Scolaris Content Display Scolaris Content Display

Non‐invasive high‐frequency ventilation in newborn infants with respiratory distress

Esta versión no es la más reciente

Contraer todo Desplegar todo

Abstract

This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:

To determine effects of non‐invasive high‐frequency ventilation compared with other forms of respiratory support in newborn infants with respiratory distress in terms of need for intubation or reintubation and morbidity and mortality in term and preterm newborn infants.

We will examine use of nHFV in the following situations.

  • Initial respiratory support compared with other forms of invasive respiratory support (intubation and any ventilation mode).

  • Initial respiratory support compared with other forms of non‐invasive respiratory support (CPAP, nIPPV, HFNC) or air/oxygen.

  • An alternative to invasive endotracheal intubation for ventilation following failure of other types of non‐invasive respiratory support (CPAP, nIPPV, HFNC) in newborn infants with respiratory distress.

  • An alternative to other forms of non‐invasive respiratory support (nCPAP, nIPPV, HFNC) or air/oxygen following extubation.

We will include any interface used to deliver nHFV, including unilateral/bilateral and short/long nasal prongs, nasopharyngeal tube, face mask, and laryngeal mask airway.

Air/oxygen may involve low‐flow nasal prongs, a head box, or humidicrib oxygen delivery methods.

Background

Description of the condition

Respiratory distress

Respiratory distress occurs in 7% of newborn infants (Kumar 1996). Causes vary between preterm and term infants. Respiratory distress syndrome and infection are responsible for around half of cases in preterm infants, and infection, pulmonary hypoplasia, meconium aspiration syndrome (MAS), congenital heart disease, and diaphragmatic hernia contribute to half of all cases of respiratory distress in term infants (Kumar 1996). Respiratory distress syndrome is a disease that occurs predominantly in preterm infants and is associated with surfactant deficiency, dysfunction, or inactivation (Pfister 2009; Soll 2010). The term 'hyaline membrane disease' is used synonymously with 'respiratory distress syndrome' to describe respiratory distress that occurs in preterm infants (Stedman 2000). Transient tachypnoea of the newborn (TTN) is a common cause of respiratory distress in term infants, particularly after caesarean section.

Management of respiratory distress and its complications

Despite varied causes, the goals of managing respiratory distress include maintaining airway patency and providing respiratory support to deliver oxygen and remove carbon dioxide. In severe respiratory distress, these goals are often achieved through mechanical ventilation (Sarnaik 2011). Bronchopulmonary dysplasia (BPD) is one of the sequelae of mechanical ventilation of greatest concern. The term 'BPD' is used interchangeably with the term 'chronic lung disease (CLD)' (Jobe 2001). BPD is a chronic pulmonary condition caused by incomplete resolution or abnormal repair of lung injury during the neonatal period. A factor that contributes to BPD is that mechanical ventilation leads to volutrauma and barotrauma, causing fluid and protein transudation in the alveoli (Jobe 2001). Insufficiently opened lung areas may be damaged by shear forces that occur during the respiratory cycle through repetitive opening and closing of alveoli (atelectotrauma). These different traumas in turn stimulate the release of pro‐inflammatory cytokines and an inflammatory cascade causing biotrauma to the lungs. In addition, high inspired oxygen can cause oxidative stress and inflammation (Neumann 2014). Furthermore, the endotracheal (ET) tube used in mechanical ventilation causes trauma during introduction, leading to loss of defence mechanisms, including mucociliary clearance, and increasing risk for bacterial colonisation and respiratory infection (Aly 2008). Prolonged use of the ET tube can lead to subglottic stenosis and oedema, resulting in subsequent failure to extubate.

Description of the intervention

Non‐invasive ventilation techniques have the potential to minimise BPD caused by invasive endotracheal ventilation and have been reported to reduce BPD in some cases (DiBlasi 2011). Several methods of non‐invasive ventilation can be used, including nasal continuous positive airway pressure (nCPAP) (Ho 2002a; Ho 2002b; Rojas‐Reyes 2012; Subramaniam 2005); nasal intermittent positive‐pressure ventilation (nIPPV) (Davis 2001; Davis 2003); and high‐flow nasal cannula (HFNC) (Wilkinson 2011).

Non‐invasive high‐frequency ventilation (nHFV) is another non‐invasive ventilation strategy that addresses some of the disadvantages of mechanical ventilation. Three modes of high‐frequency ventilation (HFV) are used (Allan 2010).

  • High‐frequency jet ventilation (jet HFV).

  • High‐frequency oscillatory ventilation (oscillatory HFV).

  • High‐frequency percussive ventilation (percussive HFV).

High‐frequency jet ventilation delivers tidal volumes of 1 to 3 mL/kg at rates between 240 and 660 breaths per minute. Exhalation during jet HFV is passive, which is similar to that of conventional mechanical ventilation. Jet HFV has been shown to be a more effective means of treating pulmonary interstitial emphysema (Keszler 1991) and decreasing CLD in infants with severe respiratory distress (Keszler 1997) when compared with rapid rate conventional mechanical ventilation. Oscillatory HFV differs from jet HFV in that smaller volumes are delivered at a faster rate of about 8 to 15 Hz (Donn 2009). Percussive HFV involves small pulses of gas at ≥ 60 breaths/min that accumulates to form a 'low'‐frequency tidal volume breath; this technique was initially used in burn units and may have an application in neonatal ventilation (Allan 2010).

It has been postulated that coupling of HFV with a non‐invasive nasal delivery method may produce a synergistic effect that enhances the benefit of HFV. In a single case report, non‐invasive high‐frequency ventilation was shown to be effective for managing pulmonary emphysema in a premature infant (Al Tawil 2011). Non‐invasive delivery of HFV to newborn infants has been achieved successfully with the use of nasal prongs (De Luca 2010) and a nasopharyngeal tube (Colaizy 2008); benefits for CO2 removal have been observed. Non‐invasive high‐frequency ventilation was predicted to be superior to nIPPV for lung CO2 elimination in a newborn mannequin model, although it is not clear how nHFV achieves more effective CO2 elimination, or whether it provides adequate gas exchange in neonates (Mukerji 2013).

nHFV may be used in a similar fashion to HFV in several scenarios (Bhuta 1998; Cools 2009; Cools 2010; Joshi 2006). It may be used:

  • as initial respiratory support;

  • for respiratory support following planned extubation; and/or

  • following failure of initial non‐invasive therapy.

Non‐invasive high‐frequency ventilation has some potential problems. The pressure amplitude in nHFV is dampened by varying diameters of the circuit, nasopharyngeal tube, and airways; this makes it difficult to estimate the extent of the dampening variable. Resulting leakages and changes in airway patency may cause sudden undesirable changes in pressure delivery, leading to under‐ventilation or over‐ventilation (Carlo 2008).

How the intervention might work

Nasal high‐frequency ventilation operates at high frequency and low tidal volumes to allow gas exchange; this distinguishes it from conventional ventilation, which relies on large changes in pressure and volume (Ghazanshahi 1986; Habre 2010). In animal models, this ventilation method has been reported to result in more uniform lung inflation, to improve oxygenation, and to reduce the severity of lung pathology produced by conventional ventilation (Yoder 2000). The expiration phase in jet HFV is passive; this allows jet HFV to be used with lower mean airway pressure without risk of airway collapse (Brown 2011). In contrast, the expiration phase in oscillatory HFV is active, reducing expiratory time and preventing air trapping (Wheeler 2007).

Nasal high‐frequency ventilation introduced via a less invasive interphase (e.g. nasal or nasopharyngeal tube) may achieve adequate gas exchange, and may prevent intubation in newborn infants with respiratory distress or prevent extubation failure. A recent retrospective study reported the feasibility of nHFV in preventing intubation or facilitating extubation (Mukerji 2015).

Why it is important to do this review

Respiratory distress is a condition seen in newborn infants for whom ventilation is a major management strategy. However, application of positive‐pressure ventilation for an extended duration increases the likelihood of BPD (Ramanathan 2008).

Despite significant advances in neonatal intensive care, BPD is associated with higher risks of morbidity and mortality. Newborn infants surviving BPD are at increased risk for respiratory infection, asthma‐like disease, and pulmonary hypertension. They are more likely to be admitted to hospital during the first two years of life for lower respiratory tract infection (Greenough 2002), and they suffer more deficits in somatic growth and neurodevelopmental follow‐up (Reiterer 2013). Adult survivors of BPD have persistently greater impairment in general and respiratory health compared with adults born at term (Gough 2012). Neurodevelopmental impairment is also strongly associated with neonatal BPD (Singer 1997; Singer 2001; Vohr 2000).

Although comparison between modes of ventilation delivered by non‐invasive means such as nCPAP and nIPPV following extubation in preterm infants has been the topic of various Cochrane reviews (Davis 2001; Davis 2003; Lemyre 2014), to date no other systematic review has compared nHFV with other ventilation techniques.

A systematic review will gather available evidence for use of non‐invasive high‐frequency ventilation (nHFV) in newborn infants compared with other ventilation modes delivered invasively or non‐invasively.

Objectives

To determine effects of non‐invasive high‐frequency ventilation compared with other forms of respiratory support in newborn infants with respiratory distress in terms of need for intubation or reintubation and morbidity and mortality in term and preterm newborn infants.

We will examine use of nHFV in the following situations.

  • Initial respiratory support compared with other forms of invasive respiratory support (intubation and any ventilation mode).

  • Initial respiratory support compared with other forms of non‐invasive respiratory support (CPAP, nIPPV, HFNC) or air/oxygen.

  • An alternative to invasive endotracheal intubation for ventilation following failure of other types of non‐invasive respiratory support (CPAP, nIPPV, HFNC) in newborn infants with respiratory distress.

  • An alternative to other forms of non‐invasive respiratory support (nCPAP, nIPPV, HFNC) or air/oxygen following extubation.

We will include any interface used to deliver nHFV, including unilateral/bilateral and short/long nasal prongs, nasopharyngeal tube, face mask, and laryngeal mask airway.

Air/oxygen may involve low‐flow nasal prongs, a head box, or humidicrib oxygen delivery methods.

Methods

Criteria for considering studies for this review

Types of studies

We will consider randomised, quasi‐randomised, and cluster‐randomised trials eligible for inclusion in this review.

Types of participants

Preterm and term newborn infants with or at risk of respiratory distress requiring ventilation as initial support, following extubation or as a rescue mode.

Types of interventions

Separate comparisons of different modes of nHFV, including oscillatory nHFV, percussive nHFV, and jet nHFV:

  • for initial respiratory support;

  • following failure of initial non‐invasive therapy; or

  • for respiratory support following planned extubation.

We will compare modes of HFV with other types of ventilation delivered invasively (intermittent positive‐pressure ventilation (IPPV), high‐frequency oscillatory ventilation (HFOV)) or non‐invasively (nCPAP, nIPPV, HFNC, air, or oxygen). We will also compare different types of nHFV (oscillatory, percussive, and jet) with one other.

We intend to make the following separate comparisons for each of the following indications.

  • For initial respiratory support.

    • Non‐invasive high‐frequency ventilation versus invasive endotracheal intubation (plus ventilation by any mode).

    • Non‐invasive high‐frequency ventilation versus nasal continuous positive airway pressure.

    • Non‐invasive high‐frequency ventilation versus nasal intermittent positive airway pressure.

    • Non‐invasive high‐frequency ventilation versus high‐flow nasal cannula.

    • Non‐invasive high‐frequency ventilation versus air/oxygen.

    • Oscillatory nHFV versus percussive nHFV.

    • Oscillatory nHFV versus jet nHFV.

    • Percussive nHFV versus jet nHFV.

  • After failure of initial non‐invasive therapy.

    • Non‐invasive high‐frequency ventilation versus nasal continuous positive airway pressure.

    • Non‐invasive high‐frequency ventilation versus nasal intermittent positive airway pressure.

    • Non‐invasive high‐frequency ventilation versus high‐flow nasal cannula.

    • Oscillatory nHFV versus percussive nHFV.

    • Oscillatory nHFV versus jet nHFV.

    • Percussive nHFV versus jet nHFV.

  • For respiratory support following planned extubation.

    • Non‐invasive high‐frequency ventilation versus invasive endotracheal intubation (plus ventilation by any mode).

    • Non‐invasive high‐frequency ventilation versus nasal continuous positive airway pressure.

    • Non‐invasive high‐frequency ventilation versus nasal intermittent positive airway pressure.

    • Non‐invasive high‐frequency ventilation versus high‐flow nasal cannula.

    • Non‐invasive high‐frequency ventilation versus air/oxygen.

    • Oscillatory nHFV versus percussive nHFV.

    • Oscillatory nHFV versus jet nHFV.

    • Percussive nHFV versus jet nHFV.

Types of outcome measures

Primary outcomes

  • All‐cause mortality before hospital discharge

  • Need for endotracheal intubation or reintubation

Secondary outcomes

  • All‐cause mortality at 28 days

  • All‐cause mortality to follow‐up (≥ 1 year of age)

  • Failure of respiratory support or failure of extubation as defined by respiratory support failure criteria (e.g. partial pressure of carbon dioxide (pCO2) ≥ 60 mmHg and/or blood pH < 7.20; increased oxygen requirement; apnoea that is frequent or severe, leading to additional ventilatory support) or as defined by trial authors

  • Chronic lung disease, defined as need for oxygen or respiratory support at 36 weeks' postmenstrual age (Shennan 1988)

  • Trauma to the nostrils and upper airway

  • Patent ductus arteriosus

  • Pulmonary air leak syndromes, including pulmonary interstitial emphysema (PIE) and gross extrapulmonary air leak (such as pneumothorax)

  • Proven sepsis

  • Necrotising enterocolitis (NEC) (Bell stage ≥ 2) (Bell 1978)

  • Intraventricular haemorrhage (any and severe ‐ Papile grade 3/4) (Papile 1978)

  • Periventricular leukomalacia

  • Retinopathy of prematurity (any and severe ‐ stage ≥ 3) (International Committee 2005)

  • Neurodevelopmental disability at least 18 months' postnatal age (defined as neurological abnormality including cerebral palsy on clinical examination, or developmental delay more than two standard deviations below population mean on a standardised test of development, for instance, the Denver developmental screening test); blindness (visual acuity < 6/60); or deafness (any hearing impairment requiring amplification) at any time after term corrected)

  • Length of hospital stay (days)

Search methods for identification of studies

We will use the standard search strategy of the Cochrane Neonatal Review Group. Unpublished studies will be eligible for review. The search of MEDLINE and PREMEDLINE (via OVID interface) will include the following MeSH terms and free text words: (ventilation OR respiratory support) AND ((non‐invasive OR noninvasive OR nasal OR nasopharyngeal) AND high frequency), plus database‐specific limiters for randomised controlled trials (RCTs) and neonates (see Appendix 1 for full search strategies for each database). We will apply no language restrictions.

We will adapt the search terms above to suit other electronic sources.

Electronic searches

We will search the following electronic databases.

  • Cochrane Central Register of Controlled Trials (CENTRAL).

  • MEDLINE and PREMEDLINE (1946 to current) via OVID interface.

  • Embase (1974 to current).

  • Oxford Database of Perinatal Trials.

  • Cumulative Index to Nursing and Allied Health Literature (CINAHL; 1982 to current).

  • GoogleScholar (to locate studies not yet catalogued in the formal literature databases).

Searching other resources

We will conduct additional searches of the following.

  • Ongoing trials at the following trial registries.

  • Abstracts from the following conferences.

    • Proceedings of the Pediatric Academic Societies (American Pediatric Society/Society for Pediatric Research and European Society for Paediatric Research) from 1990 to current in the journal Pediatric Research and Abstracts Online.

    • Proceedings of the European Academy of Paediatric Societies (EAPS), the European Society for Paediatric Research (ESPR), the European Academy of Paediatrics (EAP), and the European Society of Paediatric and Neonatal Intensive Care (ESPNIC) from 2003 to current from Abstracts Online.

    • Proceedings of the Perinatal Society of Australia and New Zealand (PSANZ) from 1996 to current (handsearch).

  • Reference lists: We will screen the reference lists of these papers after reading identified individual studies that examined effects of nHFOV on morbidity or mortality, or both, in newborn infants at risk of respiratory distress, to identify other relevant studies.

  • Personal communications.

    • If we identify any unpublished trial, we will contact the corresponding investigator for information on unpublished trials potentially eligible for inclusion. Unpublished studies will be eligible for review.

    • We will contact the corresponding authors of identified RCTs for additional information about their studies when data provided in the studies are deemed insufficient.

    • We will contact study authors who published in this field to ask about possible unpublished articles.

    • Medical ventilator companies: We will contact companies that develop high‐frequency ventilators to ask about possible unpublished studies using their product.

Data collection and analysis

We will use the standardised review method of the Cochrane Neonatal Review Group (CNRG) for conducting a systematic review (http://neonatal.cochrane.org/en/index.html). We will enter and cross‐check data using Review Manager 5.3 software (RevMan 2014).

Selection of studies

Review authors will independently review the titles and abstracts of potentially relevant studies identified by the search strategy against the inclusion and exclusion criteria to assess eligibility for inclusion in this review. We will retrieve full‐text versions when we wish to examine eligible studies more closely, or when inadequate information is provided in the abstract.

Data extraction and management

Review authors will independently extract data using specially designed data extraction forms. We will use retrieved information to determine trial eligibility by extracting methods and data from eligible trials and by requesting additional unpublished information from authors of original reports. We will enter and cross‐check data using RevMan 5.3 software (RevMan 2014) and will compare extracted data for any differences. If noted, we will resolve differences through mutual discussion and consensus.

Assessment of risk of bias in included studies

We will use the standardised review methods of the CNRG (http://neonatal.cochrane.org/en/index.html) to assess the methodological quality of included studies.

At least two review authors will independently assess study quality and risk of bias using the following criteria, as documented in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Sequence generation (to assess selection bias)

For each included study, we will analyse the method used to generate the allocation sequence, to determine whether it should produce comparable groups. We will assess the method as:

  • low risk (random component in the sequence generation process, e.g. random number table; coin tossing; throwing dice);

  • high risk (non‐random component in the sequence generation process, e.g. date of birth; date of admission; hospital record number); or

  • unclear risk.

Allocation concealment (to assess selection bias)

For each included study, we will analyse the method used to conceal the allocation sequence, to determine whether intervention allocation could have been foreseen in advance of or during recruitment, or changed after assignment. We will assess the method as:

  • low risk (participants and investigators enrolling participants could not foresee assignment of allocations e.g. telephone allocation; opaque, sealed envelopes);

  • high risk (participants or investigators enrolling participants could possibly foresee assignment of allocations e.g. unsealed envelopes); or

  • unclear risk

Blinding of participants and personnel (to assess performance bias)

For each included study, we will analyse the method used to blind participants and personnel from knowledge of which intervention a participant received. We will assess the method as:

  • low risk (e.g. blinding of participants and key study personnel);

  • high risk (e.g. no blinding or incomplete blinding of participants and key study personnel); or

  • unclear risk.

However, we anticipate that bias is likely, as blinding of clinicians was not likely to occur in these studies owing to the nature of the intervention.

Blinding of outcome assessment (to assess detection bias)

For each included study, we will analyse the method used to blind outcome assessors from knowledge of which intervention a participant received. We will assess the method as:

  • low risk (e.g. blinding of outcome assessment ensured);

  • high risk (e.g. no blinding of outcome assessment); or

  • unclear risk.

Incomplete outcome data (to assess attrition bias)

For each included study, we will analyse the completeness of data, including attrition and exclusions from analysis. We will state whether attrition and exclusions were reported, numbers included in the analysis at each stage (compared with total numbers of randomised participants), reasons for attrition or exclusion when reported, and whether missing data were balanced across groups or were related to outcomes. We will assess completeness of data as:

  • low risk (e.g. no missing outcome data, < 20% missing data);

  • high risk (e.g. reason for missing outcome data likely to be related to true outcome, ≥ 20% missing data); or

  • unclear risk.

Selective reporting (to assess reporting bias)

For each included study, we will investigate the possibility of selective outcome reporting bias. We will assess the method as:

  • low risk (e.g. study protocol is available and all of the study’s prespecified (primary and secondary) outcomes have been reported in the prespecified way);

  • high risk (e.g. not all of the study’s prespecified primary outcomes have been reported); or

  • unclear risk.

Other bias

For each included study, we will analyse any other possible sources of bias. We will assess the method as:

  • low risk (the study appears to be free of other sources of bias);

  • high risk (e.g. the study had a potential source of bias related to the specific study design used); or

  • unclear risk.

When necessary, we will request additional information and clarification of published data from the authors of individual trials. We will resolve discrepancies through discussion and consensus; if necessary, we will provide levels of agreement between review authors and/or details of resolution of differences.

Measures of treatment effect

We will analyse study results using Review Manager 5.3 (RevMan 2014).

We will report each continuous outcome as a weighted mean difference (WMD) with a 95% confidence interval (CI). We will report each categorical outcome as a risk ratio (RR) and a risk difference (RD) with a 95% CI. For results that are statistically significant, we will use 1/RD to calculate the number needed to treat for an additional benefit outcome (NNTB) or the number needed to treat for an additional harmful outcome (NNTH).

Unit of analysis issues

For parallel‐group trial designs by which infants are randomised to receive one or more different types of ventilation (nHFV, invasive ventilation, or an alternative type of non‐invasive ventilation such as nCPAP, NIPPV, HFNC, or air/oxygen), the unit of analysis for both short‐term and long‐term outcomes will be the infant.

For trial designs in which the unit of randomisation is the individual ventilation episode, and in which infants may be allocated more than once, the unit of analysis will be the ventilation episode, and we will report only short‐term outcomes. We will analyse and report these trials separately from parallel‐design trials. When trials report cross‐over of nHFV versus invasive ventilation or an alternative non‐invasive ventilation (nCPAP, NIPPV, HFNC, air/oxygen), we will report only short‐term outcomes for the first ventilation episode.

Cluster‐randomised trials

The unit of analysis for cluster‐randomised trials will be the randomised treating centre or cluster. We plan to include cluster‐randomised trials in the analyses, using an estimate of the intracluster correlation coefficient (ICC) derived from the trial (if possible), or from another source, as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). If ICCs from other sources are used, we plan to report this and to conduct sensitivity analyses to investigate effects of variation in the ICC. If we identify both cluster‐randomised trials and individually randomised trials, we plan to synthesise the relevant information. We will consider it reasonable to combine the results of both types of studies if we note little heterogeneity between study designs, and if interaction between effects of the intervention and choice of randomisation unit is considered unlikely.

Dealing with missing data

We will obtain missing data from study authors when possible. When this is not possible, we will conduct analyses using available data (i.e. ignoring missing data). In addition, we will document missing data and will perform a sensitivity analysis.

For dichotomous outcomes, we will conduct both best‐case and worst‐case scenarios as well as intention‐to‐treat (ITT) analysis with imputation. We will compare results obtained from two analysis options to gain a better understanding of the robustness of results relative to different analytical approaches. We will consider an imputation approach for best‐case (i.e. all missing participants in the intervention group did not experience poor outcomes (e.g. death, BPD), and all missing participants in the control group experienced poor outcomes) and worst‐case (i.e. all missing participants in the intervention group experienced the event, and all missing participants in the control condition did not) scenarios. We will conduct sensitivity analysis to compare results based on different imputation assumptions (i.e. best‐case vs worst‐case scenarios).

Assessment of heterogeneity

We will use RevMan 5.3 software (RevMan 2014) to assess the heterogeneity of treatment effects between trials. We will undertake this assessment using the following two formal statistical models.

  • Chi2 test, to assess whether observed variability in effect sizes between studies is greater than would be expected by chance. As this test has low power when few studies are included in the meta‐analysis, we will set the probability at the 10% level of significance.

  • I2 statistic, to ensure that pooling of data is valid. We will grade the degree of heterogeneity as follows: none (< 25%); low (25% to 49%); moderate (50% to 74%); or high (≥ 75%). When we find evidence of apparent or statistical heterogeneity, we will assess the source of heterogeneity by performing sensitivity and subgroup analyses, while looking for evidence of bias or methodological differences between trials.

Assessment of reporting biases

If we identify 10 or more studies that include a specific intervention (comparison) and report on the same outcome, we will assess reporting and publication biases by examining the degree of asymmetry of a funnel plot in RevMan 5.3 (RevMan 2014).

Data synthesis

We will perform statistical analyses according to recommendations of the CNRG (http://neonatal.cochrane.org/en/index.html). We will analyse all newborn infants randomised on an intention‐to‐treat basis and will examine treatment effects in individual trials using fixed‐effect modelling to combine data. For meta‐analyses of categorical outcomes, we plan to calculate typical estimates of RR and RD, each with 95% CI; for continuous outcomes, we plan to calculate the WMD if outcomes are measured in the same way between trials, and standardised mean difference (SMD) if trials measuring the same outcome using different scales are combined. When meta‐analysis is judged to be inappropriate, we will analyse and interpret individual trials separately. If we note high heterogeneity, we will report a typical effect.

Quality of evidence

We will use the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach, as outlined in the GRADE Handbook (Schünemann 2013), to assess the quality of evidence for the following (clinically relevant) outcomes: all‐cause mortality before hospital discharge; failure of respiratory support for extubation; need for endotracheal intubation or reintubation; and chronic lung disease.

Two review authors will independently assess the quality of the evidence for each of the outcomes above. We will consider evidence from RCTs as high quality but will downgrade the evidence one level for serious (or two levels for very serious) limitations based upon the following: design (risk of bias), consistency across studies, directness of the evidence, precision of estimates, and presence of publication bias. We will use the GRADEpro GDT Guideline Development Tool to create a ‘Summary of findings’ table to report the quality of evidence.

The GRADE approach yields an assessment of the quality of a body of evidence according to one of four grades.

  • High: We are very confident that the true effect lies close to the estimate of effect.

  • Moderate: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of effect but may be substantially different.

  • Low: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of effect.

  • Very low: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.

Subgroup analysis and investigation of heterogeneity

If sufficient data are available, we will explore potential sources of clinical heterogeneity by analysing whether results differed for newborn infants:

  • with gestational age ≥ 37 weeks (term), < 37 weeks (preterm), < 32 weeks (very preterm), or < 28 weeks (extremely preterm);

  • ventilated with lower (< 10 cm H2O) versus higher airway pressures (≥ 10 cm H2O);

  • ventilated with lower (< 10 Hz) versus higher frequencies (≥ 10Hz); or

  • with interface used to deliver nHFOV: unilateral/bilateral and short/long nasal prongs, nasopharyngeal tube, face mask, laryngeal mask airway.

Sensitivity analysis

If sufficient data are available, we will explore methodological heterogeneity by performing sensitivity analyses. When possible, we will conduct sensitivity analyses to assess any change in the direction of effect caused by inclusion of studies of lower quality, based on assessment of: allocation concealment, adequate randomisation, blinding of treatment, greater than 10% loss to follow‐up, and intention‐to‐treat analyses.