We performed this meta-analysis in accordance with the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) [21].
Literature Search
Published RCTs that evaluated the effects of IA Mg on postoperative pain management after arthroscopic knee surgeries were searched in the PubMed, Medline, Embase, Web of Science, and Cochrane library electronic databases inclusive until September 30, 2019. The words and MESH terms "Intra-Articular", "Magnesium", "Arthroscopy", "Postoperative", and “Pain” were searched individually and in different combinations. A manual search of references from eligible and relevant studies was performed to find additional trials. No restrictions were imposed regarding language or publication status.
Inclusion and Exclusion Criteria
Eligible studies were required to meet the following inclusion criteria: (1) RCTs; (2) patients undergoing arthroscopic knee surgery; (3) administration of Mg through the IA route; (4) including an experimental group of IA Mg or IA Mg plus a local anesthetic; and (5) including a control group of saline or local anesthetic alone. The exclusion criteria were: (1) non-RCTs; (2) reviews, letters, abstracts, case series, or editorials; (3) the administration of Mg not through the IA route; and (4) studies with insufficient data.
Study Selection
Two authors (Lijun Shi and Jinhui Ma) independently assessed the initial search results to exclude irrelevant trials and identify eligible studies according to the inclusion and exclusion criteria by screening titles and abstracts. Full-texts of any potentially useful studies were reviewed. Any discrepancies were resolved by consulting with a third author (Wei Sun).
Data Abstraction
Two authors (Lili Shi and Tengqi Li) independently evaluated the included studies and extracted trial details using special data collection forms developed for this investigation. Disagreements were resolved by consensus or consultation with a third author.
We first extracted data from tables or text. For data not reported numerically, we extracted them from available figures using the software GetData (http://getdata-graph-digitizer.com/index.php). Continuous data were reported using means and standard deviations (SD), and data presented in terms of the median and range were converted to means and SD [22]. For trials that involved more than one experimental group in comparison with a single control group, the relevant comparisons to the comparator were split for primary analysis.
The data extracted from trials included the first author, year of publication, sample size, patient baseline characteristics, type of surgery, type of anesthesia, IA Mg dose, pain scores at rest and with movement (postoperative 2, 4, 12 and 24 h), cumulative opioid consumption, time to first rescue analgesic request (min), and adverse events. The pain intensity was measured using the 10-point visual analogue scale (VAS), where 0 means no pain and 10 means the most severe pain. The numerical rating scale (NRS) of pain was converted to a VAS score. Postoperative opioid consumption within 24 h was converted to the equivalent dosage of intravenous (IV) morphine [23].
The primary outcomes of interest were the pain VAS scores at rest and with movement at different postoperative time points and total opioid consumption (IV morphine equivalent, mg) in the first 24-h postoperative period. The secondary outcomes included the time to first analgesic requirement (min) and the incidence of side effects.
Assessments of the Risk of Bias and Methodological Quality
Two senior authors (Fuqiang Gao and Wei Sun) independently evaluated the methodological quality of the included studies using the Cochrane Collaboration’s Risk of bias tool [24], which contains seven domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other sources of bias. The risk of bias was defined as high, low, and unclear. Disagreements were resolved by discussion.
The quality of evidence for each outcome was judged with the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology [25], which consists of five items: study limitations, inconsistency of results, indirectness of evidence, imprecision, and reporting bias. This methodology categorizes the strength of evidence as high, moderate, low, or very low, and each of these items may be used to define the quality level. This process was conducted using Grade Profiler software (GRADEpro version 3.6).
Statistical Analysis
All statistical analyses were conducted using Review Manager software (RevMan version 5.3). Continuous variables are reported as mean differences (MD) with 99% confidence intervals (CIs). As the incidence of adverse events was very low, only qualitative analysis and description was performed. Statistical heterogeneity was measured and reported as I2, which describes the percentage of the total variability caused by heterogeneity rather than by chance. The I2 values ranged between 0% and 100%, where values above 50% and 75% represent substantial and considerable heterogeneity, respectively. If the heterogeneity was significant (p < 0.05, I2 > 50%), the random-effects model was used. Otherwise, the fixed effects model was adopted (p > 0.05, I2 < 50%). Sensitivity analysis was further performed by removing one trial at a time to explore possible explanations for heterogeneity and to identify the influence of a single RCT on the overall mean differences.