The PICOS (patient, intervention, comparison, outcome, and study design) framework will be used to inform the eligibility criteria of studies [21,22]. Studies will be excluded if they meet any of the following criteria: (1) duplicate studies; (2) nonrandomized controlled trials; (3) literature reviews; (4) case reports; (5) studies comparing different types of Tui na; and (6) animal experiments.

Any randomized controlled trial (RCT) involving Tui na therapy in the treatment group for patients with CNLBP will be included in this study.

Adults (aged 18 years and older) who were diagnosed with chronic, nonspecific (with no diagnosable underlying pathology) LBP will be included in this study. There are no limits on race, gender, age, nationality, etc.

All trials evaluating Tui na intervention will be included. Experimental intervention can be any type of Tui na or Tui na combined with other therapies.

Different control interventions such as no-treatment control, placebo, and other currently used interventions (such as health education, behavior therapy, acupuncture, moxibustion, physical therapies, medications, and gentle touch) will be included.

Primary outcomes will include improvement of pain, reduction in analgesic medication, improvement of functional disability, and degree of satisfaction with the intervention. A secondary outcome will be any adverse events (AEs) associated with Tui na.

An electronic literature search will be applied to the following databases: PubMed, Embase, MEDLINE, Cochrane Library, Springer, Scopus, World Health Organization International Clinical Trials Registry Platform, Physiotherapy Evidence Database (PEDro), Clarivate Analytics, and Chinese biomedical databases (the China National Knowledge Infrastructure, Wan-fang database, Chinese Scientific Journals Database, and Chinese Biomedical Literature Databases), from database inception to December 2020, limited to English and Chinese language only. Peer-reviewed journal articles and reference lists of final included studies, as well as grey literature including conference proceedings, dissertations, and white papers, will be used to help identify all applicable studies for inclusion.

The comprehensive literature search will be designed and carried out by an experienced librarian at Mayo Clinic, Rochester campus, with input from the study’s principal investigator. Controlled vocabulary and keywords will be used to search for publications describing the effect and safety of Tui na on patients with CNLBP.

The results of the literature search, including the abstract and citation, will be imported into EndNote X9 (Thomson Reuters Clarivate Analytics), and duplicate studies will be removed prior to the literature screening. Two authors will screen titles and abstracts on inclusion criteria independently. Publications that are not relevant or do not meet the inclusion criteria will be removed. The whole selection process will be presented in the PRISMA flowchart of Figure 1. Any selection divergence will be resolved by the consensus between the two authors or by consulting the original corresponding author.

To ensure reliability, the risk of bias of each study included in our study will be assessed independently by two authors using the special assessment tool recommended by the Cochrane Collaboration [23], including selection, performance, attrition, detection, reporting, and other biases. An assessment disagreement between the two authors will be discussed between the two authors or by involving a third reviewer until a consensus is reached. Any discrepancies will be resolved through discussion or by involving a third reviewer.

Important data associated with our study will be extracted from the included studies using a designed and piloted data collection Excel spreadsheet (Microsoft Corp) by two authors independently. Data will include characteristics of the studies including the first author, published year, sample size, population, and outcome measurement, and detailed information about the interventions such as observation group, control group, provider, frequency, efficacy, AEs, and follow-up.

The AE severity is evaluated with the Common Terminology Criteria for Adverse Events (CTCAE) [24]. The CTCAE Scale uses Grades 1 through 5 as follows: Grade 1 (mild AE), Grade 2 (moderate AE), Grade 3 (severe AE), Grade 4 (life-threatening or disabling AE), and grade 5 (death AE). Disagreements will be resolved by discussions. If necessary, the corresponding author will be contacted for clarification.

Funnel plots will visually reveal the publication bias if the number of included trials for data analysis is sufficient (a minimum of 10 trials). Egger and Begg tests will be carried out to check the asymmetry of the funnel plot. A symmetric funnel plot represents a low risk of reporting bias, while a dissymmetric funnel plot represents a high risk.

Descriptive analysis or meta-analysis will be conducted according to the interventions, the measurements, and heterogeneity levels. Quantitative findings will be descriptively reported. Continuous variables are analyzed using mean difference and 95% CI, while dichotomous variables will be analyzed using odds ratio. If outcome measure scales are different, the standardized mean difference and 95% CI will be calculated. Study heterogeneity will be calculated within each pairwise comparison by Q test and I² statistic; higher values indicate higher heterogeneity. I²=0% indicates no heterogeneity; I²<25% indicates low, 25%<I²<50% indicates moderate, while 50%<I²<75% indicates substantial, and 75%<I²<100% indicates considerable heterogeneity [25]. If at least 2 included trials are sufficiently homogenous in terms of study design, comparator, and outcome measurement, a meta-analysis will be performed using a random-effects model. Review Manager software (RevMan 5.3) provided by the Cochrane Collaboration will be applied for the meta-analysis. If substantial heterogeneity between studies is present and there are sufficient RCTs, subgroup analyses will be performed to explain clinical heterogeneity effects on study intervention, pain intensity, and measurement tool.

The proposed review will only synthesize previous publications. No research ethics committee review approval is required. Results from this study will be disseminated as a peer-reviewed journal article and presented in conferences.