Our review followed the guidelines for Meta-analyses of Observational Studies in Epidemiology and the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA), and the protocol was registered in PROSPERO (registration No CRD42021240303) (Stroup et al., 2000; Moher et al., 2009). Two independent researchers (PH and JJL) searched Medline, Embase, Web of Science, and Cochrane databases updated to 15 September for meta-analyses. The detailed search strategy is shown in Supplementary Appendix S1. We processed another search at the end of the study on 16 February 2022. We would only replace the included study when the newly published study had a different conclusion. Two researchers (RL and YD) independently screened the titles and abstracts, and articles satisfying the inclusion criteria were accessed for full-text review. They independently reviewed full-text articles for eligibility. When data were incomplete, the corresponding author was contacted by email and invited to provide additional information. Reference lists of eligible reviews and meta-analyses were searched for additional citations.

We only included meta-analyses of RCTs or meta-analyses of observational studies discussing TXA. Post hoc analyses and systematic reviews without meta-analysis were excluded. Articles were included when the exposure group was divided based on the dosage or administration method, and patients who just received the placebo or saline were included in the controlled group. We included studies with comparisons of different administration methods and dosages (high versus low dose and any versus none). Bayesian meta- and network meta-analysis were excluded unless we could extract the complete statistical data separately from a subgroup analysis. Included studies must be in accordance with the Declaration of Helsinki and approved by the respective ethics committees.

Two researchers (SR and JJL) independently extracted data from eligible articles. Our data extraction methodology was recommended and revised by the Joanna Briggs Institute (Supplementary Appendix S2) (Aromataris et al., 2014). Outcomes were classified as blood loss (total blood loss, intraoperative blood loss, and postoperative blood loss), risk of adverse effects (risk of massive hemorrhage, risk of transfusion requirement, risk of postoperative edema and ecchymosis, risk of rebleeding, and risk of significant adverse effects), and other outcomes (surgical field score, satisfaction with the surgical field, operative time, intraoperative blood pressure, hemoglobin drop, hematocrit drop, and length of hospital stay). We only compared the preoperative dose, which was more important than the maintenance dose. Considering that more variables would bring heterogeneity and that the maintenance dose was also correlated with the preoperative dose, we did not compare the maintenance dose. If an article presented separate meta-analyses for more than one health outcome, we included each of these separately. Standard mean difference (SMD) or weighted mean difference (WMD) was used for continuous variable statistics, and relative risks (RR), odds ratios (OR), or risk difference (RD) was used for discontinuous variable statistics.

We used “assessing the methodological quality of systematic reviews (AMSTAR) 2” to assess the methodological quality of the included meta-analyses (Supplementary Appendix S3) (Shea et al., 2017). We also used the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) tool to evaluate the quality of evidence for each included outcome ( Supplementary Appendix S4 ) (Guyatt et al., 2011). The GRADE tool classifies evidence for outcomes from systematic reviews and meta-analyses into “high,” “moderate,” “low,” and “very low.” The study design of the included studies decides the baseline quality of evidence, but other factors can decrease or increase the quality level (for the detailed appraisal rule, see Supplementary Appendix S4). Two researchers (RL and JJL) rated the methodological quality of the reviews.

The heterogeneity in the results of meta-analyses was assessed by chi-square and I² tests, followed by appropriate analysis models (fixed-effect or random-effect). Chi-squared p < 0.05 and I² >50% indicated high heterogeneity, and random-effects models were used in these outcomes, whereas chi-squared p > 0.05 and I²<50% were considered acceptable heterogeneity, and a fixed-effects model was used instead. Most of the meta-analyses included in our review used software (Stata and RevMan) to process their statistical analyses. Publication bias was evaluated by funnel plots and further confirmed by Egger’s and Begg’s test, and p < 0.05 was considered as the statistically significant risk of bias (Sterne et al., 2000). Small-study effects that lead to potential reporting or publication bias could also be avoided by Egger’s test. However, if the data of Egger’s test were not reported, then the meta-analyses, including no more than five studies, were considered high publication bias. Sensitivity analysis was performed in the meta-analysis by excluding each study one at a time to check whether the effectiveness of the outcome was determined by individual studies. Overlapping reviews happened frequently when two or more studies evaluated the same exposure and outcome. To determine the potential overlapping, a graphical cross-tabulation was made to calculate the corrected covered area (CCA), which could quantify the degree of overlap (Pieper et al., 2014). Detailed rules and standards of calculation are explained in Supplementary Appendix S5.

Besides, Shojania et al. suggested that more than 50% of systematic reviews were outdated after 5.5 years (Shojania et al., 2007). Hence, if the gap between the year of publication of potential overlapping studies is >5 years or if meta-analysis only included studies published more than 15 years ago, the previous one would be excluded directly. Update of eligible reviews was recommended by Garner et al. (2016) and satisfied the following criterion: the review achieved a minimum rating of the Moderate AMSTAR score. We identified newly published studies that met the inclusion criteria and matched the keywords of previous studies; the outcomes from the newly published one would change the conclusion or credibility of the review.

Figure 1 (flowchart of literature search) shows the procedure for the literature search and selection of eligible studies. Overall, 736 unique records (excluded duplicates) were searched across 4 databases, and 292 full-text articles were screened in full text. After the preliminary screening, 154 meta-analyses met the inclusion criteria. Supplementary Appendix S5 describes citation matrices of CCA for reviews with overlapping associations. In all, 107 meta-analyses reported 25 overlapping topics. After removing overlapping and outdated reviews, 47 studies were included in our umbrella review with 319 unique outcomes. Supplementary Appendices S6,S7 show the characteristics of included studies and excluded studies with their reasons for exclusion. Supplementary Appendices S8,S9 show specific data of the detailed outcomes in Result and Discussion.

The included meta-analyses were divided into eight categories as per the types of surgery: joint replacement surgery, other orthopedic surgery, cerebral surgery, cardiac surgery, nasal surgery, obstetrics and gynecological surgery, other types of surgery, and complex antifibrinolytic agents. The 47 included studies were published from 2012 to 2022, and 42 of them were published within 6 years (2016 and later). A total of 45 of 47 studies were rated as High quality in AMSTAR, and the remaining two were rated as Moderate quality. All the studies of High quality were meta-analyses of RCTs. AMSTAR scores for individual studies included in the umbrella review are shown in Supplementary Appendix S10. Among 319 outcomes, there were 101 outcomes assessed as at least of Moderate quality by GRADE, and 58 of them were rated as High quality. GRADE classification of the quality of evidence is listed in Supplementary Appendix S11.

Blood loss, being the core outcome of our study, had complicated data for subgroup analysis. We divided outcomes into those of the WMD group and the SMD group. The WMD group included subgroups of total blood loss, intraoperative blood loss, and postoperative blood loss (see Table 1).

Nine types of surgeries provided WMD outcomes with a relatively higher quality (at least Moderate GRADE), and two of them were High GRADE. IV TXA in spine surgery (WMD -214.68ml, 95% CI, -281.10 to -148.27) and orthognathic surgery (WMD -153.97ml, 95% CI, -166.52 to -141.41) could reduce perioperative bleeding and intraoperative bleeding, respectively (Zhao et al., 2019; Chen et al., 2021). Besides, Huang et al. (2015) reported that antifibrinolytic drugs could significantly reduce blood loss in THA (WMD -389.14, 95% CI, -483.95 to -295.23). In conclusion, TXA could effectively reduce blood loss and play a more significant role in major orthopedic surgery, including THA, TKA, and spine surgery.

As for subgroups, antifibrinolytic drugs could lead to a reduction of both intraoperative and postoperative blood loss (see Supplementary Appendix S12). A few studies that discussed SMD blood loss are shown in Supplementary Appendix S12d. IV TXA could significantly reduce intraoperative blood loss in endoscopic sinus surgery (SMD -0.66, 95% CI, –0.86 to -0.46) (Kim et al., 2019). Du and Feng concluded that IV or topical TXA could reduce intraoperative blood loss in spinal fusion surgery (SMD -0.32, 95% CI, -0.58 to -0.06) (Du and Feng, 2018).

Blood transfusion is an emergency treatment and could reflect the critical situation of the patient. In all, 14 surgeries provided outcomes with a relatively higher quality, and 16 outcomes were rated as High (see Tables 2, 3). According to 42 RCTs, Huang et al. (2014) concluded that TXA would reduce the transfusion rate in major orthopedic surgery by 49% (RR 0.51, 95% CI, 0.46–0.56). Besides, data from 31 RCTs demonstrated that IV or topical TXA in cardiac surgery could reduce the transfusion rate by 29% (RR 0.71, 95% CI, 0.65–0.78) (Guo et al., 2019). We also summarized incidences of adverse events containing seizure (a neurologic event), death, and vascular adverse events (see Table 4). Vascular adverse events included deep vein thrombosis (DVT), stroke, myocardial infarction (MI), and pulmonary embolism (PE). Five outcomes were rated as High. Zhang S. et al. (2019) reported that IV TXA could effectively reduce the rate of all adverse events in calcaneal fracture surgery (RR 0.26, 95% CI, 0.15–0.42). Li et al. indicated that IV antifibrinolytic drugs decreased the risk of DVT in spine surgery by 64%. Besides, TXA decreased the mortality of upper gastrointestinal bleeding by 41% (Twum-Barimah et al., 2020). In contrast, Guo et al. (2019) displayed that high-dose TXA increased the risk of seizure by nearly five times in cardiac surgery, and Zhang S. et al. (2019) also demonstrated an increased incidence of seizure by more than six times. Therefore, we concluded that TXA could reduce the transfusion rate and that TXA is also a protective factor from vascular adverse events and death, but it is a risk factor for seizures.

We also evaluated outcomes of reoperation rate, operative time, length of hospital stay, and hemoglobin (Hb). Re-operation was the additional surgical intervention due to poor postoperative recovery. The reduction in the postoperative reoperation rate could avoid unnecessary pain and the economic burden on patients. As shown in Supplementary Appendix S13a, Zhang Y. et al. (2019) reported that IV TXA led to a 54% decrease in the reoperation rate in coronary artery bypass grafting (CABG) (RR 0.46, 95% CI, 0.31, 0.68). Similarly, Guo et al. (2019) showed that IV or topical TXA averts 38% of reoperation in cardiac surgery.

Besides, the operative time and length of hospital stay were also significantly shortened with TXA treatment (see Supplementary Appendix S13b,c). Zhao et al. demonstrated the decreasing operation time of orthognathic surgery with IV TXA (WMD -16.18, 95% CI, -19.60 to -12.75) (Zhao et al., 2019). Tsai et al. also reported that IV TXA could shorten the length of hospital stay for hemoptysis patients by 1.62 days (Tsai et al., 2020). These outcomes demonstrated the hemostatic effect of TXA, with faster postoperative recovery.

Furthermore, postoperative Hb outcomes have been displayed in Supplementary Appendix S14a,b. Chen et al. determined that the IV TXA group gained a higher Hb level after spine surgery (SMD 0.20, 95% CI, 0.02–0.38) (Chen et al., 2021). Yao et al. (2019) and Guo et al. (2018) also suggested that patients receiving TXA showed fewer Hb decreases. However, in terms of the coagulation function, there is no significant difference in partial thromboplastin time and prothrombin time (see Supplementary Appendix S14c).

We summarized the recommended application of TXA in different types of adult surgery after the overall data analysis. IV-combined topical TXA (1–1.5 g IV + 1–3 g topical) was recommended for joint replacement surgery, and 10–15 mg/kg IV TXA was recommended in spinal surgery. 10 mg/kg IV TXA was recommended for cardiac surgery, and 2 g IV TXA was recommended in cerebral hemorrhage. Other detailed data are presented in Table 5.

We analyzed data from 47 meta-analyses and demonstrated that TXA effectively reduced blood loss and the transfusion rate in various surgeries. TXA resulted in a higher Hb level, shorter operative time and length of hospital stay, and a decreased rate of reoperation. Besides, TXA did not increase the risk of death and vascular adverse events, but it was a risk factor for seizure in cardiac surgery. Therefore, we believe that TXA is a safe choice for hemostasis in major surgeries, in particular with a high risk of allogeneic blood transfusion. We also recommended the dosage of TXA in different types of adult surgery.