervention [32, 35, 37, 38] and four studies evaluated helmet NIV [33, 34, 36, 39]. Only one study examined patients with COPD [33], while the remaining seven studies examined helmet NIV in patients with hypoxic respiratory failure. Of the studies evaluating hypoxic patient populations, two focused on patients with COVID-19 infection [32, 35], one evaluated patients with haematological malignancies [38] and one assessed immunocompromised patients [39]. We summarise the risk of bias for the observational studies in supplementary table E2b. Most studies were adjudicated to have low risk of bias except for two studies [32, 35] that did not match their comparison cohorts.
Outcomes
We summarise the GRADE certainties and pooled estimates for pooled outcomes in supplementary table E3.
Helmet NIV versus facemask NIV
Compared to facemask NIV, helmet NIV may reduce mortality (relative risk 0.56, 95% CI 0.33–0.95; low certainty) (figure 2) and intubation (relative risk 0.35, 95% CI 0.22–0.56; low certainty) (figure 3). Observational data were consistent with these findings, yet of lower certainty (supplementary figures E1 and E2). Pooled data from RCTs suggested that helmet NIV has an uncertain effect on ICU length of stay (MD 0.29 days less, 95% CI 2.31 days less to 1.74 days more; very low certainty) (figure 4) and may have no effect on duration of NIV (MD 0.02 days less, 95% CI 0.15 days less to 0.11 days more; low certainty) (figure 5). Observational data were again consistent with these findings but of lower certainty (supplementary figures E4 and E5)
Helmet NIV has an uncertain effect on the risk of skin necrosis/pressure sores compared to facemask NIV (relative risk 0.50, 95% CI 0.19–1.37; absolute risk reduction 8.1% lower, 95% CI 13.2% lower to 6.0% higher; very low certainty) (supplementary figure E7). All other complications are summarised in table 2 as they were too variably reported to allow for pooling. The most common complications were skin necrosis/pressure sores and gastric distension. Similarly, whether and how patient comfort scales were documented across trials did not allow for statistical synthesis, so these are summarised in table 2.
Helmet NIV versus HFNC
Compared to HFNC, low certainty of evidence from RCTs suggests that helmet NIV may reduce intubation (relative risk 0.59, 95% CI 0.39–0.91) (figure 6), but has an uncertain effect on mortality (relative risk 0.72, 95% CI 0.40–1.28; very low certainty) (figure 7).
The pooled estimates from observational studies for both intubation (relative risk 0.69, 95% CI 0.27–1.73) (supplementary figure E5) and mortality (relative risk 0.77, 95% CI 0.16–3.75) (supplementary figure E6) are consistent in demonstrating uncertainty based on very low certainty of evidence.
Subgroup and sensitivity analysis
For the outcome of intubation, we did not identify credible subgroup effects when comparing patients with hypercapnic respiratory failure to those with hypoxaemic respiratory failure or when comparing high versus low or intermediate risk of bias trials in pooled analysis from either RCTs or observational studies (figure 3, and supplementary figures E2 and E8). For the outcome of intubation, we also did not identify any credible subgroup effects when comparing high versus low or intermediate risk of bias trials (supplementary figure E11). The remaining pre-planned subgroup analyses were not feasible due to lack of study-level aggregate data (only one study included immunocompromised patients and two included patients with COVID-19).
Publication bias
There was minimal publication bias for the comparison of helmet NIV to facemask NIV in terms of the outcomes of mortality and intubation (supplementary figures E9 and E10). We did not perform funnel plots for the comparison of helmet NIV to HFNC due to the small number of included studies.
Discussion
Although the use of helmet NIV has steadily increased [13], the evidence supporting its use remains sparse. This systematic review and meta-analysis found that while available studies demonstrate that helmet NIV may be associated with lower intubation rates and mortality compared to facemask NIV, the certainty of these estimates remains low. The effect of helmet NIV on other clinically important outcomes, including ICU length of stay, duration of NIV and adverse events such as facial ulceration, is uncertain. There was limited evidence to compare helmet NIV to HFNC and therefore we conclude that high-quality RCTs are required to establish the net clinical benefits or harms of helmet NIV.
Compared to previous reviews, this systematic review and meta-analysis adds a number of new studies examining the role of helmet NIV in ARF [40] (12 new studies including seven new RCTs [7, 21, 22, 24, 25, 30, 31]). Despite this, all included trials and observational studies were small. For example, the largest trial examining helmet NIV use was an RCT of 188 patient that compared helmet NIV to HFNC [25]. Furthermore, two included trials were only published in abstract form [24, 25], and two trials were of a crossover design and only examined short-term outcomes [7, 31]. Although pooled data from this systematic review suggest that helmet NIV may be preferable to facemask NIV, the information size and event rates are low, contributing to important imprecision which limits the strength of inferences that can be made. Comparisons between the effects of helmet NIV versus HFNC are even more uncertain. Overall, this systematic review highlights the critical need for large, high-quality RCTs comparing helmet NIV to both facemask NIV and HFNC, including patient-important outcomes and attention to possible adverse events.
Many questions regarding the net clinical benefits of helmet NIV remain. Although some trials and studies reported complications and patient-reported comfort with helmet NIV, we were unable to pool the majority of data on these end-points due to infrequent and variable outcome reporting. Similarly, while current best trial evidence supports the use of facemask NIV in selected populations (patients with COPD, CHF, immunocompromised, etc.) [1], there is currently a relative dearth of evidence regarding the effects of helmet NIV in these patient populations. Specifically in patients with hypercapnic respiratory failure, worsening hypercapnia, ventilator asynchrony and under-assistance are common concerns [33, 41]. However, at least one study of helmet NIV has shown that adequate carbon dioxide clearance can be achieved with high gas flow rates [41] and a few others have shown that helmet NIV reduces inspiratory effort [7, 8]. Regardless, to address the aforementioned concerns, we compared patients with hypercapnic respiratory failure versus those with hypoxaemic ARF in a pre-specified subgroup analysis. Although we did not find any credible subgroup effects based on available data, imprecision and low number of events underscore the need for further investigation.
The ability to provide a better seal compared to a facemask mask and not obscure a full facial view also provides the helmet with a few unique applications. For pandemic-related illnesses, such as COVID-19 and severe acute respiratory syndrome, the helmet may be a safer route to provide noninvasive respiratory support. To this end, simulation studies have demonstrated benefits of the helmet interface when compared to other noninvasive modes of respiratory support in the context of exhaled viral dispersion [6, 42], although this aerosolisation has not been rigorously evaluated in patients. For patients with ARF who are post-extubation, HFNC can be concurrently applied with helmet NIV and other nasal respiratory support devices. Moreover, helmet NIV permits a full facial view, speaking and nasogastric feeding tubes, which is often not possible with facemask NIV. Whether these features translate into enhanced comfort, fewer cutaneous complications and other benefits remains unknown, as patient-reported outcomes are lacking in this field. In addition, both CPAP and pressure support ventilation modes have been used with helmet NIV for various causes of respiratory failure. While it is likely that certain modes will provide no benefit for certain conditions (e.g. CPAP for COPD), the ideal mode for each cause of respiratory failure remains unknown. Finally, the cost-effectiveness of this new technology has not been examined. Although the helmet interface costs more than the traditional facemask interface, a previous costing study based on the RCT by Patel et al. [3] suggested that by reducing intubation and ICU length of stay, the helmet interface may actually be associated with cost savings; however, further clinical studies and a more comprehensive cost-effectiveness study is needed to confirm or refute these findings.
To the best of our knowledge, this is the largest and most comprehensive systematic review and meta-analysis to assess helmet NIV compared to facemask NIV and HFNC. Strengths of this study include pre-registration, incorporation of a comprehensive search, assessment of GRADE certainty allowing for appropriate contextualisation of results and inclusion of 12 additional studies (including seven RCTs) compared to a previously conducted review including 13 studies [40]. This review also has limitations. First, the total number of included patients and the number of events are small. Second, by including all studies that compared helmet NIV to either HFNC or facemask NIV, there was considerable clinical and methodological heterogeneity across trials, which nonetheless was not associated with statistical heterogeneity (inconsistency) for most outcomes. Acknowledging different design features informing this review, we analysed studies that compared helmet NIV to facemask NIV and HFNC separately, and RCTs and observational studies separately. However, considerable clinical heterogeneity remained as we were unable to conduct most predefined subgroup analyses due to insufficient data. In particular, we were unable to separate studies that examined hypoxic respiratory failure by the underlying varying pathophysiological mechanisms. While this highlights the need for further study on how specific causes of ARF respond to helmet NIV, the lack of inconsistency across our outcomes of interest seems to suggest that the effect of helmet NIV is likely similar regardless of the cause of ARF.
Conclusions
Compared to facemask NIV, helmet NIV may reduce mortality and intubation; however, the effect of helmet compared to HFNC remains uncertain. As application of this technology increases, large, well-designed RCTs comparing helmet NIV to both facemask NIV and HFNC in patients with both hypoxaemic and hypercapnic respiratory failure will be needed to help inform practice.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary tables and search strategy information ERJ-01269-2021.Supplement
Supplementary figures ERJ-01269-2021.Figures
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-01269-2021.Shareable
Acknowledgements
The authors wish to thank Library Services, St Joseph's Healthcare Hamilton (Hamilton, ON, Canada) for their assistance with conducting literature searches.
Footnotes
The protocol for this systematic review is registered with PROSPERO (CRD42020222942)
Conflict of interest: D. Chaudhuri has nothing to disclose.
Conflict of interest: R. Jinah has nothing to disclose.
Conflict of interest: K.E.A. Burns has nothing to disclose.
Conflict of interest: F. Angriman has nothing to disclose.
Conflict of interest: B.L. Ferreyro is supported by a Vanier Canada Graduate Scholarship.
Conflict of interest: L. Munshi has nothing to disclose.
Conflict of interest: E. Goligher has nothing to disclose.
Conflict of interest: D. Scales has nothing to disclose.
Conflict of interest: D.J. Cook has nothing to disclose.
Conflict of interest: T. Mauri received personal fees from Drager, Fisher & Paykel and BBraun, all outside of the submitted work.
Conflict of interest: B. Rochwerg has nothing to disclose.
- Received May 4, 2021.
- Accepted July 17, 2021.
- Copyright ©The authors 2022. For reproduction rights and permissions contact permissions{at}ersnet.org