Efficacy and safety of sugammadex for neuromuscular blockade reversal in pediatric patients: an updated meta-analysis of randomized controlled trials with trial sequential analysis

Two independent authors (BL and QZ) searched PubMed, Embase, Cochrane Library, and CNKI (China National Knowledge Infrastructure) databases up to April 24, 2021. Moreover, we considered potentially useful studies in Google Scholar as additional sources of information. The search terms we used included infant, child, adolescent, sugammadex, org 25,969, bridion and randomized controlled trial (Appendix S1). Only human studies were involved, and there were no restrictions of language.

The studies meeting the following conditions were selected for further analysis:

Two authors (BL and QZ) conducted the data extraction and identified quality and eligibility of studies. After removing the duplicates from different databases, those obviously irrelevant records were excluded by titles and abstracts screening. The full texts of the remaining studies were obtained and perused. To collect the general characteristics of enrolled studies, a table was designed and filled by us (Table 1). The risk of bias in RCTs was evaluated by the Cochrane risk of bias tool [20], using the following domains: random sequence generation (generation of the randomization sequence), allocation concealment, blinding of outcome assessment, incomplete outcome data, and selective reporting. All articles could have the following domain classifications: high risk of bias, low risk of bias, uncertain risk (without information for judgment). Any disagreements were resolved by consensus through discussion.

Assessment of quality of evidence and strength of recommendations was conducted by using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) methodology [39]. The quality of outcomes was independently assessed by two authors (BL and QZ). On the basis of risk of bias, inconsistency, indirectness, imprecision, and publication bias, the quality was classified as high, moderate, low, or very low. The GRADE profiler (version 3.6) software was used.

Statistical analyses were performed by using Review Manager software (Version 5.3.3, the Cochrane Collaboration 2014, the Nordic Cochrane Centre). Mean difference (MD) with 95% confidence interval (CI) were used to estimate continuous variables, and risk ratio (RR) with 95% confidence interval (CI) and the Mantel–Haenszel method (fixed or random models) were used to analyze dichotomous data. The I-squared (I²) test was chosen to weigh the impact of heterogeneity on the results. If significant heterogeneity (present at I² > 50%) existed, the sensitivity analysis was performed by omitting each study individually, and the random effects model was chosen; otherwise, the fixed-effects model was chosen. Publication bias were evaluated by using Begg's test and Egger’s test if the number of included studies exceeds 10. Evaluation was performed using version 1.2.4 of the metabias program, Stata/MP 12.0 for Windows (StataCorp LP, 4905 Lakeway Drive, College Station, TX 77,845, USA). A P value < 0.05 was considered statistically significant.

Sparse data and the repeated significance testing with new studies updating may lead to type-1 errors (false-positive outcomes) and type-2 errors (false-negative outcomes) of meta-analyses. To eliminate the risks from type-1 and type-2 errors, Trial sequential analysis (TSA), which can adjust the statistical threshold by controlling P value and widening confidence intervals, was performed by us. TSA can estimate the required information size (RIS) and trial sequential monitoring boundaries. The cumulative Z curve entering the futility area or crossing the trial sequential monitoring boundary may indicate that the present evidences of intervention effects are at a sufficient level, and further trials will be unnecessary. Otherwise, evidences are insufficient to draw the conclusion if Z curve does not cross any boundaries or reach the RIS [40]. And the TSA was performed using Trial Sequential Analysis Viewer Software (version 0.9.5.10 beta; http://​www.​ctu.​dk/​tsa).

After screening in databases and additional sources of information, a total of 187 relevant items were identified initially. 65 duplicate records were removed, and 96 records were excluded by titles and abstracts reviewing. In these 96 excluded items, 49 were studies conducted in adult patients, 20 were protocols or registered trials, 11 were reviews, 9 were irrelevant studies, 3 were conferences news, 2 were case reports or letters, 2 were previous systematic reviews published in 2016 and in 2017. And then 8 items were excluded by full-text screening, five of them reported the uncorrelated outcomes, and three of them were owing to the inappropriate comparisons. Eventually, 18 studies were chosen in consequent analysis [21‐38]. The process of literatures identification is described in PRISMA flowchart (Fig. 1).

The enrolled studies were published from 2009 to 2020, and a total of 1,065 eligible pediatric patients (ages ranged from 7 days to 18 years) were included in analysis. The outcome “time interval from administration of reversal agents to train-of-four ratio” was reported in 17 studies [21‐31, 33‐38], and the outcome “extubation time” was reported in 14 studies [23‐25, 28‐38]. 0.6 mg/kg rocuronium was given in all patients except patients in Veiga RG et al. study [22] (Rocuronium 0.45 mg/kg). And most of studies focused on evaluation in sugammadex versus combination of acetylcholinesterase inhibitors and anticholinergics, only two studies compared sugammadex with placebo. The main characteristics of all enrolled studies were summarized in Table 1.

We used Cochrane Collaboration’s risk of bias tool to evaluate the validity and quality of these enrolled studies [20]. In random sequence generation domain, 12 studies had low risk [21, 25, 27‐34, 36, 37], and 6 studies had unclear risk [22‐24, 26, 35, 38]. In allocation concealment domain, 6 studies had low risk of bias [21, 26, 29, 30, 32, 34], and 12 studies had unclear risk [22‐25, 27, 28, 31, 33, 35‐38]. Ten studies had low risk of bias [21, 24, 26, 28, 30‐34, 36] and rest of studies had unclear risk of bias [22, 23, 25, 27, 29, 35, 37, 38] in blinding of participants and personnel domain. One study had a high risk of bias [28], 9 studies had low risk of bias [21, 24, 26, 30‐34, 36], and 8 studies had unclear risk [22, 23, 25, 27, 29, 35, 37, 38] in blinding of outcome assessment domain. Sixteen studies had low risk of bias in incomplete outcome data [21, 23, 25‐38] and rest of studies had unclear risk of bias [22, 24]. In selective reporting domain, 16 studies had low risk [21, 23‐31, 33‐38], and two studies had unclear risk of bias [22, 32].

Seventeen studies including 995 pediatric patients described the time from NMB reversal to recovery of the TOF ratio to 0.9. The I² of 99% indicated that substantial heterogeneity was existed, but the source could not be attributed clearly to one particular study by sensitivity analysis; thus, the random effects model was used. According to present analysis with larger sample size, the use of sugammadex was associated with significantly shorter duration from administration of reversal agents to TOFr > 0.9 compared to traditional acetylcholinesterase inhibitors or placebo (MD -14.42 with 95% CI [-17.08, -11.75], P < 0.00001, I² = 99%) (Fig. 2A). Publication bias was detected in analysis by both Begg's test (P = 0.001) and Egger's test (P = 0.000) (Fig. 4A). In order to estimate and adjust for the number and outcomes of missing studies, we performed Duval's trim and fill method [41] by using version 1.0.5 of the metatrim program, Stata/MP 12.0 for Windows (StataCorp LP, 4905 Lakeway Drive, College Station, TX 77,845, USA). The trim-and-fill method showed no trimming performed and data unchanged. The information about trim and fill procedure was provided in Appendix S2. The outcome of TSA indicated that the cumulative Z curves crossed the conventional boundary, trial sequential monitoring boundary, and also the required information size (calculated as 358). It revealed that the sample size of patients was enough, and further studies would be unlikely to change the conclusion (Fig. 2B). According to GRADE summary of findings table, the quality of evidence for this outcome was low. It might be resulted from inconsistency (I² > 50%) and existed publication bias (Table S1).

A total of 14 studies involving 883 pediatric patients reported the duration from NMB reversal to extubation. By the same token, I² of 99% existed the significant heterogeneity. However, all attempts to reduce the value of I² to below 50% by excluding one single study were not successful in sensitivity analysis, therefore, the random effects model was used by us. The use of sugammadex was associated with shorter interval from reversal from NMB to extubation compared to acetylcholinesterase inhibitors or placebo (MD -13.98 with 95% CI [-16.70, -11.26], P < 0.00001, I² = 99%) (Fig. 3A). However, results from Begg's test (P = 0.002) and Egger’s test (P = 0.000) indicated that publication bias was existed in the analysis (Fig. 4B). Duval's trim and fill method was conducted, and results showed no trimming performed and data unchanged. The information about trim and fill procedure was provided in Appendix S2. The result from TSA indicated that with a required information size of 747, firm evidence was in place in favor of sugammadex (Fig. 3B). The GRADE summary of findings table indicated that quality of evidence for present outcome was low. Inconsistency (I² > 50%) and publication bias may be considered as main factors (Table S1).

Adverse effects including postoperative nausea and vomiting (PONV), bradycardia, pain, spasm, dry mouth, apnea, and oxygen desaturation were considered as our secondary outcomes. The results indicated that use of sugammadex was associated with significantly lower incidence of PONV (RR = 0.30; 95%CI: 0.20 to 0.46), bradycardia (RR = 0.09; 95%CI: 0.02 to 0.46), and dry mouth (RR = 0.14; 95%CI: 0.05 to 0.38) compared to acetylcholinesterase inhibitors or placebo. For other adverse effects, no significant differences were found between the two groups. The results of publication bias were (P = 0.088, Begg’s test and P = 0.004, Egger’s test) (Fig. 4C), however, the trim-and-fill method to adjust for funnel plot asymmetry showed no trimming performed and data unchanged. Owing to absence of statistical heterogeneity (I² < 50%) in secondary outcomes, the fixed-effects model was used. The details of secondary outcomes were demonstrated in Table 2. The results from GRADE summary of table revealed that quality of evidence for most of secondary outcomes was low and imprecision (lack of events number) was served as the main reason. The details were provided in Table S1.