Mortality and Thrombosis in Injured Adults Receiving Tranexamic Acid in the Post-CRASH-2 Era

Introduction The CRASH-2 trial demonstrated that tranexamic acid (TXA) reduced mortality with no increase in adverse events in severely injured adults. TXA has since been widely used in injured adults worldwide. Our objective was to estimate mortality and adverse events in adults with trauma receiving TXA in studies published after the CRASH-2 trial. Methods We systematically searched PubMed, Embase, MicroMedex, and ClinicalTrials.gov for studies that included injured adults who received TXA and reported mortality and/or adverse events. Two reviewers independently assessed study eligibility, abstracted data, and assessed the risk of bias. We conducted meta-analyses using random effects models to estimate the incidence of mortality at 28 or 30 days and in-hospital thrombotic events. Results We included 19 studies and 13 studies in the systematic review and meta-analyses, respectively. The pooled incidence of mortality at 28 or 30 days (five studies, 1538 patients) was 10.1% (95% confidence interval [CI], 7.8–12.4%) (vs 14.5% [95% CI, 13.9–15.2%] in the CRASH-2 trial), and the pooled incidence of in-hospital thrombotic events (nine studies, 1656 patients) was 5.9% (95% CI, 3.3–8.5%) (vs 2.0% [95% CI, 1.8–2.3%] in the CRASH-2 trial). Conclusion Compared to the CRASH-2 trial, adult trauma patients receiving TXA identified in our systematic review had a lower incidence of mortality at 28 or 30 days, but a higher incidence of in-hospital thrombotic events. Our findings neither support nor refute the findings of the CRASH-2 trial but suggest that incidence rates in adults with trauma in settings outside of the CRASH-2 trial may be different than those observed in the CRASH-2 trial.


INTRODUCTION
Hemorrhage is the primary cause of death in the first 24 hours after trauma and is responsible for 40% of all traumarelated deaths. 1,2 Tranexamic acid (TXA), an antifibrinolytic drug that blocks plasmin-mediated fibrin clot breakdown, attenuates excessive bleeding. In patients undergoing surgery, TXA has been shown to decrease blood product transfusion requirements. 3,4 The success of TXA in the surgical setting led to the CRASH-2 trial, an international, randomized controlled trial of the early administration of TXA to bleeding adult trauma patients. 5 estimate the incidence of mortality and thrombotic events in injured adults in the post-CRASH-2 era. 9 Our objective was to evaluate the incidence of mortality and adverse events in studies published after the CRASH-2 trial results were published.

Search Strategy
We searched PubMed, Embase, MicroMedex, and ClinicalTrials.gov for studies that included adult trauma patients who received TXA and reported mortality and/or adverse events (Supplemental File). References of potentially eligible articles identified in the search were further screened for relevant references missed in the database search.

Inclusion and Exclusion
We included all studies that assessed mortality and/or adverse events in adult trauma patients receiving TXA. We included studies regardless of TXA dosing or clinical setting (e.g., prehospital, military, civilian) and studies that reported only mortality or adverse events. We excluded case reports and review articles, studies that were not trauma-related or that included primarily children, and studies that did not report mortality or adverse events. We also excluded studies that were secondary analyses of the CRASH-2 trial.

Study Selection
We screened studies for inclusion initially by titles, abstracts, and then full texts. Each study title and abstract was reviewed independently by two authors. When consensus could not be reached on screened titles and abstracts, a third reviewer independently adjudicated the discrepancies. Fulltext discrepancies were resolved by group consensus during in-person meetings. Prior to independent author screening, we piloted the study selection procedures as a group for several studies to enhance standardization of the selection protocol. Our study selection procedure is reported for in Figure 1 according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (See Figure 1).

Data Abstraction and Quality Assessment
Two authors independently abstracted data from each of the included studies. We abstracted study characteristics that included the following: year, country, setting (e.g., civilian, military, prehospital), design, inclusion criteria, TXA dosing, and outcomes measured. We abstracted outcome measures including mortality during any time frame, thrombotic events, and other adverse events reported, as well as adverse event definitions provided by the authors of the included study. Extraction was piloted as a group on several studies. We resolved disagreements in extracted data by group discussion and by consensus of all authors.
We assessed the quality of included studies using a quality assessment instrument previously developed by the National Heart, Lung, and Blood Institute. 11 This instrument included nine points assessing the clarity of the study objective and study population, the sequence of enrollment (consecutive vs. nonconsecutive), the comparability of subjects, the clarity of the study intervention and the outcome measures, the adequacy of length of follow-up, the appropriateness of statistical methods, and the clarity of reported results.

Outcome Measures
Our primary outcomes were mortality at 28 days and in-hospital thrombotic events, as these were outcomes reported in the CRASH-2 trial. 5 We identified studies that reported mortality at 30 days and thus expanded this outcome to include mortality at both 28 and 30 days. In-hospital thrombotic events for the CRASH-2 trial were defined as any vascular occlusive event including myocardial infarction, stroke, pulmonary embolism (PE) and deep vein thrombosis (DVT). We accepted any definition of an in-hospital thrombotic event as reported by the included studies. Since the majority of the studies reported total thrombotic events rather than the number of patients with thrombosis, we reported on total number of thrombotic events (i.e., if one patient had both a DVT and stroke identified, it would count as two thrombotic events). Secondary outcomes included mortality at 24 hours, in-hospital mortality, and in-hospital PE or DVT.

Analysis
Prior to pooling the data, we assessed studies for clinical heterogeneity based on study population, setting, design, intervention, and outcome assessment. All authors participated in group discussions to determine which studies should be excluded from the meta-analyses due to significant clinical heterogeneity compared to the other studies. We performed meta-analyses using the random effects model to report incidence with 95% confidence intervals (CI). Statistical heterogeneity was assessed with I 2 where a value >75% represents considerable heterogeneity. 12 Forest plots were ordered along the Y-axis by descending sample size. 13 We did not construct a funnel plot to assess for publication bias as these have been shown to be inaccurate for assessing incidence and may cross the 0 and 100% boundaries. 14 Statistical analysis was performed using Stata 14.0 (College Station, Texas).
We did not include six studies in the meta-analyses due to significant clinical heterogeneity compared to the other studies. Reasons for exclusion of these studies were as follows: only included patients with injury by firearm or explosive; 16 only included patients with hyperfibrinolysis; 24 only included combat patients who survived 24 hours after injury 27 or survived to receive treatment at a U.S. military hospital after transport from a combat hospital; 18,19 or only included patients with traumatic brain injury. 34
The pooled incidence of mortality at 28 or 30 days was 10.1% (95% CI, 7.8 to 12.4%; I 2 = 42.7%) ( Figure 2). This was lower than reported in the CRASH-2 trial, which had an incidence of mortality at 28 days of 14.5% (95% CI, 13.9 to 15.2%) in patients receiving TXA. The pooled incidence of in-hospital thrombotic events was 5.9% (95% CI, 3.3 to 8.5%; I 2 = 87.6%) (Figure 3). This was higher than reported in the CRASH-2 trial, which had an incidence of in-hospital thrombotic events of 2.0% (95% CI, 1.8 to 2.3%). The pooled incidences of the secondary outcomes of mortality at 24 hours, in-hospital mortality, and PE and/or DVT are reported in Figures 4 to 6.

DISCUSSION
Our study demonstrated some interesting findings, particularly in comparison to the CRASH-2 trial. In our study, trauma centers demonstrated a wide variation of TXA administration including dosing (bolus vs bolus + maintenance), total bolus dose (1 g vs 2 g), and timing (within three hours vs eight hours from injury). The CRASH-2 trial administered TXA as a 1 g IV bolus infusion over 10 minutes and a 1 g maintenance infusion over eight hours within eight hours from the time of injury. The varying timing of TXA adminstration noted in our study is likely a result of an exploratory analysis that demonstrated increased benefit in preventing hemorrhagic death with earlier TXA administration given within one hour. 6 The greatest benefit occurs when TXA is given within <1 hour from injury, diminished benefit if given within one to three hours from injury, and no benefit if given after three hours from injury. 6 In contrast to the CRASH-2 trial, three studies primarily administered only a bolus dose of TXA as opposed to a bolus dose with a subsequent maintenance dose. 16,20,21 This may contribute to different thrombosis and mortality rates. Current clinical trials are evaluating different TXA doses in injured patients. 15 Compared to the CRASH-2 trial, our pooled results demonstrated a lower incidence of mortality at 28 or 30 days and a higher incidence of in-hospital thrombotic events. We do not conclude from our findings that the effectiveness and harm of TXA is different than what was demonstrated in the CRASH-2 trial, as our study included primarily observational studies. Our results instead suggest    Overall, (95% CI) that the incidence rates observed in settings outside of the CRASH-2 trial might be different than what was observed in the CRASH-2 trial. This is particularly true for the incidence of thrombotic events, where our pooled results demonstrated a higher incidence of thrombotic events than what was observed in the CRASH-2 trial. The incidence rates reported in our included studies are likely biased towards under-reporting thrombotic events due to the high proportion of retrospective studies (more difficult to identify thrombotic events) and less-comprehensive definition of thrombotic event (often did not include arterial thromboses such as myocardial infarction or stroke) compared to the CRASH-2 trial. It is unclear why there are differences in thrombotic events seen in the CRASH-2 trial and our study. It is possible that the injury severity of the two study populations is different. We were unable to compare overall patient characteristics of our included studies with those of the CRASH-2 trial. It is also possible that sites included in the CRASH-2 trial screened less for thrombotic events compared to sites included in our study. Other large trauma clinical trials enrolling similarly injured populations have also reported higher thrombotic event rates compared to the CRASH-2 trial. [35][36][37] Several ongoing trauma clinical trials evaluating TXA should provide additional insight into the incidence of thrombotic events in this population. 15 Future studies evaluating TXA use in patients with hemorrhagic injuries may consider work to identify patients where the potential efficacy of TXA use is maximized and exposure to harm is minimized. Identification may be based on clinical characteristics, transport time or modality, or laboratory measurements such as thromboelastography. 38,39

LIMITATIONS
Our results should be interpreted in the context of some limitations. First, the included studies demonstrated clinical heterogeneity, limiting the numbers of studies that could be included in the meta-analyses. Second, studies had varying definitions of in-hospital thrombosis, which may lead to differences in reported incidence rates. Third, the majority of the studies were retrospective, and this may result in less accurate data abstraction compared to prospective studies. 40 This limitation is more relevant for the thrombosis outcome measure, which may be difficult to ascertain from retrospective chart review, than for the mortality outcome measure, which is easy to ascertain regardless of study design. Fourth, the chart abstractors were not blinded to the study hypotheses. This may have led to biases during study selection and data abstraction. Finally, the incidence of thrombotic events is ideally measured with the number of patients with any thrombotic event as the numerator and the total number of patients as the denominator. However, since included studies primarily reported total number of thrombotic events, we used the total number of thrombotic events as the numerator and the total number of patients as the denominator for calculating incidence.
Mortality and Thrombosis in Injured Adults Receiving TXA in the Post-CRASH-2 Era Benipal et al.

CONCLUSION
Compared to the CRASH-2 trial, adult trauma patients receiving TXA identified in our systematic review had a lower incidence of mortality at 28 or 30 days, but a higher incidence of in-hospital thrombotic events. Our findings neither support nor refute the findings of the CRASH-2 trial. They merely suggest that incidence rates observed in settings outside of the CRASH-2 trial may be different than those observed in the CRASH-2 trial.