Efficacy and safety of tranexamic acid in cardiac surgery: a systematic review and network meta-analysis

Abstract

Background

Tranexamic acid (TXA) is recommended for reducing blood loss and transfusion in cardiac surgery. However, there are concerns regarding the safety profile of TXA, especially its proconvulsant effects. We conducted this study to investigate the efficacy and safety of tranexamic acid in cardiac surgery.

Methods

We searched PubMed, Embase and Cochrane Central Register of Controlled Trials from inception to December 11, 2024. Randomised controlled trials assessed the hemostatic effects of TXA in cardiac surgery were included. Two authors independently selected studies and assessed the quality of eligible trials. The main endpoints were red blood cell transfusion and thrombotic outcomes. The results were calculated with pairwise and network meta-analysis.

Results

Data was provided by 18,141 participants from 64 trials. High-dose continuous (OR: 0.38, 95%CI: [0.31, 0.47]), low-dose continuous (OR: 0.44, 95%CI: [0.34, 0.56]), high-dose single (OR: 0.50, 95%CI: [0.43, 0.57]), and low-dose single (OR: 0.52, 95%CI: [0.40, 0.67]) TXA significantly reduce the rate of red blood cell transfusion. Furthermore, in high-risk patients, high-dose continuous administration further reduces transfusion risk compared to low-dose continuous administration (OR: 1.22, 95%CI: [1.01, 1.75]). Topical TXA does not significantly reduce the rate of red blood cell transfusion (OR: 0.80, 95%CI: [0.60, 1.07]); conversely, it is associated with a higher rate of red blood cell transfusion compared to intravenous administration. Both intravenous and topical TXA administration reduce postoperative blood loss. High-dose continuous administration further reduces the risk of reoperation (OR: 1.70, 95%CI: [1.03, 2.80]) and the need for fresh frozen plasma transfusion (OR: 1.33, 95%CI: [1.01, 1.74]) compared to low-dose continuous administration. Neither intravenous nor topical TXA increases the incidence of postoperative thrombotic complications. High-dose single administration is associated with a significantly increased risk of postoperative seizures (OR: 6.66, 95%CI: [1.85, 24.02]).

Conclusions

This meta-analysis further confirms that intravenous TXA administration, regardless of dose or administration regimen, significantly reduces postoperative blood loss and red blood cell transfusions in adult cardiac surgery, without increasing the incidence of serious adverse events except for seizures. Future studies should incorporate patient-specific factors, comorbidities, and bleeding risks to determine the optimal TXA dosing strategy that balances risks and benefits.

Trial registration

Our prespecified protocol was registered with PROSPERO (CRD42022380404).

Supplementary Information

The online version contains supplementary material available at 10.1186/s12871-025-03365-8.

Background

Surgical trauma and cardiopulmonary bypass (CPB) during cardiac surgery can lead to the destruction of coagulation factors and platelets, inhibition of platelet function, and subsequent hyperfibrinolysis, which collectively contribute to an increased risk of perioperative bleeding, allogeneic blood product transfusion, and the need for secondary thoracotomy to control bleeding [1, 2]. Moreover, blood transfusion and secondary thoracotomy are strongly associated with poor prognosis in cardiac surgery patients [3]. Anti-fibrinolytic medications effectively inhibit plasmin activation and preserve platelet function, thereby reducing postoperative bleeding [4]. Through reversibly binding to the lysine-binding sites of plasminogen and plasmin, tranexamic acid prevents plasmin from binding to the lysine residues of fibrin, thereby inhibiting fibrinolysis. Additionally, tranexamic acid further protects platelet function by attenuating the effect of plasmin on platelet glycoprotein Ib receptors [5].

Although tranexamic acid is widely recognized for its efficacy in reducing bleeding and blood transfusion requirements during cardiac surgery, the optimal dosing regimen remains uncertain [3, 6]. A high-quality clinical trial previously demonstrated that a regimen of 10 mg/kg followed by 1 mg/kg/h over 12 h effectively reduced bleeding and transfusion risks, with no additional hemostatic benefit observed at higher doses [7]. Based on pharmacokinetic modeling [8, 9], some researchers have proposed regimens targeting specific plasma concentrations of tranexamic acid, which in vitro studies have shown to completely inhibit fibrinolysis [10]. These regimens involve substantial doses, such as 30 mg/kg followed by 16 mg/kg/h during surgery, with 2 mg/kg added to the pump prime, or a preoperative bolus of 100 mg/kg. The relationship between in vivo drug concentrations, fibrinolytic inhibition and blood loss reduction has not been definitively established. Furthermore, tranexamic acid seemed to be associated with an increased risk of postoperative seizures [11] that were believed to be dose-related [11–13]. Myles et al. [3] reported a prospective, randomized trial of TXA versus placebo and found that patients receiving a TXA dose of 50 or 100 mg/kg had a significantly higher rate of seizures (0.7%) when compared to placebo (0.1%) (P = 0.002). A 2019 meta-analysis found that lower dose TXA equally decreased transfusion requirements compared to higher dose TXA. However, the higher dose group had a 4.83 times higher risk of seizures than the lower dose group [14]. Determining the optimal dose of TXA administration during cardiac surgery remains a challenge.

The topical application of tranexamic acid in cardiac surgery is not a novel concept. Previous small-scale trials in both non-cardiac [15, 16] and cardiac surgery [14] have indicated that topical tranexamic acid effectively reduces perioperative bleeding compared to placebo. This approach is associated with lower plasma concentrations of tranexamic acid than intravenous administration, potentially mitigating the risk of seizures [17, 18]. However, it remains unclear whether the topical use of tranexamic acid in cardiac surgery can reduce seizure risk without compromising hemostatic efficacy.

This study aims to summarize evidence from all trials published before and try to provide information on the optimal dosage and delivery methods which is effective with the least adverse outcomes.

Methods

This systematic review was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement extension for network meta-analysis [19]. Our prespecified protocol was registered with PROSPERO (CRD42022380404).

Search strategy

We searched five electronic databases: PubMed/MEDLINE, Embase, Cochrane Central Register of Controlled Trials, Scopus, and Web of Science, from their inception to December 11, 2024, for relevant articles published in any language. A detailed description of the search strategy is provided in the supplemental methods section. We manually searched the references of the related meta-analyses and articles of interest to identify further eligible studies.

Eligibility criteria

Randomized controlled trials assessing the efficacy and safety of tranexamic acid in adult patients undergoing elective heart surgery were considered eligible. The interventions in this study were categorized as follows: (1) High-dose continuous TXA infusion (HD-CI); (2) Low-dose continuous TXA infusion (LD-CI) [6]; (3) High-dose bolus TXA (HD-Bolus); (4) Low-dose bolus TXA (LD-Bolus) [3, 20]; (5) Intrapericardial TXA [21]. Table 1. presented specific definitions of the interventions included in this study.

Table 1.

Definitions of interventions included in this study

Intervention	Definition
High-dose continuous TXA infusion (HD-CI)	Patients received continuous infusions of tranexamic acid at doses larger than a loading dosage of 10 mg kg−1 and a maintenance dose of 2 mg kg−1 h−1.
Low-dose continuous TXA infusion (LD-CI)	Patients received continuous infusions of tranexamic acid at doses less than a loading dosage of 10 mg kg−1 and a maintenance dose of 2 mg kg−1 h−1.
High-dose bolus TXA (HD-Bolus)	Patients received a single bolus of tranexamic acid larger than 2 g or 50 mg per kg of body weight.
Low-dose bolus TXA (LD-Bolus)	Patients received a single bolus of tranexamic acid less than 2 g or 50 mg per kg of body weight.
Intrapericardial TXA	Tranexamic acid was diluted in 0.9% saline and poured into mediastinal cavity before closing the chest.

The exclusion criteria were as follows: (1) patients aged less than 18 years or who underwent non-cardiac surgery or urgent surgery; (2) studies with other hemostatic strategies; (3) incomplete or erroneous data that could not be merged; (4) Case reports, conference abstracts, letters, or animal trials; (5) sample size of less than 30.

Data extraction and outcomes

Two reviewers independently extracted data from the original trial reports using a specifically designed form that captured information on study design, patient characteristics (age, sex, weight), sample sizes, details of the interventions, surgery characteristics (duration of surgery and CPB, type of surgery, transfusion policy), and outcomes. When relevant data were missing, we attempted to contact the corresponding authors.

The primary efficacy outcome was the proportion of patients who received any red blood cell transfusion after the start of operation. The primary safety outcome was the rate of a composite of perioperative cardiovascular thromboembolic events, defined as any deep vein thrombosis, pulmonary embolism, myocardial ischemia/infarction, or cerebral ischemia/infarction. The secondary outcomes included postoperative 24 h blood loss, fresh frozen plasma transfusion rate, platelet transfusion rate, incidence of reoperation due to major bleeding, incidence of postoperative mortality or seizures, and postoperative platelet counts.

Risk of bias assessment

The revised Cochrane risk of bias tool for randomized trials (ROB2) [22] was used to assess the quality of the eligible studies. It comprises five primary domains: randomization procedure, intended intervention, missing outcome data, outcome evaluations, and selective reporting. Each domain was categorized as low risk, high risk, or some concerns. Two reviewers independently assessed the risk of bias and any differences were resolved by a third reviewer.

Statistical analysis

All statistical analyses were performed using Stata 17.0. Given the possibility of heterogeneity among trials, a random-effects model was chosen for analysis. We calculated summary mean differences (MD) with 95% confidence intervals (CIs) for continuous data and odds ratios (OR) with 95% CIs for categorical data. Initially, we conducted a conventional pairwise meta-analysis and assessed statistical heterogeneity using the I2 statistic, with values more than 50% suggesting substantial heterogeneity [23]. Secondly, we conducted a frequentist random-effects network meta-analysis and created a corresponding forest plot [24].

A prediction interval plot was drawn to evaluate the heterogeneity among studies (see Supplemental Fig. 2. for details). If there is no heterogeneity, the boundaries of the prediction intervals should match the CIs obtained from the random-effects model [25]. We created a network plot and network/treatment characteristic descriptive data to describe the geometry of the network for each outcome. Publication bias was examined by the funnel plot method. The surface under the cumulative ranking (SUCRA) curve was used for ranking, with larger SUCRA values indicating a greater likelihood to rank the best [26, 27]. The global inconsistency was evaluated for the entire network using the Wald Chi2 test obtained by fitting the inconsistency model. The node-splitting method was applied for the local inconsistency check. We evaluated the possible heterogeneity of treatment effects and the robustness of our findings with subgroup network meta-analyses using transfusion risk. High transfusion risk cardiac surgery was defined as one of the following surgical types: repeat cardiac surgery; combined procedures (e.g., valvular operation plus CABG); and other complex procedures (e.g., multiple valve replacement or repair, ascending aortic graft) [28–30]. The sensitivity analyses we conducted either excluded studies with a high risk of bias or included only trials with clear transfusion protocols. The GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach was employed to rate the certainty of evidence for each network estimate [31].

Fig. 2.

Practical flow chart for optimal tranexamic acid dose in cardiac surgery

Results

Systemic review

Initially, we identified 730 papers from electronic databases and 3 additional papers from other sources. After eliminating duplicates, 412 papers were screened on titles and abstracts, and 339 papers were not in accordance with the eligibility criteria. The full texts of 73 papers were obtained and analyzed. Ultimately, 64 papers involving 18,141 patients were selected for inclusion in this network meta-analysis (NMA). The selection procedure is summarized in Fig. 1.

Fig. 1.

PRISMA flow chart

Characteristics of included studies

Supplemental Table 1 shows the features of the included studies. These 64 trials were conducted in 23 countries and were published between 1995 and 2024. Double-blinded (k = 45 studies), placebo-controlled (k = 58 studies), and two-arm (k = 62 studies) trials constituted the bulk of the included trials. Most study participants were patients who underwent CABG, with an average surgical duration of 241.4 min. The overall male-to-female ratio in these studies was 52.3% [29–94%], with the median age of the participants being 58.4y [135y-71y]. The sample size varied from 30 to 4631, with 15 trials involving more than 180 participants. Within the eligible trials, there was considerable variation in the transfusion criteria.

Risk of bias assessments

Sequence generation and allocation concealment were performed in 36 and 40 trials, respectively. All trials followed the allocated interventions and were deemed to have a low risk of bias. In two of the included trials, the dropout rate was significant, while it was unclear in four trials. Owing to the improper approach employed to quantify postoperative 24 h blood loss, potential measurement bias was discovered in one study. Only 15 trials provided registration numbers and reported predetermined outcomes. Conversely, the majority of studies were assessed as having a high or uncertain risk of selective reporting bias. Overall, 9 studies were found to have a low risk of bias, 44 had some concerns, and 11 had a high risk.

Network meta-analysis

Primary efficacy outcomes

Red blood cell transfusion rate

In terms of transfusion needs, 40 studies (n = 15 971 patients) were included in the NMA [Table 2]. Intrapericardial TXA (OR 0.80, 95% CI 0.60 to 1.07) did not significantly lower the risk of red cells transfusion, whereas HD-CI (OR 0.38, 95% CI 0.31 to 0.47), LD-CI (OR 0.44, 95% CI 0.34 to 0.56), HD-Bolus (OR 0.50, 95% CI 0.40 to 0.67) and LD-Bolus (OR 0.52, 95% CI 0.40 to 0.67) were more effective than placebo in this regard. HD-Bolus was linked to an increased incidence of transfusion requirements compared with HD-CI TXA (OR 1.30, 95% CI 1.03, 1.65). Patients receiving intrapericardial TXA required transfusions at a considerably higher incidence than those receiving HD-CI (OR 2.11, 95% CI, 0.50 to 2.95), LD-CI (OR 1.83, 95% CI, 1.26 to 2.66), HD-Bolus (OR 1.62, 95% CI, 1.20 to 2.18), or LD-Bolus (OR 1.55, 95% CI 1.29 to 1.85).

Table 2.

Network meta-analysis for transfusion needs and thromboembolic events

Intrapericardial TXA	1.12(0.72,1.76)	0.85(0.46,1.58)	0.83(0.39,1.39)	0.79(0.38,1.65)	0.76(0.41,1.41)
>1.55(1.29,1.85)	LD-Bolus	0.76(0.47,1.24)	0.74(0.39,1.39)	0.70(0.37,1.32)	0.68(0.42,1.10)
1.62(1.20,2.18)	1.05(0.80,1.37)	HD-Bolus	0.97(0.63,1.51)	0.92(0.60,1.43)	0.90(0.77,1.04)
1.83(1.26,2.66)	1.18(0.85,1.65)	1.13(0.84,1.53)	LD-CI	0.95(0.80,1.13)	0.92(0.61,1.40)
2.11(1.50,2.95)	1.36(1.00,1.85)	1.30(1.03,1.65)	1.15(0.94,1.41)	HD-CI	0.97(0.64,1.47)
0.80(0.60,1.07)	0.52(0.40,0.67)	0.50(0.43,0.57)	0.44(0.34,0.56)	0.38(0.31,0.47)	placebo

The results of red-cell transfusion needs were presented in the left lower half and the results of thromboembolic events in the upper right half. Comparisons between treatments should be read from left to right. HD-CI high-dose continuous tranexamic acid infusion; LD-CI low-dose continuous tranexamic acid infusion; HD-Bolus high-dose bolus of tranexamic acid; LD-Bolus low-dose bolus of tranexamic acid; TXA tranexamic acid

HD-CI had the highest SUCRA value (97.6), followed by LD-CI (SUCRA 73.9), HD-Bolus (SUCRA 57.2), and LD-Bolus (SUCRA 51.3). Patients receiving intrapericardial TXA appeared to yield the least benefit from a decreased incidence of transfusion requirements (SUCRA 18.6).

Primary safety outcomes

A composite of thromboembolic events

Thirty-four studies comprising 15 614 individuals reported the risk of postoperative complications. Compared with placebo, no intervention dramatically lowered the probability of postoperative composite complications (HD-CI: OR 0.97, 95% CI 0.64 to 1.47, LD-CI: OR 0.92, 95% CI 0.61 to 1.40, HD-Bolus: OR 0.90, 95% CI 0.77 to 1.04, LD-Bolus: OR 0.68, 95% CI 0.42 to 1.10, and intrapericardial TXA: OR 0.76, 95% CI 0.41 1.41).

LD-Bolus had the highest SUCRA value (83.5), followed by intrapericardial TXA (SUCRA 64.6), HD-Bolus (SUCRA 51.6), and LD-CI (SUCRA 45.6). HD-CI had the lowest SUCRA value (32.0).

Secondary outcomes

Postoperative 24 h blood loss

Regarding the amount of blood lost in the 24 h following surgery, 35 trials with 3907 participants were included in this NMA. Compared with patients getting placebo, patients receiving intrapericardial TXA (MD −283.09, 95% CI −406.01 to −160.18), HD-CI (MD −224.31, 95% CI −325.55 to −123.08), LD-CI (MD −158.40, 95% CI −266.31 to −50.48), HD-Bolus (MD −198.85, 95% CI −317.93 to −79.76) and LD-Bolus (MD −127.79, 95% CI −231.66 to −23.92) lost less blood after surgery. A comparative analysis of the various interventions revealed no statistically significant differences in decreasing postoperative blood loss.

With a ranking likelihood of 89.9, intrapericardial TXA was ranked the highest, followed by HD-CI (SUCRA 72.1). LD-Bolus (SUCRA 32.2) came last, while the SUCRA scores for HD-Bolus (SUCRA 61.8) and LD-CI (SUCRA 43.8) were comparable.

Fresh frozen plasma transfusion rate

Twenty-nine studies comprising 7359 individuals reported the incidence of fresh frozen plasma transfusion. Intrapericardial TXA (OR 1.14, 95% CI 0.42 to 3.10) did not significantly reduce the fresh frozen plasma transfusion rate, while HD-CI (OR 0.37, 95% CI 0.29 to 0.47), LD-CI (OR 0.49, 95% CI 0.35 to 0.68), HD-Bolus (OR 0.31, 95% CI 0.19 to 0.49) and LD-Bolus (OR 0.35, 95% CI 0.22 to 0.57) were more effective than placebo in this regard. Patients receiving intrapericardial TXA required fresh frozen plasma transfusions at a significantly higher incidence than those receiving HD-CI (OR 3.09, 95% CI 1.08 to 8.87), HD-Bolus (OR 3.73, 95% CI 1.21 to 11.51) or LD-Bolus (OR 3.25, 95% CI 1.05 to 10.12). HD-CI further lowered the risk of fresh frozen plasma transfusion as compared to LD-CI (OR 1.33, 95% CI 1.01 to 1.75).

HD-Bolus had the highest SUCRA value (86.5), followed by LD-Bolus (SUCRA 75.5), HD-CI (SUCRA 71.5), and LD-CI (SUCRA 44.3). Intrapericardial TXA had the lowest SUCRA value (10.3).

Platelet transfusion rate

In terms of platelet transfusions, the NMA aggregated data from 20 studies (n = 5994 participants). HD-CI (OR 0.52, 95% CI 0.31 to 0.86) and HD-Bolus (OR 0.31, 95% CI 0.17 to 0.55) were linked to a significant decrease in the risk of platelet transfusion, while LD-CI (OR 0.67, 95% CI 0.37 to 1.19), LD-Bolus (OR 0.56, 95% CI 0.18 to 1.70) and intrapericardial TXA (OR 0.68, 95% CI 0.19 to 2.47) did not significantly impact the platelet transfusion rate. No significant differences were observed among the interventions in terms of reducing the platelet transfusion rate (Table 3).

Table 3.

Network meta-analysis for the incidence of fresh frozen plasma and platelet transfusion

Intrapericardial TXA	1.22(0.22,6.64)	2.22(0.54,9.99)	1.02(0.25,5.78)	1.32(0.33,5.25)	0.68(0.19,2.47)
3.25(1.05,10.12)	LD-Bolus	1.82(0.57,5.79)	0.84(0.24,2.93)	1.08(0.32,3.67)	0.56(0.18,1.70)
3.73(1.21,11.51)>	2.32(0.62,2.13)	HD-Bolus	0.46(0.20,1.05)	0.59(0.26,1.29)	0.31(0.17,0.55)
2.32(0.79,6.80)	0.71(0.39,1.29)	0.62(0.35,1.10)	LD-CI	1.29(0.87,1.90)	0.67(0.37,1.19)
3.09(1.08,8.87)	0.95(0.55,1.63)	0.83(0.49,1.41)	1.33(1.01,1.74)	HD-CI	0.52(0.31,0.87)
1.14(0.42,3.10)	0.35(0.22,0.57)	0.31(0.19,0.49)	0.49(0.35,0.68)	0.37(0.29,0.47)	placebo

The results of fresh frozen plasma transfusion rate were presented in the left lower half and the results of platelet transfusion rate in the upper right half. Comparisons between treatments should be read from left to right. HD-CI high-dose continuous tranexamic acid infusion; LD-CI low-dose continuous tranexamic acid infusion; HD-Bolus high-dose bolus of tranexamic acid; LD-Bolus low-dose bolus of tranexamic acid; TXA tranexamic acid

With a ranking likelihood of 92.6, HD-Bolus was ranked the highest, followed by HD-CI (SUCRA 63.5). LD-CI (SUCRA 28.1) came last, while the SUCRA scores for LD-Bolus (SUCRA 56.0) and intrapericardial TXA (SUCRA 45.3) were comparable.

Reoperation due to major hemorrhage

The NMA evaluated the risk of re-exploration due to severe bleeding by incorporating 36 trials with 15 627 participants. Table 4. showed that while the effect of LD-CI (OR 0.80, 95% CI 0.45 to 1.43) and intrapericardial TXA (OR 0.53, 95% CI 0.23 to 1.19) on reoperation was not statistically significant, HD-CI (OR 0.47, 95% CI 0.25 to 0.90), HD-Bolus (OR 0.46, 95% CI 0.30 to 0.71), and LD-Bolus (OR 0.42, 95% CI 0.20 to 0.88) considerably reduced the frequency of reoperation due to severe bleeding. In addition, HD-CI was associated with a lower risk than LD-CI (OR 1.70, 95% CI 1.03 2.80).

Table 4.

Network meta-analysis for postoperative blood loss and reoperation

Intrapericardial TXA	1.24(0.74,2.09)	1.14(0.50,2.57)	0.66(0.24,1.77)	1.12(0.41,3.07)	0.53(0.23,1.19)
−155.30 (−155.30, 5.61)	LD-Bolus	0.91(0.40,2.06)	0.53(0.21,1.30)	0.90(0.34,2.33)	0.42(0.20,0.88)
−84.25 (−255.35, 86.86)	71.05 (−61.35, 203.46)	HD-Bolus	0.58(0.28,1.21)	0.98(0.46,2.11)	0.46(0.30,0.71)
−124.69 (−288.24, 38.85)	30.61 (−111.21,172.42)	−40.45 (−198.28,117.38)	LD-CI	1.70(1.03,2.80)	0.80(0.45,1.43)
−58.78 (−217.99,100.43)	96.52 (−46.91,239.95)	25.47 (−129.90,180.84)	65.92 (−67.09,198.92)	HD-CI	0.47(0.25,0.90)
−283.09 (−406.01,−160.18)	−127.79 (−231.66,−23.92)	−198.85 (−317.93,−79.76)	−158.40 (−266.31,−50.48)	−224.31 (−325.55,−123.08)	placebo

The results of postoperative blood loss were presented in the left lower half and the results of repeat surgery in the upper right half. Comparisons between treatments should be read from left to right. HD-CI high-dose continuous tranexamic acid infusion; LD-CI low-dose continuous tranexamic acid infusion; HD-Bolus high-dose bolus of tranexamic acid; LD-Bolus low-dose bolus of tranexamic acid; TXA tranexamic acid

With a ranking likelihood of 77.4, LD-Bolus was ranked the highest, followed by HD-Bolus (UCRA 70). LD-CI (SUCRA 22.8) came last, while the SUCRA scores for HD-CI (SUCRA 68.8) and intrapericardial TXA (SUCRA 54.7) were comparable.

Rate of postoperative seizure

Seizures were documented in 10 trials with a total of 12 631 patients. The results of this network meta-analysis revealed a greater likelihood of postoperative seizures in patients with HD-Bolus (OR 6.66, 95% CI 1.85 to 24.02). Intrapericardial TXA did not exhibit a significant lower incidence of postoperative seizures compared with HD-CI (OR 0.28, 95% CI 0.01 to 5.47), LD-CI (OR 0.58, 95% CI 0.03 to 2.93), HD-Bolus (OR 0.11, 95% CI 0.01 to 1.38), or LD-Bolus (OR 0.11, 95% CI 0.11 to 1.14). No significant differences were observed in the comparison of high-dose and low-dose TXA.

Intrapericardial TXA had the highest SUCRA value (89.6), followed by LD-CI (SUCRA 66.2), LD-Bolus (SUCRA 42.4), and HD-CI (SUCRA 31.6). HD-Bolus had the lowest SUCRA value (8.6), making it more likely than other therapies to cause early postoperative seizures.

Incidence of mortality

In terms of mortality, the NMA aggregated data from 23 studies (n = 10 047 participants). Table 5. showed that intrapericardial TXA (OR 0.31, 95% CI 0.10 to 0.93) and LD-Bolus (OR 0.28, 95% CI 0.09 to 0.86) were linked to a significant decrease in postoperative death. HD-CI (OR 0.48, 95% CI 0.19 to 1.24), LD-CI (OR 0.66, 95% CI 0.25 to 1.79), and HD-Bolus (OR 0.72, 95% CI 0.26 to 1.99) did not significantly impact mortality compared with placebo. No significant differences were observed in the other head-to-head comparisons.

Table 5.

Network meta-analysis for postoperative seizures and death

Intrapericardial TXA	1.11(0.51,2.43)	0.44(0.14,1.38)	0.47(0.11,2.06)	0.65(0.15,2.78)	0.31(0.10,0.93)
0.36(0.11,1.14)	LD-Bolus	0.39(0.14,1.10)	0.42(0.09,1.88)	0.59(0.14,2.53)	0.28(0.09,0.86)
0.11(0.01,1.38)	0.29(0.03,12.49)	HD-Bolus	1.08(0.26,4.49)	1.50(0.37,6.03)	0.72(0.26,1.99)
0.58(0.03,12.49)	1.59(0.09,27.76)	5.41(0.66,27.76)	LD-CI	1.38(0.76,2.52)	0.66(0.25,1.79)
0.28(0.01,5.47)	0.77(0.05,12.05)	2.61(0.37,18.60)	0.48(0.22,1.03)	HD-CI	0.48(0.19,1.24)
0.71(0.05,9.38)	1.96(0.19,19.88)	6.66(1.85,24.02)	1.23(0.23,6.55)	2.55(0.58,11.31)	placebo

The results of postoperative seizures were presented in the left lower half and the results of mortality rate in the upper right half. Comparisons between treatments should be read from left to right. HD-CI high-dose continuous tranexamic acid infusion; LD-CI low-dose continuous tranexamic acid infusion; HD-Bolus high-dose bolus of tranexamic acid; LD-Bolus low-dose bolus of tranexamic acid; TXA tranexamic acid

LD-Bolus (SUCRA 84.0) ranked the highest among the SUCRA values, followed by intrapericardial TXA (SUCRA 77.2), HD-CI (SUCRA 60.2), LD-CI (SUCRA 34.9), and HD-Bolus (SUCRA 32.3).

Postoperative platelet counts

In terms of postoperative platelet counts, the NMA aggregated data from 28 studies (n = 3022 participants). Compared with placebo, no intervention dramatically increased platelet counts (HD-CI: MD 12.96, 95%CI −2.68 to 28.60, LD-CI: 14.68, 95%CI −5.45 to 34.82, HD-Bolus: 2.35, 95%CI −18.52 to 23.22, LD-Bolus: 8.98, 95%CI −6.19 to 24.15, intrapericardial TXA: 6.00, 95%CI −53.05 to 41.05).

Intrapericardial TXA (SUCRA 70.2) ranked the highest among the SUCRA values, followed by HD-Bolus (SUCRA 61.9), LD-Bolus (SUCRA 40.3), HD-CI (SUCRA 28.5), and LD-CI (SUCRA 24.4). [Table 6].

Table 6.

Network meta-analysis for red-cell transfusion volume and postoperative platelet counts

Intrapericardial TXA	−14.98 (−64.42,34.46)	−8.35 (−59.82,43.13)	−20.68 (−71.86,30.50)	−18.96 (−68.55,30.63)	−6.00 (−53.05,41.05)
−0.87 (−1.71 , −0.03)	LD-Bolus	6.63 (−17.30,30.56)	−5.71 (−30.92,19.51)	−3.98 (−25.77,17.81)	8.98 (−6.19,24.15)
−0.44 (−1.14,0.25)	0.43 (−0.40,1.26)	HD-Bolus	−12.33 (−41.34,16.67)	−10.61 (−36.69,15.47)	2.35 (−18.52,23.22)
−0.27 (−0.94,0.41)	0.60 (−0.19,1.40)	0.18 (−0.48,0.83)	LD-CI	1.72 (−18.70,22.15)	14.68 (−5.45,34.82)
0.21 (−0.51,0.93)	1.08 (0.25,1.91)	0.65 (−0.05,1.36)	0.48 (−0.08, 1.03)	HD-CI	12.96 (−2.68,28.60)
−1.11 (−1.61 , −0.61)	−0.24 (−0.91, 0.43)	−0.67 (−1.15 , −0.18)	−0.84 (−1.29 , −0.39)	−1.32 (−1.84 , −0.80)	Placebo

The results of red-cell transfusion volume were presented in the left lower half and the results of postoperative platelet counts in the upper right half. Comparisons between treatments should be read from left to right. HD-CI high-dose continuous tranexamic acid infusion; LD-CI low-dose continuous tranexamic acid infusion; HD-Bolus high-dose bolus of tranexamic acid; LD-Bolus low-dose bolus of tranexamic acid; TXA tranexamic acid

Subgroup and sensitive analyses

With pooled OR of 1.22 (95%CI: 1.05 to 1.43) and 0.93 (95% CI: 0.66 to 1.32), respectively, subgroup analysis on transfusion risk revealed that HD-CI significantly reduced the incidence of blood transfusions compared with LD-CI in high-risk cardiac surgery but not in low-risk cardiac surgery. The lack of sufficient eligible trials in this study prevented us from performing subgroup analysis to identify the impact of transfusion risk on the incidence of peri-operative thromboembolic events. The data for subgroup analysis are summarized in Supplemental Fig. 2.

Sensitivity analyses did not identify significantly different patterns from the crude findings. It is noteworthy that the results of the sensitivity analyses revealed a lack of robustness in the significance of intrapericardial TXA, with larger confidence intervals and less accuracy. The sensitivity analysis results are displayed in Supplemental Figs. 3. and 4.

Heterogeneity and inconsistency

In this study, no indication of a statistically significant global inconsistency was identified. Node-splitting analysis revealed the presence of local inconsistency for postoperative 24-hour blood loss between LD-CI and placebo groups and reoperation due to major bleeding between HD-Bolus and placebo groups. Given that all variables had I2 values far below 50%, there was little evidence of substantial heterogeneity (Supplemental Table 5 and 6.).

Publication bias

The comparison-adjusted funnel plots for each endpoint appeared to be symmetrical, indicating that there was no evidence of small study effects or publication bias. The detailed results are shown in Supplemental Fig. 5.

Discussion

This NMA represents the most comprehensive analysis of currently available data regarding the efficacy and safety of TXA in patients undergoing cardiac surgery [Fig. 2]. The analysis yield five key findings. First, intravenous TXA, regardless of dose or administration regimen, significantly reduces postoperative blood loss and the rate of red blood cell transfusion. Furthermore, in high-risk patients, high-dose continuous TXA further reduces the risk of red blood cell transfusion compared to low-dose continuous TXA. Second, topical TXA does not significantly reduce the rate of red blood cell transfusion; on the contrary, compared to intravenous administration, topical administration is associated with a higher risk of red blood cell transfusion. Third, high-dose continuous TXA, when compared to low-dose continuous administration, further reduces the rates of reoperation and platelet transfusion. Fourth, neither intravenous nor topical TXA administration increases the incidence of postoperative thrombotic complications or mortality. In contrast, topical administration and low-dose single administration are associated with reduced mortality rates. Fifth, high-dose single TXA significantly increases the risk of postoperative seizures. Based on the aforementioned results, we have developed a practical flowchart to recommend the optimal dose of TXA for administration in cardiac surgery with different risk factors (see in Fig. 2).

Extracorporeal circulation induces fibrinolysis through plasmin activation [32–35]. Tranexamic acid functions as a competitive inhibitor of plasminogen and exhibits noncompetitive inhibition of plasmin at higher concentrations [32, 33]. Initial research exploring TXA’s antifibrinolytic effects employed a wide dosage range (20 mg/kg to 20 g) [7]. In 1995, Horrow et al. [7] conducted a pivotal prospective randomized double-blind study to establish the optimal TXA dosage in cardiac surgery. Their findings demonstrated that a 10 mg/kg loading dose followed by 1 mg/kg/h maintenance infusion for 12 h postoperatively achieved comparable hemostatic effect to regimens using two- or four-fold higher doses. Subsequent research comparing the Horrow protocol with an intensified regimen (6.6 mg/kg loading dose, 6 mg/kg/h maintenance infusion, and 40 mg CPB circuit priming) revealed no significant differences in postoperative mediastinal blood loss or transfusion requirements between groups [36]. However, Shi et al. [6] reported superior efficacy of continuous high-dose TXA infusion in reducing postoperative blood transfusions compared to low-dose regimens. This ongoing controversy regarding the optimal TXA dosage in cardiac surgery warrants further investigation. Our study demonstrated the clinical benefit of a higher-dose bolus followed by continuous infusion during cardiac procedures. While our primary endpoint (proportion of patients requiring red-cell transfusion) showed no significant difference between regimens, the higher-dose protocol significantly reduced fresh frozen plasma transfusion requirements and reoperation rates for hemostasis control. These secondary outcomes carry substantial clinical implications. Notably, existing pharmacodynamic data remain limited, with all published studies referencing a single source to support target plasma concentrations exceeding10 µg/ml [10]. Comprehensive investigations are imperative to establish evidence based TXA dosing guidelines for cardiac surgery.

The determination of optimal TXA dosage in cardiac surgery necessitates careful consideration of patient-specific factors and transfusion risk stratification. Pharmacokinetic analysis in chronic kidney disease (CKD) patients demonstrated that reduced TXA doses (25-30 mg/kg bolus) achieve plasma concentrations sufficient for complete fibrinolysis suppression [9] while minimizing complication risks associated with elevated TXA levels in this population [37, 38]. Given the variability in body surface area (BSA) among patients, it is unlikely that a single specific dose of TXA will be universally applicable. Individualized dosing strategies, including BSA-adjusted regimens, should be considered to optimize therapeutic efficacy and safety [39]. Future research should investigate the pharmacokinetic and pharmacodynamic profiles of TXA in relation to BSA and other patient-specific factors to inform evidence-based dosing guidelines [40, 41]. Current evidence supports differential dosing strategies based on transfusion risk: high-risk patients benefit from intensive TXA regimens (30 mg/kg bolus + 16 mg/kg/h infusion), whereas low-risk patients achieve adequate hemostasis with conservative dosing (10 mg/kg bolus + 1-2 mg/kg/h infusion) [6, 38]. Our subgroup analysis, stratified by transfusion risk, revealed comparable overall transfusion rates between high-dose and low-dose regimens in low-risk patients. However, continuous high-dose TXA administration demonstrated superior efficacy in reducing transfusion requirements specifically among high-risk patients. This differential response may be attributed to pharmacokinetic variations. While previous studies established that low-dose TXA regimens, such as those proposed by Horrow et al., achieve therapeutic concentrations sufficient for antifibrinolytic effects in low-risk patients [7], these concentrations progressively decline during surgery, potentially compromising efficacy in high-risk scenarios [8]. Maintenance of consistent therapeutic TXA levels throughout cardiac surgery proves crucial for high-risk patients, necessitating augmented dosing regimens. Future advancements incorporating viscoelastic testing, fibrinolytic pattern analysis, and therapeutic concentration monitoring may facilitate the development of personalized perioperative antifibrinolytic strategies in cardiac surgery [8].

Previous research has demonstrated that a single bolus of tranexamic acid may prove insufficient for patients undergoing prolonged cardiac procedures [42]. In contrast, continuous tranexamic acid infusion maintains more stable plasma concentrations with reduced peak levels compared to single-dose administration, potentially enhancing antifibrinolytic efficacy while minimizing adverse effects [42]. Our findings indicate that tranexamic acid demonstrates hemostatic efficacy in cardiac surgery through both single-bolus and continuous administration. However, high-dose continuous tranexamic acid administration exhibits superior antifibrinolytic properties, significantly reducing red cell transfusion requirements relative to single-dose regimens. Notably, single high-dose tranexamic acid administration correlates with increased postoperative seizure incidence, while high-dose continuous administration shows no significant association with increased seizure risk.

Intrapericardial administration of TXA has been proposed as a targeted delivery strategy to potentially mitigate postoperative hemorrhage [21]. Preliminary studies in cardiac surgery have demonstrated that topical TXA application reduces bleeding [43, 44], with a meta-analysis of seven trials confirming decreased 24-hour postoperative blood loss [45]. However, transfusion events were insufficiently frequent to establish definitive conclusions. The DEPOSITION trial, as the first large-scale study directly comparing topical and intravenous TXA administration in cardiac surgery, revealed contrasting findings to previous smaller trials, indicating an increased transfusion risk with topical TXA administration compared to intravenous delivery [18]. This phenomenon may be attributed to lower plasma TXA concentrations upon intensive care unit (ICU) admission following topical administration, with the antifibrinolytic effect diminishing rapidly during the initial ICU hours [17]. While Lamy et al. hypothesized that topical TXA administration might reduce seizure risk through lower mean plasma concentrations compared to intravenous administration [18], the study was prematurely terminated due to safety concerns before reaching the target sample size, failing to demonstrate significant differences in seizure incidence. Our findings similarly revealed no significant reduction in seizure incidence with topical TXA administration compared to intravenous administration. Given that the available data primarily stem from the DEPOSITION trial, definitive conclusions cannot be drawn from our study.

There is increasing concern regarding the safety profile of TXA [46], particularly in the context of its widespread use following the discontinuation of aprotinin. This shift has underscored certain limitations of TXA, especially with respect to neurological morbidity [47]. To address these concerns and elucidate the safety of TXA in cardiac surgery, Myles et al. conducted a prospective, randomized controlled trial comparing TXA to placebo—the Aspirin and Tranexamic Acid for Coronary Artery Surgery (ATACAS) trial [3]. The findings demonstrated that TXA significantly reduced the need for blood transfusions (P < 0.001) and re-exploration (P = 0.001) in patients undergoing coronary artery bypass grafting (CABG) compared to placebo. The ATACAS trial also sought to evaluate the potential prothrombotic effects of TXA, aiming to provide novel insights into its safety profile. Notably, the study found no significant difference in a composite endpoint of death and thrombotic complications—including myocardial infarction, stroke, pulmonary embolism, renal failure, and bowel infarction—within30 days postoperatively (P = 0.22). Furthermore, a one-year follow-up analysis confirmed no difference in graft patency rates between the TXA and placebo groups. In our study, we observed that the use of TXA, regardless of dosage or administration method, did not increase the risk of thrombotic complications or postoperative mortality. However, it is important to note that the incidence of thrombotic complications and mortality was notably low across the included clinical studies. Additionally, the majority of these studies were not primarily designed to evaluate the impact of TXA on thrombotic events or mortality in cardiac surgery patients. The limited sample size and low event rates in our study preclude robust statistical analysis of adverse events. Therefore, no conclusive inferences can be made from our results concerning the safety of TXA in relation to thrombotic complications or mortality. Future research with larger sample sizes and studies specifically designed to evaluate these outcomes is warranted to further clarify the safety profile of TXA in cardiac surgery.

Seizures are a well-known adverse event of TXA [11, 48]. There is a dose-dependent reported incidence of seizures ranging from 2.7 to 7.6% [13, 49, 50], with doses of TA of 100 mg/kg and greater linked to an increased risk of seizures [48]. Most of these studies used a single bolus of TXA. Having a seizure is not trivial, particularly following cardiac surgery: patients who experience seizures are more likely to get postoperative neurological complications, such as stroke or delirium, increased length of stay in intensive care unit, and increased intensive care unit mortality [12].

Limitations

This network meta-analysis is subject to several limitations that merit careful consideration. First, Residual risk of adverse events. Despite the large sample size, the heterogeneity in patient comorbidities and surgical factors may contribute to the residual risk of bleeding or thrombosis. This highlights the need for individualized approaches to TXA dosing and further research in specific subpopulations. Second, Insufficient data to draw definitive conclusions. The incidence of mortality and thromboembolic complications in the included studies was notably low and the majority of the included studies were not specifically designed or powered to evaluate mortality or thromboembolic events as primary endpoints. These may limit the ability to draw definitive conclusions regarding the impact of TXA on mortality and thromboembolic outcomes. Furthermore, the heterogeneity in study designs, patient populations, and TXA dosing regimens across the included trials may further confound the interpretation of these results. Third, Potential selection bias. Potential selection bias may have been introduced by excluding RCTs with sample sizes below 30 participants.

Conclusion

This meta-analysis further confirms that intravenous TXA administration, regardless of dose or administration regimen, significantly reduces postoperative blood loss and red blood cell transfusions in adult cardiac surgery, without increasing the incidence of serious adverse events except for seizures. Future studies should incorporate patient-specific factors, comorbidities, and bleeding risks to determine the optimal TXA dosing strategy that balances risks and benefits.

Abbreviations

TXA

Tranexamic acid

CPB

Cardiopulmonary bypass

HD-CI

High-dose continuous tranexamic acid infusion

LD-CI

Low-dose continuous tranexamic acid infusion

HD-Bolus

High-dose bolus of tranexamic acid

LD-Bolus

Low-dose bolus of tranexamic acid

SUCRA

Surface under the cumulative ranking

CABG

Coronary artery bypass grafting

NMA

Network meta-analysis

CKD

Chronic kidney disease

Authors’ contributions

XP: This author performed the search, screened the search results and separately collected the reasons for inclusion or exclusion, evaluated the risk of bias in the included studies and draft the original manuscript. MT: This author screened the search results and extracted data independently using an electronic data extraction form. ZX: This author extracted and analyzed data independently. HY: This author evaluated the risk of bias in the included studies and analyzed data independently. JH: This author helped design the study, supervise the data collection, and revise the manuscript. PL: This author helped design the study, acquire funding, supervise the data collection, and revise the manuscript.

Funding

This work was supported by the Natural Science Foundation of the Sichuan Province (grant numbers 2023NSFSC0676).

Data availability

All data generated during this study are included in the text and supplementary materials. Available upon request from the corresponding author at liangpengwch@scu.edu.cn.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.