The safety and efficiency of intravenous administration of tranexamic acid in coronary artery bypass grafting (CABG): a meta-analysis of 28 randomized controlled trials

Background The safety and efficiency of intravenous administration of tranexamic acid (TXA) in coronary artery bypass grafting (CABG) remains unconfirmed. Therefore, we conducted a meta-analysis on this topic. Methods We searched the Cochrane Central Register of Controlled Trials (CENTRAL), PUBMED and EMBASE for randomized controlled trials on the topic. The results of this work are synthetized and reported in accordance with the PRISMA statement. Results Twenty-eight studies met our inclusion criteria. TXA reduced the incidence of postoperative reoperation of bleeding (relative risk [RR], 0.46; 95% confidence interval [CI]; 0.31–0.68), the frequency of any allogeneic transfusion (RR, 0.64; 95% CI, 0.52–0.78) and the postoperative chest tube drainage in the first 24 h by 206 ml (95% CI − 248.23 to − 164.15). TXA did not significantly affect the incidence of postoperative cerebrovascular accident (RR, 0.93; 95%CI, 0.62–1.39), mortality (RR, 0.82; 95%CI, 0.53–1.28), myocardial infarction (RR, 0.90; 95%CI, 0.78–1.05), acute renal insufficiency (RR, 1.01; 95%CI, 0.77–1.32). However, it may increase the incidence of postoperative seizures (RR, 6.67; 95%CI, 1.77–25.20). Moreover, the subgroup analyses in on-pump and off-pump CABG, the sensitivity analyses in trials randomized more than 99 participants and sensitivity analyses that excluded the study with the largest number of participants further strengthened the above results. Conclusions TXA is effective to reduce reoperation for bleeding, blood loss and the need for allogeneic blood products in patients undergoing CABG without increasing prothrombotic complication. However, it may increase the risk of postoperative seizures. Electronic supplementary material The online version of this article (10.1186/s12871-019-0761-3) contains supplementary material, which is available to authorized users.


Background
Excessive bleeding is a common complication which may lead to exposure to the risk of homologous blood transfusion and increased morbidity in patients undergoing cardiac operations [1]. Tranexamic acid (TXA), an antifibrinolytic agent, has been widely used and proved to be effective in reducing risk of blood loss and transfusion among patients undergoing cardiac surgery [2]. However, whether it reduced the incidence of reoperation for life-threatening bleeding which are strongly associated with poor outcomes after cardiac surgery remains controversial.
Despite of the effectiveness in reducing the risk of blood loss and transfusion, it may potentially increase the risk of myocardial infarction, stroke, and other thrombotic complications after cardiac surgery especially in patients undergoing coronary artery bypass grafting (CABG) surgery who are commonly characterized by systemic arteriosclerosis or stenosis [3,4]. It was reported that TXA was associated with the increased risk of postoperative neurologic events such as stroke and seizures in cardiac surgery [5,6]. Some studies have suggested that TXA is associated with reduction in cerebral blood flow and increase the risk of cerebral infarction [5,7]. A multi-center study suggested that TXA was associated with a higher risk of postoperative seizures in GABG surgery [8]. A meta-analysis in 2011 has shown that TXA is associated with reduced blood transfusion in off-pump CABG surgery [9]. However, the safety of TXA in off-pump CABG surgery could not be confirmed due to the small population sample size.
An increasing number of studies that investigated the effectiveness and safety of TXA in CABG surgery have been conducted in recent years with varying results [8,[10][11][12][13][14][15][16][17][18]. Therefore, we conducted a meta-analysis of existing studies to estimate the safety and efficiency of TXA in CABG surgery focusing on the incidence of postoperative cerebrovascular accident, seizures and reoperation for bleeding.

Methods
The meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement in this study [19].

Search strategy
A systematic and comprehensive search was conducted in the Cochrane Central Register of Controlled Trials (CENTRAL), PUBMED and EMBASE from database established to February 8, 2018 with no language limitation. The search strategy included the following MED-LINE subject heading terms: tranexamic acid and cardiac surgical procedures. The above subject heading terms were connected by "AND". The initial searches of PUBMED and EMBASE were unrestricted to maximize sensitivity and a filter which primarily identifies randomized controlled trials was adopted to improve the specificity. Moreover, we also checked the reference lists of relevant articles for potential relevant studies.

Eligibility criteria
Randomized controlled trials that compared the effectiveness or safety of the intravenous administration of TXA with that of placebo in adult CABG surgery were included in this meta-analysis. Studies were eligible for inclusion, regardless of the publication language. We excluded studies which were conducted on underage patients or in which TXA was topically applied in mediastinum.

Selection of included studies
Retrieved studies were imported into Endnote (version X7; Thomson Reuters), where duplications were detected and deleted automatically. Two authors independently scanned the titles and abstract of retrieved studies according to the established eligibility criteria to exclude the obvious unrelated studies. The full-text was further evaluated if the judgement could not easily be decided based on its title or abstract. The disagreements between reviewers were settled by a third reviewer. The relevant data of included studies was extracted by these reviewers independently using a standard data sheet. Study characteristics included author, publication year, sample size, sex ratio, type of CABG, duration of anticoagulant medication discontinued before surgery, outcome data, drug dose and treatment regimens.

Assessment of risk of bias in included studies
The Cochrane risk of bias tool which is recommended by the Cochrane Collaboration for risk of bias assessment was adopted in this study [20]. There are seven domains in the Cochrane risk of bias tool, including the random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other bias. The judgment of each domain is presented as "low risk", "high risk" or "unclear risk" based on the instruction of Cochrane Collaboration. Two reviewers independently assessed each domain of included studies and any disagreements were adjudicated by a third reviewer.
Quality of the evidence GRADE (Grades of Recommendation, Assessment, Development and Evaluation) Working Group system was adopted to evaluate the quality of the evidence [21]. Two reviewers independently assessed the quality of each outcome. The five categories used for the GRADE quality assessment were: limitations of design, inconsistency, indirectness, imprecision, and publication bias. We used GRADE profiler (GRADEpro) software to create the "Summary of findings" table, which includes the following outcomes: incidence of postoperative cerebrovascular accident, seizures, reoperation for bleeding, mortality, myocardial infarction, acute renal insufficiency, the frequency of any allogeneic transfusions and 24-h postoperative chest tube drainage.

Study outcomes
All outcomes were described a priori, according to the principles of the PRISMA statement. The primary outcome was incidence of postoperative cerebrovascular accident, seizures and reoperation for bleeding. The second outcomes included postoperative mortality, myocardial infarction, acute renal insufficiency, the frequency of any allogeneic transfusions and 24-h postoperative chest tube drainage.

Statistical methods
In some studies, continuous variables was presented as median, range and/or interquartile range. To facilitate meta-analysis, we estimated the sample mean and standard deviation from median, range and/or interquartile range by using the calculator with a compiled formula recommended by Luo and colleagues [22]. The risk ratio (RR) with the corresponding 95% confidence interval (95% CI) was calculated for dichotomous data and continuous data were analyzed by using mean difference (MD) with the corresponding 95% CI. Data analyses followed the guidelines established by the Cochrane Collaboration regarding statistical methods. The statistical heterogeneity was evaluated by reviewing the I 2 statistic and Chi 2 test. If either the Chi 2 test resulted in P < 0.10 or the I 2 statistic was greater 50%, random-effect model was used to evaluate outcomes, otherwise a fixed-effect model was used. For all tests, two-tailed P-values < 0.05 were considered significant. Funnel plots were conducted to evaluate reports for publication bias when more than 10 studies were included. Considering the activation effect of cardiopulmonary bypass (CPB) on the fibrinolytic pathway, subgroup analysis was performed based on CABG with/without CPB. Moreover, Sensitivity analyses was performed in studies randomized more than 99 patients to avoid the possibility that the rare incidences of complication were underestimated due to the included studies with small population size.
Sensitivity analyses that excluded the study with the largest number of participants were conducted to estimate the effect of that study on the overall effect of metaanalysis. All data analysis was conducted using Review Manager (RevMan; version 5.2), Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2012.

Results of search
Two hundred twenty-seven studies were identified from our initial search and 146 of them remained after duplicates were removed. One hundred eight of the remaining studies were excluded during title and abstract screening. Thirty-eight studies were identified for full-text assessment according to our inclusion and exclusion criteria and 10 of them were removed because of non-RCT, topical application of TXA or without placebo group. Finally, 28 studies [3, 4, 8, 10-12, 14-18, 23-39] were included in this meta-analysis. The study selection process is shown in Fig. 1.

Risk of bias within studies
The results of bias risk assessment were showed in Fig. 2a and b. Fourteen studies [3, 11, 12, 14, 16, 23-25, 27-30, 35, 38] did not provide a satisfactory description of their random processes. Blinding process was at high risk of bias in one study [39] and unclear risk of bias in 7 studies [11, 12, 23-25, 29, 35] due to unclear description. Three studies [16,25,31] had unclear or incomplete descriptions of their outcome data. Two studies [3,36] were considered to be at high risk of selective reporting bias because the reported outcome indicators were inconsistent with the planed outcome indicators.

Publication bias
Publication bias was evaluated by funnel plots in the following outcomes: postoperative cerebrovascular accident, reoperation for bleeding, mortality, myocardial infarction, acute renal insufficiency, the frequency of any allogeneic transfusions and 24-h postoperative chest tube drainage (Additional file 1: Figure S1, Additional file 2: Figure S2, Additional file 3: Figure S3, Additional file 4: Figure S4, Additional file 5: Figure S5 and Additional file 6: Figure S6 and Additonal file 7: Figure  S7). All of the plots showed a symmetrical shape which suggested low risk of publication bias of the above outcomes.
Nine studies with a total of 5939 participants were included in the sensitive analysis of studies that randomized not less 100 participants. The conclusion that TXA would not increase cerebrovascular accident incidence was strengthened by the sensitivity analysis [RR = 0.87(0.57-1.33), P for effect = 0.53, P for heterogeneity = 0.95, I 2 = 0%]. Sensitivity analysis that excluded the study with the largest number of participants furether strengthened the above conclusion [RR = 0.95(0.43-2.10), P for effect = 0.90, P for heterogeneity = 0.86] ( Table 2).

Seizures
In total, 5 studies with 5043 participants reported the incidence of seizures after CABG. The summary RR for postoperative seizures with the use of TXA versus Fig. 3 Forest plot of cerebrovascular accident placebo was 5.99 (95% CI 1.77-20.24) which suggested that tranexamic acid would increase the incidence of seizures after CABG (Fig. 4).
Ten studies with 1143 participants were included in on-pump CABG, the result of meta-analysis suggested no significant difference of reoperation for postoperative bleeding between TXA and placebo [16/569 vs 26/574, RR = 0.64 (0.35-1.15), P for effect = 0.14, P for heterogeneity = 0.62, I 2 = 0%]. In off-pump subgroup, 4 studies  with 384 participants were included and only one patient suffered reoperation in placebo group (Fig. 5). Eight trials were included in sensitivity analysis of studies randomized not less than100 patients. The sensitivity analysis supported the result that TXA decreased incidence of reoperation for bleeding in CABG surgery when compared with placebo [29/2812 vs 59/2821, RR = 0.49 (0.32-0.77), P for effect< 0.01, P for heterogeneity = 0.58, I 2 = 0%]. While sensitivity analysis that excluded the study with the largest number of participants did not supported the above conclusion [RR = 0.59 (0.34-1.04), P for effect = 0.07, P for heterogeneity = 0.72] ( Table 2).

Mortality
The overall analysis showed that TXA did not significantly decrease the mortality in patients receiving CABG when compared with placebo [33/3196 deaths in the TXA group vs 41/3218 deaths in the placebo group, RR = 0.82(0.53-1.28), P for effect = 0.38, P for heterogeneity = 0.82, I 2 = 0%, with 18 trails included] (Fig. 6).
Sub-analysis in the settings of on-pump CABG also showed no statistically significant effect of TXA on mortality [6/639 vs 7/663, RR = 0.93 (0.36-2.38), P for effect = 0.88, P for heterogeneity = 0.62, I 2 = 0%, with 12 trials included]. Sub-analysis in the settings of off-pump included 5 trials, but only one of them reported one patient died in each group (Fig. 6).
Sensitivity analysis of studies randomized more than 99 patients supported the results that TXA did not significantly decrease the mortality in CABG surgery compared with placebo [31/2870 vs 36/2886, RR = 0.87 (0.54-1.40), P for effect = 0.56, P for heterogeneity = 0.46, I 2 = 0%, with 7 trials included]. The result of sensitivity analysis that excluded the study with maximum  Table 2).
Seven trials were included in sensitivity analysis of studies randomized not less than100 patients. The sensitivity analysis supported the result that TXA did not increase myocardial infarction in CABG surgery when compared with placebo [286/2999 vs 318/3011, RR = 0.90 (0.78-1.05), P for effect = 0.18, P for heterogeneity = 0.64, I 2 = 0%]. The result of sensitivity analysis that excluded the study with maximum sample was consistent  Table 2).

Acute renal insufficiency
There are 14 studies that reported the incidence of acute renal insufficiency in this meta-analysis. The summary RR for acute renal with the use of TXA versus placebo was 1.01 (95% CI 0.77-1.32) which suggested that tranexamic acid would not increase the incidence of acute renal insufficiency (Fig. 8).
The summary RR of sub-analysis in on-pump CABG was 0.91 (95% CI 0.36-2.29) which suggested that TXA did not have adverse effect on postoperative renal function in patients undergoing on-pump CABG. A similar result was found in the sub-analysis in off-pump CABG [RR = 0.85 (0.29-2.47), P for effect = 0.76, P for heterogeneity = 0.52, I 2 = 0%] (Fig. 8).
Sensitivity analysis in trials randomized not less than100 participants reinforced the overall analysis [RR = 1.03 (0.79-1.35), P for effect = 0.81, P for heterogeneity = 0.89, I 2 = 0%, with 7 studies included]. The result of sensitivity analysis that excluded the study with  Table 2).

Transfusion of any blood products
Eleven trails with a total of 5360 participants reported the postoperative transfusion rate of any blood product. Overall, TXA significantly reduced the transfusion of any blood products [RR = 0.64(0.52-0.78), P for effect< 0.01, P for heterogeneity< 0.01, I 2 = 76%] (Fig. 9).
In the sensitivity analysis that included all the studies that randomized more than 99 participants, TXA significantly decreased the transfusion of any blood products [RR = 0.64(0.50-0.81), P for effect< 0.01, P for hetero-geneity< 0.01, I 2 = 86%]. The result of sensitivity analysis that excluded the study with maximum sample size further enhanced the above analyses [139/396 vs 216/363, RR = 0.29 (0.20-0.40), P for effect < 0.01, P for heterogeneity < 0.01] ( Table 2).

Postoperative chest tube drainage in the first 24 h
In total, 16 studies with 6247 participants were included in the meta-analysis of postoperative chest tube drainage in the first 24 h. One of them [18] divided participants into two groups according to the difference in fluid use and reported the drainage of patients receiving TXA and placebo in both groups separately. We treated these two sets of data as two separate studies in the meta-analysis. Overall, the chest tube drainage was significantly  (Fig. 10).

Quality of the evidence
The GRADE approach was adopted to evaluate the quality of each outcome and "Summary of findings" tables were presented (Table 3). In general, the overall quality of evidence in the meta-analyses of postoperative seizures and reoperation for bleeding was high. However, the overall quality of evidence in the meta-analyses of postoperative transfusion of any blood products and chest tube drainage in the first 24 h was very low due to the problems of inconsistency and the risk of bias. The overall quality of evidence of other outcomes was moderate due to the risk of bias.

Discussion
In this meta-analysis, we found that the intravenous use of TXA was associated with lower risk of reoperation for postoperative bleeding, blood loss and blood transfusion than the placebo group. Moreover, we also found that intravenous use of TXA did not increase the risk of postoperative cerebrovascular accident, mortality or other thrombotic complication among patients undergoing CABG when compared with placebo treatment. However, it may increase the incidence of postoperative seizures. The results of most subgroup analyses of the primary results in CABG conducted under on-pump or off-pump condition were consistent with that of overall analyses. However, meta-analysis could not be performed in the sub-analyses of postoperative reoperation for bleeding, mortality and cerebrovascular accident in offpump CABG due to the small number of incidence. No significant decrease in postoperative reoperation for bleeding and transfusion of any blood products were found in on-pump group. Most of the sensitivity analyses in trails that recruited more than 99 participants or in trails that excluded the study with the largest number of participants further strengthened the conclusion of overall analyses.
The release of plasmin during cardiac surgery activates fibrinolysis and may contribute to platelet dysfunction [40]. In addition to inhibiting the transformation of plasminogen into plasmin by reversibly binding lysine binding site on plasmin, TXA can also reduce bleeding by preventing platelet activation induced by fibrinolytic enzyme [41]. A previous meta-analysis suggested that TXA was effective in reducing blood loss and the need for blood transfusion in cardiac surgery [42]. However, the incidence of reoperation for bleeding was not significantly decrease by TXA [42]. In our current analysis, we found that TXA overall reduced the transfusion of any blood products and 24-h postoperative chest tube drainage in CABG surgery which was consistent with the previous study. Moreover, the sub-analyses in the different conditions under which GABG was conducted further strengthened the above results. However, these analyses have significant heterogeneity which may due to the difference in indications of blood transfusion, drug dose and treatment regimens among different studies.
Different from the previous study, our current meteanalysis suggested that TXA significantly decrease the incidence of reoperation for bleeding in CABG surgery with low heterogeneity. In addition, the sensitivity analyses in studies randomized more than 99 participants further strengthened the conclusion that TXA reduced the incidence of reoperation for bleeding, transfusion of any blood products and 24-h blood loss suggesting that the small sample size studies included in the meta-  The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) b 4 studies with a high risk of bias were included c few studies reported this result d RR > 5 e 5 studies with a high risk of bias were included f RR < 0.5 g No explanation was provided h I2 > 75% i 2 studies with a high risk of bias were included analysis did not affect the overall effectiveness. However, the sensitivity analysis that excluded the study [8] with maximum sample size did not suggest that TXA would significantly decrease the incidence of reoperation for bleeding. This result suggested that the study with the largest number of participants largely determines the overall effect of meta-analysis. While considering the low risk of bias assessment in that study, we can still believe that TXA overall decrease the incidence of reoperation for bleeding. In the sub-analysis of on-pump GABG, TXA tended to reduce the incidence of reoperation for bleeding. However, the effect was not statistically significant. The exclusion of the study with the largest number of participants due to mixed surgical types in the sub-analysis may explain this difference.
Although lots of studies have suggested that blood transfusion and reoperation for bleeding is associated with poor outcomes after cardiac surgery, we did not find that TXA would reduce the risk of cerebrovascular accident, myocardial infarction, acute renal insufficiency or mortality despite its effectiveness in reducing transfusion and reoperation for bleeding. A previous metaanalysis had reported that TXA reduced blood transfusion in off-pump CABG and did not increased the incidence of postoperative adverse events [9]. However, the sample size in that study was not sufficient to detect the rare but clinically significant adverse events. In the current meta-analysis, enough population were included in the above analyses to detect clinically significant difference. Moreover, the above conclusion were strengthened by sensitivity analyses in trails enrolling more than 99 patients or sensitivity analyses excluded the study with largest sample size. In addition, there was no heterogeneity in above analyses from the results of heterogeneity tests and the risk of publication bias in these meta-analyses was quite low revealed by funnel plots. These unexpected results may be explained by the potential prothrombotic effects of TXA. It is well known that 5 to 15% of all grafts may be blocked in the early postoperative period even without the use of antifibrinolytic agents, which may led to recurrence of myocardial ischemia, infarction, or even death [43,44]. Perioperative inhibition of fibrinolysis may increase the rate of early graft occlusion rate [45]. The phenomenon that TXA reduced transfusion, blood loss and incidence of reoperation without decreasing postoperative morality or adverse events may be a balance of its blood conservation effect and potential prothrombotic effect.
A previous meta-analysis suggested that the risk of seizure increased in patients with TXA exposure [46]. In the current meta-analysis we found that TXA increased the incidence of postoperative seizures in CABG surgery. Several studies have suggested that the convulsant property of TXA is likely mediated by disinhibition of gama-aminobutyric acid type A (GABAA) receptors and glycine receptor, which are two major mediators of inhibition in the CNS [47,48]. Moreover, TXA did not interfere with N-methyl-Daspartate receptor and impact glutamatergic synaptic transmission [48,49]. In addition, some studies have shown that TXA reduces cerebral blood flow and increases the risk of cerebral infarction which could contribute to the postoperative seizures. However, the meta-analysis of postoperative cerebrovascular accident in current study did not supported the hypothesis that TXA increase incidence of seizures by increasing the incidence of cerebral infarction. Moreover, a growing number of studies have suggested the seizures associated with TXA to be dose related [6,50,51]. Therefore, studies that investigate the optimize dose and regime for administration of TXA are needed in the future. Moreover, a growing number of studies that investigate the efficacy and safety of topical use of tranexamic acid have been conducted in recent years due to the promise of reducing postoperative bleeding and seizures [52,53]. A recent meta-analysis showed that the topical application of TXA effectively reduces both transfusion risk and blood loss compared to placebo and no major differences were found between topical and intravenous tranexamic acid with respect to safety and efficacy [54]. However, both surgical and non-surgical trials were included in that study. While in our study, we focused on the safety and efficiency of intravenous administration of tranexamic acid in coronary artery bypass grafting (CABG).
There are some limitations in this meta-analysis. Firstly, heterogeneity due to clinical and methodological diversity was inevitable which may affect the reliability of the analysis results especially in meta-analyses of transfusion and blood loss. Secondly, some data were presented as median and interquartile range which cannot be used in performing meta-analysis. We estimated the mean and standard deviation from those data to perform meta-analysis which may compromise the reliability of analysis results. Thirdly, the postoperative incidence of adverse event was suggested to may be dose-dependent [6], while we failed to performed subanalysis in different dose setting due to the various dosage and regimens of TXA administration in current meta-analysis. Fourthly, a multicenter study that randomized 2311 participants occupied the main part of most analyses which may lead to bias. Despite the above limitations, the current study is still the most comprehensive analysis on the efficacy and safety of TXA in CABG surgery with sufficient sample size.

Conclusion
The current study systematically reviewed the existing evidence on the efficacy and safety profile of the