Article Text
PDF
Abstract
Introduction Acute hypoxic respiratory failure (AHRF) is one of the most common causes of admission to paediatric intensive care units (PICUs) around the world, posing a serious health concern for the global community. Non-invasive respiratory support (NRS) is considered effective in reducing mortality and intubation rates in adults. However, it is not yet clear whether NRS is beneficial for children and which NRS modalities are most effective. This network meta-analysis aims to summarise existing evidence and compare the efficacy and safety of different NRS modalities in paediatric patients with acute hypoxaemic respiratory failure.
Methods and analysis To identify randomised controlled trials, we will perform a systematic search of key databases (Embase, PubMed, CENTRAL, CINAHL Complete and Web of Science) and registered clinical trials (ClinicalTrials.gov, WHO ICTRP and ISRCTN). To ensure the inclusion of the latest literature, an initial pilot search was conducted on 8 July 2024, and an updated search will be conducted after the main research work of this study. AHRF in children treated with NRS will be included. Hospital mortality, intubation rate, treatment failure rate and serious adverse events are critical outcomes closely related to patient-centredness and importance. Two authors will independently select the studies and extract the data. The risk of bias will be assessed using the Cochrane risk of bias tool V.2.0. In order to compare the effects of different NRS modalities, pairwise meta-analysis and network meta-analysis will be conducted using R software. Several subgroup analyses will be conducted, including analyses of different causes of AHRF. We will conduct sensitivity analyses by excluding studies with a high risk of bias and those involving neonates. Using the Grading of Recommendations Assessment, Development and Evaluation methodology, we will assess the certainty of the evidence for the effect estimates of all the outcomes.
Ethics and dissemination Since this research is a network meta-analysis based on published literature, no formal ethics approval is required. The results will be disseminated through a peer-reviewed journal for publication.
PROSPERO registration number CRD42024529804.
- Network Meta-Analysis
- Paediatric intensive & critical care
- Randomized Controlled Trial
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
The preferred reporting items and meta-analysis protocol guidelines for systematic reviews and their extension for network meta-analyses will be strictly followed in this systematic review.
The operationalisation of the search strategy will be developed by an experienced librarian and tailored to multiple databases and registered clinical trials.
The quality of evidence on the effects of interventions will be reported using the Grading of Recommendations Assessment, Development and Evaluation to facilitate their application.
The possibility of limited studies and the low quality of some studies may affect the outcome or evidence of this systematic review.
There will be no restrictions on the language, publication year, gender, or race of the articles included.
Introduction
Acute hypoxaemic respiratory failure (AHRF) is a common paediatric problem accounting for a significant percentage of paediatric intensive care unit (PICU) admissions and deaths with a significant global health impact.1–4 Invasive mechanical ventilation (IMV) may be effective in improving outcomes for children with AHRF through the use of artificial airways connected to a ventilator. However, there are a number of adverse events associated with IMV, including ventilator-associated pneumonia, ventilator-induced lung injury, oversedation and damage to the larynx and trachea.5 Currently, the primary objective of treating paediatric patients with AHRF remains to reduce mortality and avoid unnecessary endotracheal intubation.6 Several non-invasive respiratory support (NRS) modalities have been reported to reduce mortality and prevent endotracheal intubation. However, which of these is most effective is unclear.7–9
Given the current clinical dilemma of selecting the most effective NRS for children with AHRF, a network meta-analysis (NMA) might be an appropriate approach to address this question. The NMA is a form of meta-analysis for synthesising the evidence derived from clinical trials involving multiple interventions.10 When head-to-head comparisons are scarce, NMAs allow estimation of treatment effects, as well as reducing imprecision in pooled treatment effects, by incorporating estimates from direct and indirect comparisons.11
In recent years, several network meta-analyses have examined the best NRS for children suffering from respiratory diseases. In 2021, a Bayesian NMA examined options for non-invasive ventilation (NIV) strategies in children with acute lower respiratory infections.12 This study examined multiple ventilation strategies (continuous positive airway pressure (CPAP), high flow nasal cannula (HFNC), bilevel positive airway pressure (BiPAP) and standard oxygen therapy (SOT)) and concluded that, compared with SOT, CPAP reduced the risk of intubation in children with acute lower respiratory infections. Also, both CPAP and HFNC were associated with lower rates of treatment failure than SOT. A subsequent NMA of NRS modalities choices for postextubation children also yielded similar results.13 An NMA conducted recently on the selection of NRS in patients with bronchiolitis, however, produced inconsistent results.14 The study compared HFNC with other NRS therapies in bronchiolitis and concluded that experimental evidence does not indicate that HFNC has advantages over either SOT as initial treatment or NIV as rescue treatment. Although AHRF is the common endpoint for a number of underlying diagnoses, including bronchiolitis, asthma and pneumonia,15 no NMA has been conducted to evaluate NRS selection in paediatric patients with AHRF.
In order to address these gaps in knowledge, we plan to perform a systematic review and NMA to determine the optimal NRS for paediatric patients with AHRF.
Methods
The protocol has been registered in the International Prospective Register for Systematic Reviews (PROSPERO, CRD42024529804) and is being reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) and its extension for NMA (online supplemental material 1).16 17 Research on this project began in May 2024, and results are expected to be released in December 2024.
Supplemental material
Criteria for included studies
Participants and settings
Studies that enrolled paediatric patients (28 days–18 years) with AHRF, defined as the development of new clinical signs (tachypnoea, increased breathing work), radiological signs (unilateral or bilateral chest X-ray opacities) and gas exchange changes (oxygen saturation (SpO2) of <92% in room air or SpO2 92%–94% with chest indrawing (subcostal/intercostal/suprasternal retraction) and/or depressed sensorium), will be included. Studies involving patients treated in the PICU, medical wards and emergency departments will be considered. We will exclude research involving central or peripheral neuromuscular diseases, oncological diseases and chronic respiratory conditions (ie, chronic respiratory diseases persisting for more than a year) and not reporting the outcomes of interest. The postoperative setting and postextubation setting will also be excluded. Specifically, we will exclude studies that had enrolled more than 50% neonates, acknowledging that a lower threshold would exclude too many studies.
Interventions and comparators
The interventions in this review will be any NRS modality. With HFNC, oxygen is delivered through a nasal cannula at fractional inspired oxygen (FiO2) up to 1 and a flow rate up to 60 L/min. With CPAP, the patient breathes with a constant level of positive pressure that is maintained during both inspiration and expiration. The NIV provides positive pressure ventilation using different modalities: BiPAP, pressure support ventilation, etc. Depending on the interface, a face mask, nasal mask or helmet can be used. Considering the potential differences between reported outcomes, these interfaces will be considered to be three separate interventions. The comparators will be SOT or another NRS modality. SOT (low-flow systems) consist of traditional nasal cannulas, Venturi system masks or non-rebreather masks. The respiratory modalities above can all deliver oxygen at low flow (<15 L/min) at varying levels of FiO2.
Outcomes measures
Prior to the literature search, outcomes will be selected and rated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach for patient-centredness and importance.18 We will categorise the outcomes as follows. Critical outcomes will include hospital mortality, intubation rate, treatment failure rate (intubation or escalation/crossover to another NIV mode) and serious adverse events (pneumothorax, unexpected emergency intubations, respiratory or cardiac arrest and death before hospital discharge). Important outcomes will include hospital length of stay, PICU admission rate, PICU length of stay, total duration of NRS, ventilation intolerance rate, comfort and pressure injuries. Outcomes of limited importance will include abdominal distension and the need for sedatives.
Study design and publication types
We will only include randomised controlled trials (RCTs) that explored the effects of HFNC, CPAP, NIV and SOT in the treatment of AHRF. All non-RCTs, quasi-RCTs, cohort studies, case-control studies and cross-sectional studies will be excluded from this study. We will also exclude case reports or case series, study protocols, reviews, comments, conference abstracts and letters to the editor.
Information sources and search strategy
A comprehensive search strategy will be applied to the following five electronic bibliographic databases: (1) EMBASE, (2) PubMed, (3) Cochrane Central Register of Controlled Trials, (4) EBSCO CINAHL Complete and (5) Web of Science. Additionally, we will search ClinicalTrials.gov, the WHO International Clinical Trials Registry Platform and the International Standard Randomised Controlled Trial Number Registry for all registered clinical trials. A pilot search was conducted on 8 July 2024 from the period of inception to date. In order to ensure that all the newest publications are included in the network meta-analysis, we will conduct an updated literature search before completing the analysis. The Peer Reviewed Electronic Search Strategies (PRESS) 2015 guidelines19 will guide the search strategy (requestor: YF, reviewer: JQ). The full search strategy can be found in the online supplemental material 2. Using a validated search filter from the Cochrane Handbook for Systematic Reviews of Interventions V.6.4.0, Section 4.4.7, we will screen EMBASE and PubMed for RCTs.20 For EBSCO CINAHL Complete, we will adapt a pretested search filter for RCTs from the Scottish Intercollegiate Guidelines Network.21 Language, publication year, gender or race will not be restricted. The citation management software EndNote 21 (Clarivate Analytics, Philadelphia, Pennsylvania, USA) will be used to manage all references and duplicates.
Study selection
Based on the title and abstract of the studies, two authors will independently select potential trials based on inclusion and exclusion criteria. Afterwards, we will obtain full texts of the articles that meet these prespecified criteria and review them again. It will be decided by a third investigator if there is any disagreement between the reviewers. The PRISMA flow diagram attached in figure 1 illustrates the process of study selection.22
Preferred Reporting Items for Systematic Reviews and Meta-Analyses-Protocol flow diagram.
Data extraction
Using a prepiloted Excel (Microsoft) data extraction form, the two reviewers will extract data independently. If there are any differences between the authors, they will be resolved through consensus or discussion with a third author. Study characteristics (trial design, size and funding source), patient characteristics (age, gender, aetiology of AHRF, etc), details of the interventions (location of application, duration of exposure to NRS and modality of NRS) and outcomes for each endpoint will all be extracted.
Dealing with missing data
In cases where data were missing or unclear in trial publications, we will contact trial authors via email to request additional information. Where possible, we will use the methods in Wan et al,23 with conversion calculations performed with the Excel spreadsheet accessible in this article as recommended in the Cochrane Handbook for Systematic Reviews of Interventions to estimate missing means. We will also use the calculator available in RevMan. We will omit results if there is insufficient data available for conversion.
Risk of bias assessment
The Cochrane risk-of-bias tool (RoB2) V.2 for RCTs will be used to assess bias independently. The tool assesses the following bias domains: (1) random sequence generation (selection bias), (2) allocation concealment (selection bias), (3) blinding of participants and personnel (performance bias), (4) blinding of outcome assessment (detection bias), (5) incomplete outcome data (attrition bias), (6) selective reporting (reporting bias) and (7) other sources of bias.
Data synthesis and analysis
A pairwise Bayesian meta-analysis will be performed using a random-effects model, followed by a network meta-analysis using a Bayesian framework to compare all interventions head-to-head. Markov chain Monte Carlo methods will be used to analyse event rates, treatment effect estimates and heterogeneity estimates using prior distributions derived from previous literature.24 Parameter settings: the number of chains was 6, the initial value was 2.5, the number of adaptation (or tuning) iterations was 50 000, the number of simulation iterations was 200 000 and the thinning factor was 10. Pairwise and NMA treatment effect estimates will be reported as risk ratios, using the median and 95% credible intervals. Additionally, we will rank interventions based on their surface under the cumulative ranking curve. The probability that interventions reduce or increase risks greater than SOT will also be calculated according to their apparent effectiveness. We will quantify heterogeneity in treatment effects between studies using the posterior distribution τ2. Intransitivity will be assessed by tabulating the included studies and by comparing the population, interventions, comparisons and outcomes across studies. We will estimate inconsistencies (incoherences) between direct and indirect comparisons using a node-splitting approach, contrasting estimates from direct and indirect evidence.25 26 In the presence of inconsistency, we will explore analytical strategies, such as splitting the network into subgroups or using Bayesian network meta-regression to explore sources of inconsistency and heterogeneity (see below ‘Subgroup analyses’ section).20 The Brooks-Gelman-Rubin diagnostic, trace plots and autocorrelation plots will be used to assess model convergence visually. By comparing the mean residual deviance of our final models with the number of data points contributing, calculating the percentage of node-based residuals within 1.96 of the standard normal distribution and visually inspecting the residual distribution on Q–Q plots, we will be able to assess the goodness of fit of our final models. All analyses will be performed in R V.4.3.2.
Assessment of publication bias
An examination of the publication bias in this network meta-analysis will be conducted through the use of funnel plots and Egger’s test.27 By examining the funnel plot’s shape, we assume there may be publication bias if the shape is asymmetrical. Using Egger’s test, we will measure funnel plot asymmetry formally.28
Subgroup analyses
When applicable, we will perform meta-regressions or subgroup analyses to address heterogeneity and inconsistency. In subgroup analyses, the following potential factors can be evaluated: (1) severe AHRF versus less severe AHRF (we determine the severity of AHRF using the PaO2:FiO2 ratio or respiratory rate oxygenation (ROX) or modified mROX based on the characteristics of the study), (2) immunocompromised patients (haematological malignancies, HIV infection or exposure, severe malnutrition, treatment with immunosuppressant drugs and solid organ transplant recipients) versus non-immunocompromised patients, (3) cause of respiratory failure, (4) age, (5) low-income countries versus high-income countries, and (6) PICU patients versus non-PICU patients.
Sensitivity analyses
An analysis of sensitivity will be conducted using the one-by-one elimination method in order to verify the robustness of the study conclusions. We will perform a sensitivity analysis by excluding studies with a high risk of bias. Furthermore, we will exclude studies with neonates and with non-informative prior distributions for the heterogeneity parameters in our sensitivity analyses.
Grading of recommendations
According to the GRADE working group, we will grade the quality of every direct, indirect and NMA estimate.29 The certainty of evidence will be divided into four levels: high, moderate, low and very low. Plain language statements determined using the ratings for the certainty of evidence will be presented with the ratings.30 A summary of findings table will display the network effect estimates for each of the potential comparisons and each outcome, along with the GRADE rating for certainty in the evidence.
Ethics and dissemination
This research is based on the literature in the database and does not require the approval of the ethics committee. The published article is to be uploaded to academic and public science repositories and presented at conferences, symposia and congresses.
Ethics statements
Patient consent for publication
Acknowledgments
The Lanzhou University First Hospital's Research Center for Clinical Medicine and the Lanzhou University Evidence-Based Medicine Center provided assistance and guidance in the development of this protocol.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Y-DF and Y-XL are joint first authors.
Y-DF and Y-XL contributed equally.
Contributors Y-DF and Y-XL conceived the study and contributed clinical expertise and background. Y-DF, J-ZL, JQ and Y-QY contributed with the analytical plan and the bias assessment approach. Y-DF and JQ performed the literature search plan. Y-DF and Y-XL drafted the protocol. All authors revised the protocol and approved the final version. J-ZL supervised the study. Y-QF is the guarantor.
Funding All phases of this study were supported by the National Natural Science Foundation of China (No. 82160921), National Administration of Traditional Chinese Medicine on Announcement of the Fifth Batch of National Training Projects for Clinical Talents in Traditional Chinese Medicine (Guo Zhong Yi Ren Jiao Han (2022) No. 1).
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.