This review was reported according to the criteria of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Additional file 1) [17]. The protocol for this review was registered with the International Prospective Register of Systematic Reviews (PROSPERO—CRD42021252156).
Search strategy
We searched for databases including PubMed, Medline, Embase, the Cochrane Library, Ovid, Web of Science, and websites including Clinicaltrials.gov, WHO International Clinical Trials Registry Platform (ICTRP), and Google Scholar till February 1, 2021. The following terms were used: (IPACK OR “interspace between the popliteal artery and posterior capsule of the knee”) AND (total knee arthroplasty OR knee arthroplasty OR total knee replacement OR knee replacement OR TKA OR TKR) AND ((randomize* control* trial*) OR RCT)). No language or date limits were placed on the search. We also used a manual search strategy, checked references, and contacted authors to identify additional studies. Two authors screened studies with a third author adjudicating in case of disagreement.
Trial selection
The studies had to be RCTs comparing TKA patients with IPACK. Any non-RCTs, quasi-RCTs, retrospective studies, cadaver studies, comments, letters, editorials, protocols, guidelines, surgical registries, and review papers were excluded. Follow-up reports at different time points or different comparisons in one trial will be extracted separately. Studies with multiple arms were eligible, as were studies in which multiple regional anesthetic techniques were performed, so long as an IPACK was one of the arms or one of the used techniques. There was no restriction on language or publishing year. Two investigators independently screened titles and abstracts to exclude non-relevant trials. Discrepancies were resolved by a third author. Relevant full-text articles were retrieved and analyzed for eligibility using the pre-defined inclusion criteria.
Data extraction
Data were extracted via a standardized spreadsheet according to a pre-agreed protocol. The following information was collected: first author, publication year, country, number of participants in each group, patient demographics, inclusion and exclusion criteria, and conclusions. We collected: interventions, dosages, and types of anesthesia drug administered, the method of anesthesia, pain rescue methods, multimodal analgesia protocol, surgeons, prothesis, approach, follow-up duration, and numbers of patients lost to follow. If data cannot be extracted directly or missing, we will contact the authors by email or calculate data with the Cochrane Review Manager calculator [18]. Two authors independently extracted the information, and any discrepancies were resolved by a third author. Pain scores reported on visual, verbal, or numerical rating scales were converted to a standardized 0–10 scale. All opioids were converted to oral milligram morphine equivalents via an online website (http://opioidcalculator.practicalpainmanagement.com/).
Outcomes
The primary outcome was the ambulation pain score. The secondary outcomes were rest pain score, morphine consumption, functional recovery outcomes, clinical outcomes, and complications. The morphine consumption was collected as a continuous variable (amount) and category variable (used or not). The functional recovery outcomes included the range of motion (ROM), quadriceps muscle strength (QMS), ambulation distances, and time-up-and-go test (TUG) time. The clinical outcomes included the length of hospital stay, operation time, and patient satisfaction. The complications were postoperative nausea and vomiting (PONV) and sleep disturbance.
Subgroup analyses
Our pre-defined subgroup analysis was based on multiple time points. The subgroups were as closest to 6, to 12, to 24, to 48 h and beyond one week or as the postoperative day (POD) 0, 1, and 2 described in original studies.
Trial sequential analysis
We performed Trial Sequential Analysis (TSA) using the TSA program (www.ctu.dk/tsa.) on the three critical outcomes (pain at rest, pain at ambulation, morphine consumption). TSA tests the credibility of the results by combining the estimation of information size (a cumulative sample size of included RCTs) with an adjusted threshold of statistical significance for the cumulative meta-analysis. The required information size (RIS) and meta-analysis monitoring boundaries (Trial Sequential Monitoring Boundaries) were quantified, alongside adjusted 95% confidence intervals. Diversity adjustment was performed according to an overall type I error of 5% and power of 80%.
Meta-regression
High heterogeneity not fully explained by subgroup analysis was further investigated with a post hoc mixed-model meta-regression on the primary outcome (pain at ambulation). To avoid overfitting, meta-regression was performed only in the following clinically meaningful and explanatory variables: patient number, the multimodal analgesia protocol, types of other nerve blocks, anesthesia drug.
Risk of bias assessment and publication bias
The methodology quality was independently evaluated by two reviewers using the Cochrane Collaboration’s Risk of Bias Tool [19]. The following domains were assessed and evaluated: randomization process, deviation from intended interventions, missing outcome data, measurement of outcomes, and selection of reported results. Each domain can be judged as low risk of bias, high risk of bias, or unclear, and overall risk of bias is expressed on a three-grade scale (low risk of bias, high risk of bias or unclear).
The funnel plots were used to assess publication bias when the included studies were more than 10 in the outcome, and the Egger test was further performed (when visual asymmetry was observed).
Quality of evidence
We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system to assess the certainty of the evidence in key outcomes. Study design, risk of bias, imprecision, inconsistency, indirectness, and magnitude of effect were considered. The level of evidence could be divided into four degrees: high, moderate, low, and very low. The rules for downgrade evidence were referenced in Guyatt’s studies [20,21,22,23,24,25]. We defined the following as critical outcomes: pain at ambulation, pain at rest, morphine consumption amount, the rate of rescue morphine use.
Statistical analysis
Weight mean difference (WMD) for continuous variables (Mantel–Haenszel method) and risk ratios (RR) for dichotomous variables (inverse variance method) with 95% confidence intervals (95% CIs) were used. P values of < 0.05 were considered statistically significant. A random-effect model was used in the study. The heterogeneity was reported by I2 statistics. (I2 > 70% was considered as high heterogeneity.) Sensitivity analysis will be applied to examine the effect of deleting one single study on the overall estimate when observed high heterogeneity, and Publication bias was evaluated both by a visual inspection of funnel plots and by Egger test (p < 0.05 indicating a possible publication bias) using Egger’s regression intercept to quantify publication bias. The Review Manager 5.3 was used for drafting figures of risk of bias, and STATA 13.0 was used for data analysis.