Association of Intravenous Tranexamic Acid With Thromboembolic Events and Mortality

Key Points

Question

Is intravenous administration of tranexamic acid associated with thromboembolic events in patients of all ages and of any medical discipline?

Findings

In this systematic review and meta-analysis of 216 studies of 125 550 patients undergoing surgical procedures and receiving either intravenous administration of tranexamic acid or placebo or no treatment, 1020 (2.1%) thromboembolic events in the tranexamic acid group and 900 (2.0%) total thromboembolic events in the control group were found. There was no increased risk of any thromboembolic event in patients of all medical disciplines.

Meaning

These results clarify whether vascular occlusive events are associated with administration of tranexamic acid.

Abstract

Importance

Tranexamic acid (TXA) is an efficient antifibrinolytic agent; however, concerns remain about the potential adverse effects, particularly vascular occlusive events, that may be associated with its use.

Objective

To examine the association between intravenous TXA and total thromboembolic events (TEs) and mortality in patients of all ages and of any medical disciplines.

Data Source

Cochrane Central Register of Controlled Trials and MEDLINE were searched for eligible studies investigating intravenous TXA and postinterventional outcome published between 1976 and 2020.

Study Selection

Randomized clinical trials comparing intravenous TXA with placebo/no treatment. The electronic database search yielded a total of 782 studies, and 381 were considered for full-text review. Included studies were published in English, German, French, and Spanish. Studies with only oral or topical tranexamic administration were excluded.

Data Extraction and Synthesis

Meta-analysis, subgroup and sensitivity analysis, and meta-regression were performed. This study followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline.

Main Outcomes and Measures

Vascular occlusive events and mortality.

Results

A total of 216 eligible trials including 125 550 patients were analyzed. Total TEs were found in 1020 (2.1%) in the group receiving TXA and 900 (2.0%) in the control group. This study found no association between TXA and risk for total TEs (risk difference = 0.001; 95% CI, −0.001 to 0.002; P = .49) for venous thrombosis, pulmonary embolism, venous TEs, myocardial infarction or ischemia, and cerebral infarction or ischemia. Sensitivity analysis using the risk ratio as an effect measure with (risk ratio = 1.02; 95% CI, 0.94-1.11; P = .56) and without (risk ratio = 1.03; 95% CI, 0.95-1.12; P = .52) studies with double-zero events revealed robust effect size estimates. Sensitivity analysis with studies judged at low risk for selection bias showed similar results. Administration of TXA was associated with a significant reduction in overall mortality and bleeding mortality but not with nonbleeding mortality. In addition, an increased risk for vascular occlusive events was not found in studies including patients with a history of thromboembolism. Comparison of studies with sample sizes of less than or equal to 99 (risk difference = 0.004; 95% CI, −0.006 to 0.014; P = .40), 100 to 999 (risk difference = 0.004; 95% CI, −0.003 to 0.011; P = .26), and greater than or equal to 1000 (risk difference = −0.001; 95% CI, −0.003 to 0.001; P = .44) showed no association between TXA and incidence of total TEs. Meta-regression of 143 intervention groups showed no association between TXA dosing and risk for venous TEs (risk difference, −0.005; 95% CI, −0.021 to 0.011; P = .53).

Conclusions and Relevance

Findings from this systematic review and meta-analysis of 216 studies suggested that intravenous TXA, irrespective of dosing, is not associated with increased risk of any TE. These results help clarify the incidence of adverse events associated with administration of intravenous TXA and suggest that TXA is safe for use with undetermined utility for patients receiving neurological care.

This systematic review and meta-analysis with meta-regression of 216 randomized clinical trials assesses the association between intravenous administration of tranexamic acid and vascular occlusive events in patients undergoing surgery or experiencing bleeding.

Introduction

Major surgery is commonly associated with substantial blood loss, subsequent anemia, and the need for blood transfusion. The Vascular Events in Noncardiac Surgery Patients Cohort Evaluation (VISION) study investigators¹ reported that among 40 004 patients undergoing noncardiac surgical procedures, the most common complications leading to death were major bleeding followed by myocardial injury and infection. Tranexamic acid (TXA) is an antifibrinolytic agent and widely used for prophylaxis and treatment of bleeding caused by hyperfibrinolysis. Poeran and colleagues² reported that the use of TXA increased in orthopedic surgery from almost 0% in 2006 to 11.2% in 2012, and the effectiveness to reduce surgical blood loss and associated complications has been reported.

Ker and colleagues³ performed a meta-analysis including 129 trials encompassing more than 10 000 patients suggesting that administration of TXA is associated with reductions in allogeneic blood transfusion by 38% and that further trials assessing this effect are unlikely to add new insights. However, TXA is only applied with caution because antifibrinolytic therapy may be associated with an increased risk of seizures,^4,5 myocardial infarction,⁶ and other thrombotic complications.^7,8,9 Nevertheless, the association of vascular occlusive events with TXA administration is controversial. Overall, vascular occlusive events are rare, and studies with 0 events are often excluded from meta-analysis because the assumption is that these studies may not be relevant to the treatment effect. Of 115 investigated trials in the meta-analysis by Ker and colleagues,³ 72 examined the rate of pulmonary embolism (PE) and 66 studies examined the rate of deep vein thrombosis (DVT). However, trials with 0 events were excluded from analysis. Overall, TXA was not associated with an increased risk of PE in 10 trials (risk ratio [RR] = 0.61; 95% CI, 0.25-1.47) (event rate in TXA group: 4 of 449 vs control: 8 of429) and DVT in 19 trials (RR = 0.86; 95% CI, 0.53-1.39) (event rate in TXA group: 25 of 785 vs control: 29 of 785). Based on the low number of included event rates, the issue of use of TXA and vascular occlusive events remains unaddressed.³ Several guidelines recommend the use of TXA in patients with excessive bleeding.^10,11,12,13 However, little is known about the incidence of vascular occlusive events in patients with substantial comorbidities or a history of thromboembolic events (TEs). In addition, debate is ongoing about the optimal perioperative dosing of intravenous TXA, which varies widely, ranging from 0.5 to 5 g or 10 to 100 mg/kg and might also explain the different observed incidences of vascular occlusive events.

To further explore the possible associations between intravenous TXA and vascular occlusive events in patients undergoing surgery or experiencing bleeding, we performed a comprehensive meta-analysis. We included randomized clinical trials (RCTs) irrespective of event rate and dosing regimen comparing intravenous TXA with a control group (placebo or no treatment) in our analysis. These data might help to clarify the safety of intravenous TXA and elucidate a possible dosing effect.

Methods

Search Strategy and Study Selection

We systematically searched Cochrane Central Register of Controlled Trials and MEDLINE via PubMed for eligible RCTs investigating the effect of intravenous TXA on postinterventional outcome published between 1976 and 2018, followed by a manual search through September 20, 2019. To identify RCTs published after completion of the meta-analysis, another systematic review was performed for eligible studies published between July 1, 2018, and December 31, 2020. Search terms used and additional details are available in eAppendix 1 in the Supplement.

Outcome Measures

End points were venous thrombosis (VT), PE, venous thromboembolic events (VTEs), myocardial infarction or ischemia (MI), cerebral infarction or ischemia (CII), limb ischemia, mesenteric ischemia, hepatic artery thrombosis, and composite of all vascular occlusive events (total thromboembolic events [total TEs]). In addition, we assessed overall mortality, bleeding mortality, and any nonbleeding mortality rate (eAppendix 2 in the Supplement).

Data Extraction and Statistical Analyses

This study was conducted in accordance with the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline for cohort studies.¹⁴ Because many studies showed 0 events in both groups, we assessed the risk difference (RD) to provide accurate results. The fixed-effect analysis was used for meta-analysis because we assumed that the true effect size was the same in all studies because TEs are rare and the only reason the effect size varied between studies could be caused by the number of recruited patients. However, we performed sensitivity analysis using a random-effects model to estimate whether our results were robust. Furthermore, we performed a sensitivity analysis using the RR as an effect size measure with either including (continuity correction of 0.5) or excluding studies with 0 cell frequencies. An additional analysis of total TEs with subgroups by study size was conducted. Heterogeneity was assessed by using I² statistics. A meta-regression was performed to investigate a relation between the event rate and the dosage of intravenous TXA.

We also investigated whether intravenous TXA was associated with increased risk for TE, VT, PE, VTEs, and overall mortality in patients with risks for TEs. Therefore, secondary analyses were performed only with studies including patients with a history of any TE, coronary artery disease, thrombophilia, or contraindication for TXA. Two of us (I.T. and S.C.) independently assessed the methodologic quality of included studies based on the Cochrane Risk of Bias tool. A sensitivity analysis with studies judged at low risk of selection bias was assessed for total TEs. Funnel plots were generated to detect possible evidence for small-study bias. Discrepancies were resolved by group discussion (I.T., S.C., P.M., E.H., and S.W.). Two-sided P < .05 was considered statistically significant. The Review Manager (RevMan) program, version 5.3 (The Nordic Cochrane Centre) and R software, version 3.6.1 (R Foundation for Statistical Computing) were used for analysis and graphic illustrations. Continuity corrections were performed with the R package, meta version 4.12-0. Further details appear in eAppendix 3 in the Supplement.

Results

The electronic database and manual search yielded a total of 782 studies published between 1976 and 2018. In total, 381 articles were considered for full-text review, of which 189 were excluded because of ineligible study designs (n = 55) or control groups (n = 92), ineligible end points (n = 27), missing or nontranslatable full text (n = 11), and duplicates (n = 4), leaving 192 RCTs^{6,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199,200,201,202,203,204,205} for final analysis (Figure). Included studies were published in English, German, French, and Spanish. Twenty-three authors were contacted, of whom 2 provided clarification¹¹³ or data,¹⁸⁵ whereas 21 did not reply. In total, 68 118 patients of this search were included in this meta-analysis (34 610 TXA group and 33 508 control group). Demographic data and subgroup characteristics are reported in eTable 1 and eAppendix 4 in the Supplement.

Figure. Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) Flow Diagram Showing Study Selection.

The database search was conducted for articles published between 1976 and 2018, and the manual search was conducted through September 20, 2019. An updated search for studies published between July 1, 2018, and December 31, 2020, resulted in 72 potential additional studies of which 48 studies were excluded leaving 24 additional studies. Overall, 216 eligible studies underwent analysis. RCT indicates randomized clinical trial; TXA, tranexamic acid.

Total Thromboembolic Events

In total, 176 studies (eAppendix 11 in the Supplement) provided data on total TEs in 65 900 patients (n = 33 487 TXA group vs n = 32 413 control group). We found 779 total TEs (2.3%) in the TXA group and 706 total TEs (2.2%) in the control group. Overall, administration of intravenous TXA was not associated with an increased risk for total TEs (RD = 0.001; 95% CI, −0.002 to 0.003; P = .66) (Table 1; eFigure 1 in the Supplement). However, the subgroup analysis showed that TXA was associated with a significantly increased risk for total TEs in the group of patients with neurological conditions (RD = 0.026; 95% CI, 0.007-0.045; P = .007). Sensitivity analysis using a random-effects model showed robust effect estimates for total TEs (RD = −0.001; 95% CI, −0.002 to 0.001; P = .39), but no significantly increased risk for total TEs in the subgroup of patients with neurological conditions was shown (RD = 0.018; 95% CI, −0.013 to 0.048; P = .26) (Table 1; eFigure 2 in the Supplement). Analysis of patients with neurological conditions showed a significant heterogeneity (I² = 57%). To investigate whether the heterogeneity was associated with sample size, we performed a sensitivity analysis including studies with sample sizes with more than 1000 patients. Overall, the heterogeneity remained high (I² = 73%). A sensitivity analysis was performed using the RR as an effect measure with and without studies with double-zero events revealed robust effect estimates for all subgroups (RR = 1.01; 95% CI, 0.92-1.11; P = .77 and RR = 1.02; 95% CI, 0.93-1.12; P = .71) (eFigure 3 and eFigure 4 in the Supplement). To elucidate whether incidence of total TEs increases with sample size, we performed a sensitivity analysis of studies including less than or equal to 99 patients, 100 to 999 patients, or greater than or equal to 1000 patients. Overall, administration of intravenous TXA was not associated with an increased risk for total TEs in studies with less than or equal to 99 patients (RD = 0.004; 95% CI, −0.006 to 0.014; P = .40), 100 to 999 patients (RD = 0.004; 95% CI, −0.003 to 0.011; P = .26), and greater than or equal to 1000 patients (RD = −0.001; 95% CI, −0.003 to 0.001; P = .44) (eFigure 5 in the Supplement). The results remained robust using a random-effects model (eFigure 6 in the Supplement).

Table 1. TXA and Total Thromboembolic Events.

Medical discipline	No. of included studies	TXA		Control		Model	Risk difference (95% CI)	P value	I2, %
		Events	No. of included patients	Events	No. of included patients
Cardiothoracic	16	72	3171	74	3009	Fixed effect	−0.001 (−0.009 to 0.007)	.83	0
						Random effects	−0.001 (−0.007 to 0.008)	.91
Neurological	12	282	2007	230	2000	Fixed effect	0.026 (0.007 to 0.045)	.01	57
						Random effects	0.018 (−0.013 to 0.048)	.26
Gynecological	26	35	12 356	41	12 286	Fixed effect	−0.001 (−0.002 to 0.001)	.53	0
						Random effects	−0.001 (−0.002 to 0.001)	.50
Orthopedic	101	172	4787	113	4149	Fixed effect	0.001 (−0.007 to 0.009)	.79	0
						Random effects	0.001 (−0.004 to 0.007)	.64
Major trauma	1	204	10 060	233	10 067	Fixed effect	−0.003 (−0.007 to 0.001)	.16	NA
						Random effects	−0.003 (−0.007 to 0.001)	.16
Maxillofacial	6	0	265	0	192	Fixed effect	0.000 (−0.023 to 0.023)	>.99	0
						Random effects	0.000 (−0.019 to 0.019)	>.99
Pediatric	2	0	42	0	40	Fixed effect	0.000 (−0.067 to 0.067)	>.99	0
						Random effects	0.000 (−0.064 to 0.064)	>.99
Other	12	14	799	15	670	Fixed effect	−0.004 (−0.021 to 0.013)	.62	0
						Random effects	−0.004 (−0.018 to 0.011)	.63
Total	176	779	33 487	706	32 413	Fixed effect	0.001 (−0.002 to 0.003)	.66	0
						Random effects	−0.001 (−0.002 to 0.001)	.39

Abbreviations: NA, not applicable; TXA, tranexamic acid.

Venous Thrombosis, Pulmonary Embolism, and Venous Thromboembolic Events

In total, 163 studies (eAppendix 12 in the Supplement) provided data on VTs in 59 666 patients (n = 30 334 TXA group vs n = 29 332 control group). We found 272 VTs (0.9%) in the TXA group and 213 VTs (0.7%) in the control group. Overall, administration of intravenous TXA was not associated with an increased risk for VT (RD = −0.000; 95% CI, −0.002 to 0.002; P > .99) (eTable 2, eFigure 7 in the Supplement). Sensitivity analysis using a random-effects model showed robust effect estimates for PE (RD = −0.000; 95% CI, −0.001 to 0.000; P = .26) (eTable 2; eFigure 8 in the Supplement).

In total, 129 studies (eAppendix 13 in the Supplement) provided data on PE events in 61 562 patients (n = 31 155 TXA group vs n = 30 407 control group). We found 152 PE events (0.5%) in the TXA group and 153 PE events (0.5%) in the control group. Overall, administration of intravenous TXA was not associated with an increased risk for PE (RD = −0.000; 95% CI, −0.001 to 0.001; P = .89) (eTable 3; eFigure 9 in the Supplement). Sensitivity analysis using a random-effects model showed robust effect estimates for VT (RD = −0.000; 95% CI, −0.001 to 0.001; P = .68) (eTable 3; eFigure 10 in the Supplement).

To assess the total number of VTEs, PE and VT were combined and analyzed. In total, 123 studies (eAppendix 14 in the Supplement) provided data on VTEs in 56 126 patients (n = 28 438 TXA group vs n = 27 688 control group). We found 348 VTEs (1.2%) in the TXA group and 304 VTEs (1.1%) in the control group. Overall, administration of intravenous TXA was not associated with an increased risk for VTEs (RD = −0.000; 95% CI, −0.002 to 0.002; P = .71) (eTable 4; eFigure 11 in the Supplement). Sensitivity analysis using a random-effects model showed robust effect estimates for VTE (RD = −0.001; 95% CI, −0.002 to 0.001; P = .39) (eTable 4; eFigure 12 in the Supplement).

Myocardial Infarction or Ischemia, Cerebral Infarction or Ischemia, and Other Thromboembolic Events

Overall, administration of intravenous TXA was not associated with an increased risk for MI (RD = −0.000; 95% CI, −0.001 to 0.001; P = .56) (eTable 5; eFigure 13 and eFigure 14 in the Supplement). Detailed analyses are reported in eAppendix 5 in the Supplement.

Overall, administration of intravenous TXA was not associated with an increased risk for CII (RD = −0.000; 95% CI, −0.001 to 0.000; P = .90) (eTable 6; eFigure 15 in the Supplement) or other TEs (eFigures 16-18 in the Supplement). Detailed analyses are reported in eAppendix 6 in the Supplement.

Overall Mortality

In total, 63 studies (eAppendix 15 in the Supplement) assessed the overall mortality in 55 305 patients (n = 27 865 TXA group vs n = 27 440 control group). Death occurred in 2218 patients (8%) in the TXA group and 2456 patients (9%) in the control group. Overall, administration of intravenous TXA was associated with significant reductions in overall mortality in patients of the TXA group (RD = −0.011; 95% CI, −0.015 to −0.007; P < .001) (Table 2) (eFigure 19 in the Supplement). Subgroup analysis showed a significantly decreased overall mortality in patients with trauma (RD = −0.015; 95% CI, −0.022 to −0.008; P = .004) and patients receiving care from other disciplines (RD = −0.038; 95% CI, −0.06 to −0.015; P = .001) for the TXA groups, whereas no significant differences could be detected within the remaining medical disciplines. Sensitivity analysis using a random-effects model did not show robust effect estimates for overall mortality (RD = −0.004; 95% CI, −0.008 to 0.000; P = .05) and for the subgroup of patients receiving care from other disciplines (RD = −0.024; 95% CI, −0.058 to 0.009; P = .15) (Table 2; eFigure 20 in the Supplement). Analysis of patients of other disciplines showed a significant heterogeneity (I² = 78%).

Table 2. TXA and Overall Mortality.

Medical discipline	No. of included studies	TXA		Control		Model	Risk difference (95% CI)	P value	I2, %
		Events	No. of included patients	Events	No. of included patients
Cardiothoracic	15	32	3006	46	2970	Fixed effect	−0.005 (−0.011 to 0.002)	.12	0
						Random effects	−0.003 (−0.009 to 0.003)	.30
Neurological	13	426	2017	449	2002	Fixed effect	−0.013 (−0.032 to 0.005)	.31	27
						Random effects	−0.016 (−0.041 to 0.02)	.28
Gynecological	8	227	10 871	256	10 814	Fixed effect	−0.003 (−0.007 to 0.001)	.17	0
						Random effects	−0.002 (−0.005 to 0.002)	.31
Orthopedic	16	18	844	17	652	Fixed effect	0.001 (−0.018 to 0.019)	.94	0
						Random effects	−0.002 (−0.015 to 0.011)	.76
Major trauma	1	1463	10 060	1613	10 067	Fixed effect	−0.015 (−0.022 to −0.008)	.004	NA
						Random effects	−0.015 (−0.022 to −0.008)	.004
Pediatric	1	0	40	0	42	Fixed effect	0.000 (−0.046 to 0.046)	>.99	NA
						Random effects	0.000 (−0.047 to 0.047)	>.99
Other	9	52	1027	75	893	Fixed effect	−0.038 (−0.06 to −0.015)	.001	78
						Random effects	−0.024 (−0.058 to 0.009)	.15
Total	63	2218	27 865	2456	27 440	Fixed effect	−0.011 (−0.015 to −0.007)	<.001	16
						Random effects	−0.004 (−0.008 to 0.000)	.05

Abbreviations: NA, not applicable; TXA, tranexamic acid.

Nonbleeding Mortality and Bleeding Mortality

In total, 48 studies (eAppendix 16 in the Supplement) assessed nonbleeding mortality in 46 619 patients (n = 23 458 TXA group vs n = 23 161 control group). Death occurred in 1180 patients (5%) in the TXA group and 1228 patients (5%) in the control group. Overall, administration of intravenous TXA was not associated with a decreased risk for nonbleeding mortality (RD = −0.002; 95% CI, −0.006 to 0.002; P = .29) (eTable 7; eFigure 21 in the Supplement). However, subgroup analysis showed a significant increase for nonbleeding mortality in patients with neurological conditions of the TXA group (RD = 0.044; 95% CI, 0.007-0.081; P = .02), whereas the subgroup analysis of cardiothoracic surgery showed a significant decrease for nonbleeding mortality in patients of the TXA group (RD = −0.025; 95% CI, −0.045 to −0.005; P = .02). Sensitivity analysis using a random-effects model showed robust effect estimates for nonbleeding mortality (RD = −0.000; 95% CI, −0.002 to 0.001; P = .92) but not for the subgroup of patients with neurological conditions (RD = 0.021; 95% CI, −0.014 to 0.057; P = .24) and cardiothoracic surgery (RD = −0.015; 95% CI, −0.031 to 0.002; P = .10) (eTable 7; eFigure 22 in the Supplement).

In total, 49 studies (eAppendix 17 in the Supplement) assessed the bleeding mortality in 46 702 patients (n = 23 501 TXA group vs n = 23 201 control group). Death occurred in 692 patients (3%) in the TXA group and 874 patients (4%) in the control group. Overall, administration of intravenous TXA was associated with an overall significant decrease of bleeding mortality (RD = −0.008; 95% CI, −0.011 to −0.005; P < .001) (Table 3; eFigure 23 in the Supplement). Subgroup analysis showed a significantly decreased bleeding mortality in patients with neurological conditions (RD = −0.071; 95% CI, −0.102 to −0.041; P < .001), patients with trauma (RD = −0.008; 95% CI, −0.015 to −0.002; P = .008), and patients of any other disciplines of the TXA group (RD = −0.018; 95% CI, −0.033 to −0.004; P = .02), whereas no significant differences could be detected within the remaining medical disciplines (cardiothoracic, gynecological, orthopedic, and pediatric). Sensitivity analysis using a random-effects model showed robust effect estimates for bleeding mortality (RD = −0.004; 95% CI, −0.008 to −0.001; P = .02) but not for the subgroup of other disciplines (RD = −0.01; 95% CI, −0.028 to −0.009; P = .30) (Table 3; eFigure 24 in the Supplement). Significant heterogeneity was detected for patients with neurological conditions (I² = 60%).

Table 3. TXA and Bleeding Mortality.

Medical discipline	No. of included studies	TXA		Control		Model	Risk difference (95% CI)	P value	I2, %
		Events	No. of included patients	Events	No. of included patients
Cardiothoracic	12	0	543	1	478	Fixed effect	−0.002 (−0.016 to −0.012)	.77	0
						Random effects	−0.004 (−0.012 to 0.011)	.94
Neurological	8	43	685	91	678	Fixed effect	−0.071 (−0.102 to −0.041)	<.001	60
						Random effects	−0.056 (−0.11 to −0.002)	.04
Gynecological	8	155	10 871	191	10 814	Fixed effect	−0.003 (−0.007 to −0.000)	.05	0
						Random effects	−0.002 (−0.005 to 0.001)	.12
Orthopedic	13	0	647	0	461	Fixed effect	0.000 (−0.014 to 0.014)	.77	0
						Random effects	0.000 (−0.013 to 0.013)	>.99
Major trauma	1	489	10 060	574	10 067	Fixed effect	−0.008 (−0.015 to −0.002)	.01	NA
						Random effects	−0.008 (−0.015 to −0.002)	.01
Pediatric	1	0	40	0	42	Fixed effect	0.000 (−0.046 to 0.046)	>.99	NA
						Random effects	0.000 (−0.046 to 0.046)	>.99
Other	6	5	655	17	661	Fixed effect	−0.018 (−0.033 to −0.004)	.02	53
						Random effects	−0.01 (−0.028 to −0.009)	.30
Total	49	692	23 501	874	23 201	Fixed effect	−0.008 (−0.011 to −0.005)	<.001	9
						Random effects	−0.004 (−0.008 to −0.001)	.02

Abbreviations: NA, not applicable; TXA, tranexamic acid.

Patients With Risks for Thromboembolic Events

Overall, administration of intravenous TXA was not associated with an increased risk for total TEs (RD = −0.000; 95% CI, −0.008 to 0.009; P > .99) (eFigure 25 in the Supplement), for VT (RD = 0.003; 95% CI, −0.007 to 0.013; P = .57) (eFigure 26 in the Supplement), for PE (RD = −0.001; 95% CI, −0.009 to 0.007; P = .73) (eFigure 27 in the Supplement), or for VTEs (RD = −0.000; 95% CI, −0.012 to 0.01; P = .89) (eFigure 28 in the Supplement). Administration of intravenous TXA was associated with significant reductions in overall mortality (RD = −0.038; 95% CI, −0.057 to −0.018; P < .001). Detailed analyses are reported in eAppendix 7, eFigure 29 and eFigure 30, and eTable 8 in the Supplement.

Meta-regression

A meta-regression of 143 intervention groups was conducted to assess a possible association between different intravenous dosages of TXA and VTE rate. Results from this analysis showed no association between total dosing (RD = −0.005; 95% CI, −0.021 to 0.011; P = .53), single dosing (RD = 0.018; 95% CI, −0.053 to 0.09; P = .60), or any dose of intravenous TXA (RD = −0.005; 95% CI, −0.013 to 0.003; P = .21) and incidence of VTEs. Detailed analyses are reported in eAppendix 8 in the Supplement.

Risk of Bias

Overall, 139 studies (72%) were judged with low risk and 10 (5%) at high risk for random sequence generation. The risk of bias in the remaining 43 studies (22%) was unclear because of insufficient information. Allocation was adequately concealed in 68 studies (35%), whereas 4 studies (2%) were judged at high risk because patients were not randomly assigned to intravenous TXA or the control group. Sensitivity analysis with studies judged at low risk for selection bias was performed for total TEs and showed that results remained robust (RD = -0.001, 95% CI, −0.002 to 0.003, P = .89) (eFigure 33 in the Supplement). Detailed analysis including funnel plots is reported in eAppendix 9 and in eFigures 31-36 in the Supplement.

Updated Meta-analysis

A systematic search was performed to identify RCTs published between July 1, 2018, and December 31, 2020. In total, 72 studies were considered for full-text review, of which 48 were excluded because of ineligible study designs or control groups (n = 35), ineligible end points (n = 5), missing or nontranslatable full text (n = 4), and duplicates (n = 4), leaving 24 RCTs^{206,207,208,209,210,211,212,213,214,215,216,217,218,219,220,221,222,223,224,225,226,227,228,229} including 27 888 patients (14 242 TXA group and 13 646 control group) for analysis (eTable 9 in the Supplement).

Of all trials (n = 216) comprising 125 550 patients, we found 1020 total TEs (2.1%) in the TXA group and 900 total TEs (2.0%) in the control group. The addition of 24 trials published after our analysis was completed showed that effect estimates for total TEs (RD = 0.001; 95% CI, −0.001 to 0.002; P = .49), VT (RD = −0.000; 95% CI, −0.001 to 0.001; P = .85), PE (RD = 0.000; 95% CI, −0.001 to 0.001; P = .74), VTE (RD = 0.000; 95% CI, −0.001 to 0.002; P = .85), and overall mortality (RD = −0.007; 95% CI, −0.012 to −0.004; P < .001) remained robust (eTables 10-14; eFigures 37-46 in the Supplement). Sensitivity analysis for total TEs using the RR as an effect measure including studies with (RR = 1.02; 95% CI, 0.94-1.11; P = .56) and without (RR = 1.03; 95% CI, 0.95-1.12; P = .52) double-zero events showed similar results.Detailed analyses are reported in eAppendix 10 and eFigures 37-48 in the Supplement.

Discussion

Tranexamic acid is efficient to reduce bleeding by inhibiting the enzymatic breakdown of existing fibrin blood clots and is therefore widely used in anesthesia and surgery. The utility of TXA was supported by the results of 3 trials (CRASH-2, WOMAN and CRASH-3), reporting its efficiency in reducing bleeding-associated deaths in patients with trauma,⁴¹ postpartum hemorrhage,⁴⁰ and traumatic brain injury.²⁰⁶ However, a significant survival benefit was achieved only when TXA was administered within the first 3 hours after injury or delivery.^40,41 On the basis of the CRASH-2 results, intravenous TXA was included in the World Health Organization list of essential medicine in 2011.²³⁰ Along with the positive effect, concerns about potential adverse effects, in particular vascular occlusive events, were raised. Controversial results were published, reporting from no incidence of TE⁴⁰ up to 12-fold higher rates for DVT⁷ after intravenous TXA administration. Yates and colleagues²³¹ performed a meta-analysis including RCTs with different application methods of TXA. Analysis of 20 679 patients revealed no increased risk for VTEs after administration of intravenous TXA compared with placebo or no intervention. No subgroup analysis was performed by the authors. In orthopedic patients undergoing surgery with or without intravenous TXA, a systematic review by Franchini and colleagues²³² encompassing 67 RCTs revealed no significant difference of VTEs measured as RD including studies with 0 events (RD, 0.0008) and RR excluding these studies (RR, 1.0411).

To assess any risk of TEs associated with administration of intravenous TXA, we performed a meta-analysis including studies from all medical disciplines that assessed and provided data for TEs. Analyses were performed using RD including studies with 0 events to allow a general conclusion. Although this method yields wide 95% CIs when events are rare, this approach provides precise information when analyzing all available evidence instead of excluding a large proportion of studies that provided double-zero events. Excluding these trials from analysis generates the risk of inflating the magnitude of the pooled treatment effect. However, we performed a sensitivity analysis using the RR as an effect measure with and without studies with double-zero events and revealed robust effect estimates. In total, 216 studies published between 1976 and 2020 encompassing 125 500 patients were included in the meta-analysis, and overall we found 1020 total TEs (2.1%) in the TXA group and 900 total TEs (2.0%) in the control group. There was no association between intravenous TXA and increased risk for vascular occlusive events. The bleeding-associated mortality was significantly reduced in intravenous TXA–treated patients compared with controls, whereas no difference in nonbleeding mortality was detected. Particularly, the use of intravenous TXA in patients experiencing major trauma and in patients with neurological conditions was associated with significant survival benefit concerning bleeding-related mortality. Notably, we found only 1 study providing data for patients with major trauma treated with intravenous TXA. An updated meta-analysis was performed to identify RCTs published recently. Our results remain unchanged when data from these trials for total TEs, VT, PE, VTE, and overall mortality were included in the meta-analysis.

Of all investigated subgroups, we found inconclusive results for patients with neurological conditions. In this subgroup, results obtained with the fixed-effect model did not remain robust when using the random-effects model for the end points of TE and nonbleeding mortality. However, we found a decrease in bleeding-associated mortality in patients of the TXA group compared with the control group. In total, 12 studies provided data for total TEs in patients with neurological conditions; study size varied between 24 and 2235. Heterogeneity remained high after omitting studies with more than 1000 patients. Overall, increased heterogeneity and asymmetry in funnel plots indicate that further trials are necessary to solve the uncertainty for patients with neurological conditions.

It is commonly held that trials with low numbers of recruited patients might be underpowered to detect an intervention effect. To address this possibility, we performed a sensitivity analysis and found that administration of intravenous TXA was not associated with an increased risk for total TEs in studies with less than or equal to 99 patients, 100 to 999 patients, and greater than or equal to 1000 patients. Patients with an increased risk for thromboembolism were often excluded from RCTs. Given the short elimination half-life of intravenous TXA, patients with a risk factor may benefit from treatment during surgery. We found that the risk for vascular occlusive events or overall mortality rate was not increased in studies including patients with increased risks for thromboembolism. To assess a possible effect of study size or risk of bias, we conducted sensitivity analyses for the primary endpoint of total TEs, with results remaining robust.

Overall, the administration dose of intravenous TXA varied widely from 0.5 to 5 g or 10 to 100 mg/kg. Moving away from one-dose-fits-all to weight-adapted dosing might be associated with the different treatment regimens applied worldwide in the context of trauma and surgery. Notably, we did not detect any dose-dependent association of TEs.

Limitations

Although this meta-analysis provides substantial data, this study has limitations. We cannot exclude that additional references might have been missed by our systematic search of databases. However, we believe that inclusion of further studies had no impact on our main findings regarding vascular occlusive events. The follow-up varied between trials, ranging from 24 hours to several months. However, the half-life of intravenous TXA is 1.9-2.7 hours.^233,234 Considering that postoperative thrombotic events occur 6 to 8 days after surgery,²³⁵ we hypothesized that intravenous TXA–related adverse events would be detected within even in a short period of follow-up. Furthermore, TEs were not examined using ultrasonographic screening; therefore, asymptomatic thrombosis might not have been detected in all cases, and the incidence of TEs might be underestimated in some studies. Low incidence of VTEs with an approximate rate of 1 per 1000 patients²³⁶ and the routine use of postoperative thrombosis prophylaxis might also be associated with a low detection rate in patients with and without administration of intravenous TXA. Many of the included studies did not provide sufficient information about thrombosis prophylaxis; therefore, the association of postoperative care and vascular occlusive events was not analyzed further.

Conclusions

Taken as a whole, this systematic review and meta-analysis did not find that intravenous treatment with TXA in patients of any medical discipline was associated with a significant increased risk for TEs irrespective of administered dose. The results of this study suggest that use of intravenous TXA may have utility in all medical fields, with some uncertainty for patients with neurological conditions.

Supplement.

eAppendix 1. Methods

eAppendix 2. Outcome Measure

eAppendix 3. Statistical Analyses

eAppendix 4. Demographic Data and Subgroup Characteristics

eAppendix 5. Effect of Tranexamic Acid on Myocardial Infarction or Ischaemia

eAppendix 6. Effect of Tranexamic Acid on Cerebral Infarction or Ischaemia and on Other Thromboembolic Events

eAppendix 7. Effect of Tranexamic Acid in Patients with risks for Thromboembolic Events

eAppendix 8. Meta-Regression

eAppendix 9. Risk of Bias.

eAppendix 10. Updated Meta-Analysis

eTable 1. Study Characteristics

eTable 2. Venous Thrombosis

eTable 3. Pulmonary Embolism

eTable 4. Venous Thrombosis + Pulmonary Embolism

eTable 5. Myocardial Infarction or Ischaemia

eTable 6. Cerebral Infarction or Ischaemia

eTable 7. Non-Bleeding Mortality

eTable8. Summary of Sensitivity Analysis including Patients with increased Risk for total Thromboembolic Events

eTable 9. Study Characteristics of Updated Meta-Analysis

eTable 10. Total Thromboembolic Events of Updated Meta-Analysis

eTable 11. Venous Thrombosis of Updated Meta-Analysis

eTable 12. Pulmonary Embolism of Updated Meta-Analysis

eTable 13. Venous Thrombosis + Pulmonary Embolism of Updated Meta-Analysis

eTable 14. Overall Mortality of Updated Meta-Analysis

eAppendix 11. References on Total TE

eAppendix 12. References on VT Events

eAppendix 13. References on PE Events

eAppendix 14. References on VTE

eAppendix 15. References on Overall Mortality

eAppendix 16. References on Non-bleeding Mortality

eAppendix 17. References on Bleeding Mortality

eFigure 1. Total Thromboembolic Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 2. Total Thromboembolic Event Rate in Patients receiving TXA compared to Control using the Random-Effects Model

eFigure 3. Total Thromboembolic Event Rate in Patients receiving TXA compared to Control analysed with Risk Ratio in Studies with Zero Events

eFigure 4. Total Thromboembolic Event Rate in Patients receiving TXA compared to Control analysed with Risk Ratio in Studies without Zero Events

eFigure 5. Total Thromboembolic Event Rate in Patients receiving TXA compared to Control according to Sample Size using the Fixed-Effect Model

eFigure 6. Total Thromboembolic Event Rate in Patients receiving TXA compared to Control according to Sample Size using the Random-Effects Model

eFigure 7. Venous Thrombosis Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 8. Venous Thrombosis Event Rate in Patients receiving TXA compared to Control using the Random-Effects Model

eFigure 9: Pulmonary Embolism Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect model

eFigure 10: Pulmonary Embolism Event Rate in Patients receiving TXA compared to Control using the Random-Effects Model

eFigure 11. Venous Thromboembolic and Pulmonary Embolism Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 12. Venous Thromboembolic and Pulmonary Embolism Event Rate in Patients receiving TXA compared to Control using the Random-Effects Model

eFigure 13. Myocardial Infarction Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 14. Myocardial Infarction Event Rate in Patients receiving TXA compared to Control using the Random-Effects Model

eFigure 15. Cerebral Ischaemia Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 16. Limb Ischaemia Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 17. Mesenteric Ischaemia Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 18. Hepatic Artery Thrombosis Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 19. Overall Mortality Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 20. Overall Mortality Rate in Patients receiving TXA compared to Control using the Random-Effects Model

eFigure 21. Non-Bleeding associated Mortality Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 22. Non-Bleeding associated Mortality Rate in Patients receiving TXA compared to Control using the Random-Effects Model

eFigure 23. Bleeding associated Mortality Rate in Patients receiving TXA compared to Control using the Fixed-Effect model

eFigure 24. Bleeding associated Mortality Rate in Patients receiving TXA compared to Control using the Random-Effects Model

eFigure 25. Total Thromboembolic Event Rate in High Risk Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 26. Venous Thrombosis Event Rate in High Risk Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 27. Pulmonary Embolism Event Rate in High Risk Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 28. Venous Thromboembolic Event Rate in High Risk Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 29. Overall Mortality Rate in High Risk Patients receiving TXA compared to Control using the Fixed-Effect Model

eFigure 30. Overall Mortality Rate in High Risk Patients receiving TXA compared to Control using the Random-Effects Model

eFigure 31. Risk of Bias Summary

eFigure 32. Risk of bias Graph

eFigure 33. TotalThromboembolic Event Rate in Patients receiving TXA compared to Control of Studies Judged with Low Selection Bias using the Fixed-Effect Model

eFigure 34. Funnel Plots for Total Thromboembolic Event, Venous Thrombosis, Cerebral Infarction, and Bleeding Mortality

eFigure 35. Funnel Plots for Pulmonary Embolism, Venous Thromboembolic Events, and Myocardial Infarction

eFigure 36. Funnel Plots for Overall and Bleeding associated Mortality

eFigure 37. TotalThromboembolic Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model of updated Meta-Analysis

eFigure 38. TotalThromboembolic Event Rate in Patients receiving TXA compared to Control using the Random-Effects Model updated Meta-Analysis

eFigure 39. Venous Thrombosis Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model of updated Meta-Analysis

eFigure 40. Venous Thrombosis Event Rate in Patients receiving TXA compared to Control using the Random-Effects Model of updated Meta-Analysis

eFigure 41. Pulmonary Embolism Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model of updated Meta-Analysis

eFigure 42. Pulmonary Embolism Event Rate in Patients receiving TXA compared to Control using the Random-Effects Model of updated Meta-Analysis

eFigure 43. Venous Thromboembolic and Pulmonary Embolism Event Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model of updated Meta-Analysis

eFigure 44. Venous Thromboembolic and Pulmonary Embolism Event Rate in Patients receiving TXA compared to Control using the Random-Effects Model of updated Meta-Analysis

eFigure 45. Overall Mortality Rate in Patients receiving TXA compared to Control using the Fixed-Effect Model of updated Meta-Analysis

eFigure 46: Overall Mortality Rate in Patients receiving TXA compared to Control using the Random-Effects Model of updated Meta-Analysis

eFigure 47. Total Thromboembolic Event Rate in Patients receiving TXA compared to Control analysed with Risk Ratio in Studies with Zero Events of updated Meta-Analysis

eFigure 48. TotalThromboembolic Event Rate in Patients receiving TXA compared to Control analysed with Risk Ratio in Studies without Zero Events of updated Meta-Analysis

eReferences