Study registration
This systematic review was registered on PROSPERO (CRD42020168004) on 5 February 2020. We have prepared this protocol in accordance with the Preferred Reporting Item for Systematic Review and Meta-analysis (PRISMA-P) statement [25] (Additional file 1) and we will update the PROSPERO record if there is any important amendments.
Eligibility criteria
Inclusion criteria
Type of studies
Randomized trials and quasi-randomized or prospective controlled clinical trials that have tested traditional Chinese herbal medicine with or without western medicine for COVID-19 will be included. There will be no restrictions for blinding, follow-up or publication status. Publications in English and Chinese will be included.
Type of participant
Patients diagnosed with pneumonia caused by COVID-19 without immediately life-threatening co-morbidities will be included. There will be no restrictions with respect to gender, age or ethnicity.
Type of interventions
Traditional Chinese herbal medicine involving extracts from herbs, single or mixture herbal formulas regardless of their compositions or forms. traditional Chinese herbal medicine combined with one or more other pharmacological intervention will also be included. There will be no restrictions with respect to dosage, frequency, duration or follow-up time of treatment.
Type of comparators
There will be no restrictions with respect to the type of comparator. The comparators are likely to include western medical therapies, supportive care and other therapeutic methods.
Type of outcome measurements
Our primary outcomes will be survival at the end of treatment and at the end of follow-up, and time and rate of the patient becoming negative for the COVID-19. We will also assess the following outcomes at the end of treatment and at the end of follow up: days to absence of fever; symptom score (based on fever, fatigue, cough, difficulty in breathing, poor appetite, etc.); duration of each symptom; pulmonary function; inflammation index; results of chest computerized tomography; length of stay in hospital; use (including dosage and duration) of corticosteroid; quality of life; and adverse events. If other outcomes are reported in the eligible studies, these will be extracted and reported but we will give particular attention to the possibility of selective reporting bias when using any such outcomes in our review.
Exclusion criteria
(1) Patients with life-threatening co-morbidities likely to lead to death within the trial follow-up period; (2) Duplicated data or data that cannot be extracted after contacting original authors; (3) Full text cannot be obtained after contacting original authors.
Databases and search strategy
We will search electronic databases including PubMed, Embase, the Cochrane Central Register of Controlled Trials (CENTRAL), Chinese Biomedical Literature Database (CBM), China National Knowledge Infrastructure (CNKI), Chinese Science and Technology Periodical Database (VIP), and Wanfang database (Wanfang Data) using keywords combination, such as novel coronavirus OR COVID-19 OR 2019-nCoV OR COVID-2019 pneumonia AND traditional Chinese herbal medicine OR Chinese herb OR traditional Chinese medicine. The full search strategy for PubMed is provided in Additional file 2 and similar strategies will be applied to the other electronic databases. Reference lists of relevant trials and reviews will be searched. We will manually search grey literature such as conference proceedings and academic degree dissertations, and trial registries (both through the WHO International Clinical Trials Registry Platform (ICRP) and on the websites of national registries). We will consult experts in the field for possible studies to be included.
Study selection
We will export the identified records in databases into EndNote X9 software, and use this to identify duplicates. After removing duplicates, the retrieved records will be checked independently by two reviewers (XL and DZ), who will apply the eligibility criteria based on the title and abstract. Where a study is potentially eligible, the full-text will be obtained and checked independently by two reviewers (XL and DZ) to identify the eligible studies. Any disagreements will be discussed and resolved in discussion with a third reviewer (JL).
Data extraction
In order to achieve a consistency (at least 80%) of extracted items, the data extractors will extract data from a sample of eligible studies. Results of the pilot extraction will be discussed among review authors and extractors. Two independent reviewers (YL and LG) will extract data with a predefined extraction template, which includes the following items: (1) General information: first author, title, journal, year of publication, country, funding source, study design, etc. (2) Characteristics of patients: age, gender, stage and severity of disease, syndrome differentiation, comorbidity, etc. (3) Characteristics of intervention: protocol of Chinese herbal medicine (types, dosage, frequency, duration etc.), protocol of comparators (types, dosage, frequency, duration etc.). (4) Characteristics of trial: study setting (ambulatory sector/ hospital), sample size (numbers recruited, randomized or allocated to the interventions by another method, followed up and analyzed), generation of randomization sequence, allocation concealment, blinding, studies’ length of follow-up etc. (5) Outcomes: all outcomes, main conclusions, adverse events, etc. The original authors will be contacted to request missing data where necessary. Extracted information will be cross checked by YL and LG. Any disagreements will be discussed and resolved in discussion with a third reviewer (YZ).
Assessment of risk of bias
In order to achieve a consistency (at least 80%) of risk of bias assessment, the risk of bias assessors will pre-assess a sample of eligible studies. Results of the pilot risk of bias will be discussed among review authors and assessors. Two independent reviewers (YL and DZ) will assess the risk of bias of the included studies at study level. We will follow the guidance in the latest version of Cochrane Handbook for systematic reviews of interventions[26] when choosing and using tools to assessing risk of bias for randomized trials (version 2 of the Cochrane risk-of-bias tool for randomized trials, RoB 2[27]) and non-randomized trials (The Risk Of Bias In Non-randomized Studies of Interventions, ROBINS-I tool[28]). Any disagreements will be discussed and resolved in discussion with a third reviewer (RJ). Studies with high risk of bias or unclear bias will be given less weight in our data synthesis.
Data analysis
Statistical analyses will be conducted using RevMan software (version 5.3.5) and R software (version 3.6.1). If possible, analyses for all outcomes will be done by intention-to-treat. We will perform analyses to provide effect estimates for dichotomous data and continuous data, with 95% confidence intervals. We will use risk ratios (RR) for dichotomous data and mean differences (MD) for continuous data. We will explore the heterogeneity before we perform meta-analysis for outcomes. Heterogeneity will be detected by using a standard Chi-square test with a significance level of P < 0.10. The I2 statistic will be applied to quantify inconsistency across studies and to assess the impact of heterogeneity on the meta-analyses. Mantel-Haenszel method will be used for dichotomous outcomes, and DerSimonian and Laird inverse variance method will be used for continuous outcomes. Random-effects model will be used to pool the data.
Subgroup analysis
If an adequate number of studies are identified, we will perform subgroup analysis for the following variables: age; patients with or without other diseases and COVID-19 stage at which the traditional Chinese herbal medicine was given.
We will also consider analyses for other subgroups as reported in the included studies, but we will give particular attention to the possibility of selective reporting bias when using any such subgroups in our review.
Trial sequential analysis
Trial sequential analysis provides the necessary sample size for our meta-analysis and boundaries that determine whether the evidence in our meta-analysis is reliable and conclusive[29, 30]. We will perform a trial sequential analysis to maintain an overall 5% risk of type-1 error and calculate the required sample size.
Sensitivity analysis
To check the robustness of pooled outcome results, we will carry out sensitivity analysis to explore the influence of studies with high risk of bias.
Publication bias
If sufficient number of articles are included, we will assess small study biases (e.g. publication bias) with funnel plots, the Egger 's test and Begg 's test, and Trim and Fill analysis.
Quality of evidence
Two independent reviewers (DLZ and JL) will assess the quality of evidence for each outcome with the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system[31]. Each outcome will be assessed for each of the five aspects: limitations, inconsistency, indirectness, imprecision, and publication bias. They will be rated as high, moderate, low, or very low level.