Protocol and registration
This protocol of systematic review and meta-analysis will be conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) guidelines27. See (Appendix 1) for the completed PRISMA-P checklist.
This protocol is registered in the International Prospective Register of Systematic Reviews (PROSPERO) system and can be accessed at https://www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=181234.
Eligibility criteria
1. Type of studies to be included
All relevant RCTs with or without blinding will be included. The quasi-RCTs, non-RCTs, cross-over study, and uncontrolled clinical trials will be excluded. There will be no language restrictions.
2. Types of participants
Adult participants diagnosed as SP according to the Infectious Diseases Society of America/American Thoracic Society guidelines will be inclduded28. There will be no limitation on age, gender, education, ethnicity, and economic status. Patients with serious complications will be excluded.
3. Types of interventions
On the basis of conventional treatment, the treatment intervention will be XBJ. The control intervention will be based on a placebo or no intervention. Additionally, trials that evaluate XBJ plus another therapy compared with the other therapy alone will also be included.
4. Types of outcome measures
(1) Primary outcomes
Severe pneumonia-related mortality.
(2) Secondary outcomes
The secondary outcomes will include (a) proportion of total efficiency rate including cure rate, obvious effective rate, and effective rate; (b) improvement in acute physiology and chronic health evaluation II; (c) improvement of pneumonia severity assessed by clinical scales including clinical pulmonary infection score or pneumonia severity index; (d) course of antibiotic use; (e) total duration of ICU stays and hospitalization.
(3) Safety outcomes
Adverse effects of XBJ include aphylactic reaction, impairment of liver and kidney function, nausea, and skin irritation.
Search strategy
We will search for the following databases from their inception until 1 October 2020: MEDLINE, EMBASE, CENTRAL, and the following Chinese databases and sources: CNKI, VIP, SinoMed, Wanfang Data. Additionally, we will search for dissertations, clinical trial registers, and grey literature. The following registers for ongoing or unpublished trials will be searched: World Health Organization International Clinical Trials Registry Platform (www.who.int/ictrp), ClinicalTrials.gov (www.clinicaltrials.gov/), and the Chinese Clinical Trial Registry (ChiCTR). The search strategy for Medline will be provided (Appendix 2). We will use the equivalent search words in the Chinese databases as well.
Screening and selection
According to the inclusion criteria, two review authors (MNZ, LDF) will independently screen the titles and abstracts of all searched studies and exclude reports that are obviously irrelevant. They will retrieve full-text articles to identify studies for inclusion and record reasons for the exclusion of ineligible studies. Disagreements will be resolved through discussion or by consulting a third author (QQL). The study selection procedure is shown in Figure 1.
Data extraction
Two review authors (SWZ, PJ) will independently extract data and details from included studies using a preformulated data collection form. Extracted data will be compared by two review authors for completeness and accuracy and double-checked by another review author (PH) if necessary. Disagreements will be solved through discussion with QQL. The data extraction form will include the methods of the study, patients, interventions, duration, and relevant outcomes. If one or more included studies contain missing or unclear information, the authors will be contacted directly.
Assessment of risk of bias in included studies
The methodological quality of the eligible studies will be independently conducted by two reviewers (MNZ, LDF) for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions. Any disagreements will be solved by involving another review author (QQL). Risk of bias will be assessed according to the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other bias.
We will grade the risk of bias for each domain as high, low, or unclear, and provide information from the study report, together with a justification for our judgment, in the 'risk of bias' tables.
Data synthesis and management
1. Measures of treatment effect
For continuous data, we will use the mean difference to measure the treatment effect, with a 95% confidence interval (CI). Standardized mean difference with 95% CI will be used for when different scales were used to measure a certain outcome variable. For dichotomous data, we will use the risk ratio with 95% CI to measure treatment effects.
2. Dealing with missing data
For each included study, missing data and information will be gathered by contacting the study author directly. If the missing data are unavailable, we will perform intention-to-treat and sensitivity analyses to address the potential impact of missing data. If necessary, the potential impact will be described in the ‘Discussion’ section.
3. Assessment of heterogeneity and data synthesis
Whether a fixed-effect model or a random-effect model will be used depends on the results of the I² test for heterogeneity among the trials in each analysis. A meta-analysis will be conducted if the I² less than 75% among the included trials. However, a fixed-effect model will be applied to synthesize the data when the I² value is less than 25%. The sources of heterogeneity will be estimated by sensitive analysis or subgroup analysis when I² is between 25% and 75%. If the statistical heterogeneity is successfully explained (i.e., I² <25%), the fixed-effect model will be used to synthesize the data. Otherwise, we will use the random-effect model. If there is considerable heterogeneity (i.e., I²>75%), meta-analysis will not be performed but a systematic narrative synthesis will be conducted. All statistical analyses will be performed using Review Manager V.5.3 software.
4. Analysis of subgroups or subsets
Subgroup analyses will be conducted to determine the effects of various dosages, various courses of treatment, and various drug combinations if the data are available.
5. Assessment of reporting biases
If ten or more trials are included, we will use funnel plots to assess the presence of publication bias.
Confidence in cumulative evidence
On the basis of the Grading of Recommendations Assessment, Development, and Evaluation system29, the strength of evidence in this review will be categorized as high, moderate, low, or very low by the GRADEpro software.
Patient and public involvement
Not applicable. This protocol of systematic review and meta-analysis does not directly involve patients and the general public. Data will be collected from published articles retrieved from main databases and manual search.