The protocol of this review was first registered with the International Prospective Register of Systematic Reviews (PROSPERO), and Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) statement (Supplement S1)[8] was followed while conducting and reporting the protocol of this systematic review and network meta-analysis (http://www.prisma-statement.org).
Eligibility Criteria
Type of study
The review will include randomized controlled trials (RCTs), case control study and cohort study that were reported in English or Chinese without any regional restrictions, irrespective of publication status and publication year. Case report, animal experiment, vitro cell experiment, expert experience, conference article, review, communication letter, comment and duplication publication will be excluded.
Type Of Patients
Patients with confirmed COVID-19 infection and hypoxemic respiratory failure (despite oxygen therapy) in any setting that has capacity for non-invasive ventilation (NIV) or invasive ventilation.
Type Of Intervention
In this systematic review, we defined NIV as continuous positive airway pressure (CPAP), biphasic positive airway pressure (BiPAP), proportionally assisted ventilation (PAV), and high-flow nasal catheter inhalation (HFNC).The invasive ventilation was tracheotomy, endotracheal intubation and laryngeal mask airway, while ventilator ventilation model was performed by pressure-controlled ventilation (PCV), volume-controlled ventilation (VCV), synchronous intermittent instruction ventilation (SIMV), spontaneous breathing support (PSV) and pressure-controlled volume-guaranteed ventilation (PCV-VG).
Type Of Control Group
Control treatments included early noninvasive or invasive ventilation, tracheotomy, sham operative ventilation. Studies comparing the therapeutic effects of different ventilation modes will also be included.
Type Of Outcome Measures
Studies that reported one or more of the following outcomes will be included.
Primary outcome
-
All-cause mortality.
-
Transmission of COVID-19 to health care workers and other people.
-
Length of hospital stay.
-
Length of ICU stay.
Secondary outcome
-
Complications of therapy.
-
Secondary bacterial pneumonia.
-
Need for invasive ventilation.
-
Need for tracheostomy.
-
(Time to) Recovery from COVID-19.
-
Aerosol generation and droplet dispersion of live virus at various distances and times.
-
Contextual outcomes including acceptability, feasibility, resources use and impact on equity.
-
Medical staff infected COVID-19.
Type Of Language
We will only include only English or Chinese literature.
Search Method
Two authors will independently search the electronic databases Web of Science, Embase, and the Cochrane Central Register of Controlled Trials (CENTRAL), Medline through Pubmed from December 1, 2019 to November 9, 2020. We will search the Chinese electronic databases: Chinese National Knowledge Infrastructure (CNKI), Wanfang Database, VIP Database, Chinese Biomedical Literature Database (CBM), and we will also search World Health Organization (WHO) Clinical Trials Registry, Chinese clinical registry, ClinicalTrials.gov, ProQuest Central, SciFinder, the Virtual Health Library, LitCovid, WHO covid-19 website, Centers for Disease Control (CDC) covid-19 website, Eurosurveillance, China CDC Weekly. We will also search bioRxiv (preprints), medRxiv (preprints), chemRxiv (preprints), and SSRN (preprints), Chinese Medical Journal Net (preprints), and ChinaXiv (preprints).
A subject-specific search strategy (Supplement S2) was developed in Pubmed/Medline and applied as the basic search strategy in the other electronic databases. We also scanned the reference lists and citations of the included trials and any relevant systematic reviews, referring to further literature. If necessary, we contacted trial authors to obtain additional information.
Data Collection
Two investigators independently extracted the data of all included studies using a standard data exaction form prepared by authors, on which they recorded the first author, publication year, study design, ventilation modes, population characteristics, and reported outcomes of interest. If necessary, we contacted the authors of included studies by email to obtain additional data missing from the published paper. If additional information could not be obtained in this manner, we extracted data and information from the figures. At each step of data extraction, we resolved any differences between the investigators by discussion or consulting an arbiter.
Study Selection
Two authors independently scanned the title and abstract of the identified records to exclude uncorrelated studies and accessed the full-text articles for eligibility. We resolved any disagreements regarding inclusion of the study by discussion or by consulting an arbiter. The study selection procedure will be performed according to the PRSIMA flow chart (Figure. 1).
Risk-of-bias In Individual Studies
Two authors will respectively assess the risk-of-bias of all included studies, using the Cochrane risk-of-bias tool. We will appraise each study in terms of selection bias (random sequence generation and allocation concealment), performance bias (blinding of participants and personnel), detection bias (blinding of outcome assessment), attrition bias (incomplete outcome data), selective reporting bias and other bias. Each component was judged as low risk-of-bias (green), high risk-of-bias (red), or unclear risk-of-bias (yellow).
Quality Assessment
The Grading of Recommendations Assessment, Development and Evaluation System (GRADE) will be used to evaluate the quality of the evidence for primary outcomes[9]. Two reviewers will independently assess the quality of evidence using GRADE software, the quality of evidence will be divided into four levels: high, medium, low and very low, according the GRADE standards. For a specific study, the quality will be assessed according to the risk of bias, inconsistency, indirectness, imprecision and publication bias.
Assessment Of Similarity And Consistency
The assessment of similarity and consistency will be conducted to get a reliable and valid result. At present, it is difficult to determine similarity by statistical analysis, so we will evaluate the similarity of included studies by analyzing clinical and methodological characteristics, patient characteristics, experimental design, and interventions. We will use the z-test to check for consistency, and calculate the P-value to check for inconsistencies between direct and indirect comparisons and a P-value greater than 0.05(P > 0.05) is considered to be consistent between the direct comparison and the indirect comparison, on the contrary, there is inconsistency.
Network Meta-analysis
Data synthesis and statistical analysis will be performed using R3.6.3 (https://www.r-project.org/) with the Bayesian framework in this NMA. For continuous outcomes, standard mean difference (SMD) with 95% confidence intervals (CIs) or a weighted mean difference (WMD) with 95%CIs will be applied to analyze the efficacy data, while risk ratio (RR) with 95% CIs will be used for dichotomous data. WMD will be used for the same units or for the same measurement method while SMD will be used for different scale or assessment tools.
The contribution of each study to the final effect will be assessed and demonstrated by mapping the net-heat map. P value will be used to measure the accuracy of different ventilation modes over the control group and different ventilation modes were ranked by P value. P value equal to 100 percent, it will be considered as the best treatment for patients, while P value equal to 0 percent, it will be considered the worst treatment. The results of the NMA will be shown using forest maps, the ranking results of different ventilation modes will be displayed by using the surface under the cumulative sorting curve analysis, and the network evidence plot will be applied to present the comparison between different interventions.
Assessment Of Heterogeneity
For the heterogeneity of clinical and methodological heterogeneity, the sources of heterogeneity will be assessed by careful assessment of participant characteristics, study design, randomization, blinding, interventions, and outcomes of all included studies. The I2 statistics was used to report statistical heterogeneity, which was classified by applying the Cochrane Handbook of Systematic Reviews of Interventions. When I2 > 50%, obvious heterogeneity existed, on the contrary, when I2 < 50%, heterogeneity is considered acceptable. Furthermore, if I2 > 50%, several subgroup analyses and sensitivity analyses were conducted to evaluate the source of heterogeneity and to assess the subpopulation of patients that could probably obtain benefits from the intervention. The final NMA will be performed after the removal of studies that produce primary or unacceptable sources of heterogeneity. Lastly, if the source of heterogeneity cannot be explored after subgroup analyses and sensitivity analyses, a narrative review will be provided.
Meta-regression, Subgroup Analysis And Sensitivity Analysis
The network meta-regression will be conducted to explore the sources of heterogeneity by using random effects models. We will also perform subgroup analyses based on the following items.
-
Age of participants: <18, 18 to 40, 50, 60, > 60 years
-
Interface (helmet compared to oronosal or full-face mask) or tracheotomy or tracheal intubation
-
Degree of hypoxia
-
Study design type
-
Comorbidities (diabetes, obesity, chronic kidney disease, congestive heart failure, chronic obstructive pulmonary disease, asthma, chronic liver disease, hypertension, depression, actively treated malignancy, immunosuppression (due to illness or meds))
-
Type of healthcare setting (secondary or tertiary)
-
NIV or invasive ventilation
For all outcomes, we will first include all the trials regardless of peer reviewed or non-peer reviewed, and then we will conduct sensitivity analysis limited to peer reviewed researches. Sensitivity analysis will be conducted to explore the source of heterogeneity and ensure that a stable and accuracy conclusion can be obtained.
Patients and public involvement
All data were extracted from the literature and there will be no patients and public will be directly included in this systemic review and network meta-analysis.