Skip to main content

Foundational Trials

CRASH-2

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Effects of tranexamic acid on death, vascular occlusive events, and blood transfusion in trauma patients with significant haemorrhage (CRASH-2): a randomised, placebo-controlled trial
  • Acronym: CRASH-2
  • Year: 2010
  • Journal published in: The Lancet
  • Citation: CRASH-2 trial collaborators. Effects of tranexamic acid on death, vascular occlusive events, and blood transfusion in trauma patients with significant haemorrhage (CRASH-2): a randomised, placebo-controlled trial. Lancet. 2010;376(9734):23-32.

Context & Rationale

  • Background
    Among trauma patients who reach hospital alive, haemorrhage is a leading preventable cause of early death, and effective, simple therapies that reduce fatal bleeding could have very large population impact.
  • Research Question/Hypothesis
    Does early administration of a short course of tranexamic acid (TXA) reduce death (overall and from bleeding) without increasing vascular occlusive events in adult trauma patients with significant haemorrhage or at risk of significant haemorrhage?
  • Why This Matters
    TXA is inexpensive, widely available, and logistically simple; proving a mortality benefit in trauma would justify rapid integration into damage-control resuscitation pathways and prehospital systems.

Design & Methods

  • Research Question: In adult trauma patients with significant haemorrhage (or judged at risk), does TXA (1 g bolus + 1 g infusion) reduce death compared with placebo, without increasing vascular occlusive events?
  • Study Type: Randomised, placebo-controlled, double-blind, pragmatic, international multicentre trial (274 hospitals in 40 countries); enrolment within 8 hours of injury; randomisation predominantly via local numbered treatment-pack system (with a small proportion by telephone).
  • Population:
    • Adults with trauma, within 8 hours of injury, with significant haemorrhage (e.g., hypotension or tachycardia) or considered at risk of significant haemorrhage.
    • Key physiological thresholds used for eligibility assessment included systolic blood pressure <90 mm Hg and/or heart rate >110 beats/min (as markers of haemorrhagic shock), alongside clinician judgement of bleeding risk.
    • Protocol deviations recorded in the report included enrolment >8 hours after injury (23 patients), age <16 years (5 patients), and non-traumatic haemorrhage (9 patients); four patients were removed after randomisation due to consent withdrawal.
    • Trial registration: ISRCTN86750102; ClinicalTrials.gov NCT00375258.
  • Intervention:
    • Tranexamic acid: 1 g IV loading dose infused over 10 minutes, followed by 1 g IV maintenance infusion over 8 hours.
    • Administered as early as possible after randomisation (randomisation targeted within 8 hours of injury; mean time since injury at randomisation ~2.8 hours).
  • Comparison:
    • Matched placebo (identical administration schedule: 10-minute infusion + 8-hour infusion), plus usual trauma care (including transfusion, surgery, and haemostatic adjuncts per local practice).
  • Blinding: Double-blind (patients, treating clinicians, and trial personnel masked); allocation concealed via identical numbered treatment packs; emergency unblinding was not required.
  • Statistics: Planned sample size 20,000 to detect a relative risk of 0.90 for all-cause mortality (assumed baseline mortality 20%), corresponding to an estimated 85% chance of P<0.01 and 95% chance of P<0.05; primary analyses by intention-to-treat (all randomised patients except those withdrawn post-randomisation due to consent withdrawal).
  • Follow-Up Period: In-hospital follow-up to death, discharge, or 28 days (mortality assessed up to 4 weeks after injury).

Key Results

This trial was not stopped early. Recruitment continued to the planned scale (20,211 randomised).

Outcome Tranexamic acid Placebo Effect p value / 95% CI Notes
All-cause mortality (primary) 1463/10,060 (14.5%) 1613/10,067 (16.0%) RR 0.91 95% CI 0.85 to 0.97; P=0.0035 Primary endpoint; all-cause death within 4 weeks (in-hospital follow-up up to 28 days).
Death due to bleeding 489/10,060 (4.9%) 574/10,067 (5.7%) RR 0.85 95% CI 0.76 to 0.96; P=0.0077 Cause-specific death classification as recorded by sites.
Death due to vascular occlusion (MI, stroke, PE) 33/10,060 (0.3%) 48/10,067 (0.5%) RR 0.69 95% CI 0.44 to 1.07; P=0.096 Fatal occlusive events were rare.
Any vascular occlusive event (MI, stroke, PE, DVT; fatal or non-fatal) 168/10,060 (1.7%) 201/10,067 (2.0%) RR 0.84 95% CI 0.68 to 1.02; P=0.084 Diagnosis and ascertainment were pragmatic and site-dependent.
Myocardial infarction (fatal or non-fatal) 35/10,060 (0.3%) 55/10,067 (0.5%) RR 0.64 95% CI 0.42 to 0.97; P=0.035 Rare events; multiple comparisons relevant for interpretation.
Blood product transfusion (any) 5067/10,060 (50.4%) 5160/10,067 (51.3%) RR 0.98 95% CI 0.96 to 1.01; P=0.21 Among transfused: mean units 6.06 (SD 9.98) vs 6.29 (SD 10.31).
Surgical intervention (neurosurgery, chest, abdominal, or pelvic surgery) 4814/10,060 (47.9%) 4836/10,067 (48.0%) RR 1.00 95% CI 0.97 to 1.03; P=0.79 Pragmatic endpoint; includes varied surgical indications and practices.
Dead or dependent at discharge or day 28 3453/10,060 (34.3%) 3562/10,067 (35.4%) RR 0.97 95% CI 0.93 to 1.00; P=0.12 Functional status assessed at discharge or day 28 (whichever came first).
No symptoms at discharge or day 28 1483/10,060 (14.7%) 1334/10,067 (13.3%) RR 1.11 95% CI 1.04 to 1.19; P=0.0023 Interpret with caution: multiple functional categories reported; follow-up limited to discharge/day 28.
  • TXA reduced all-cause mortality (14.5% vs 16.0%) and death due to bleeding (4.9% vs 5.7%), with no statistically significant increase in vascular occlusive events (1.7% vs 2.0%).
  • Despite lower mortality, measured transfusion exposure was similar (50.4% vs 51.3%; mean units among transfused 6.06 vs 6.29), highlighting the pragmatic nature of transfusion measurement and the likelihood that mechanism and effect are not fully captured by transfusion counts.
  • Prespecified subgroup analyses for the primary outcome showed no strong evidence against homogeneity (trial prespecified threshold P<0.001): heterogeneity P values were systolic blood pressure 0.51, Glasgow Coma Score 0.50, type of injury 0.37, and time from injury to randomisation 0.11.

Internal Validity

  • Randomisation and allocation concealment: Centralised generation with allocation via sequentially numbered treatment packs at sites (20,116 allocations) and a small telephone randomisation component (95 allocations); clinicians had no foreknowledge of assignment.
  • Blinding: Double-blind trial with matched placebo infusions; no emergency unblinding required.
  • Attrition/exclusions: Four patients were removed after randomisation because consent was withdrawn; otherwise, analyses were intention-to-treat.
  • Baseline balance: Key prognostic factors were closely matched (e.g., mean age 34.6 vs 34.5 years; male 83.6% vs 84.0%; blunt injury 67.5% vs 67.7%; mean time since injury 2.8 vs 2.9 hours).
  • Timing: Mean time since injury at randomisation 2.8 (SD 2.2) vs 2.9 (SD 2.6) hours; treated ≤1 hour 37.2% vs 36.8%; >3 hours 32.6% vs 33.4%.
  • Dose and protocol delivery: Fixed regimen (1 g bolus + 1 g infusion over 8 hours) provides clear dosing separation; detailed metrics on completion of infusion and protocol adherence beyond allocation were not reported in the main paper.
  • Separation of the variable of interest: TXA regimen (1 g over 10 minutes + 1 g over 8 hours) versus matched placebo regimen; co-interventions were similar in aggregate (e.g., transfusion 50.4% vs 51.3%; surgery 47.9% vs 48.0%).
  • Outcome assessment: Primary endpoint (all-cause mortality) is objective; cause-of-death and vascular occlusive outcomes are more susceptible to misclassification and ascertainment variability across settings.
  • Statistical rigour: Prespecified sample size and conservative subgroup inference threshold (P<0.001 for heterogeneity) reduce spurious subgroup claims; effect estimates reported with 95% CIs and two-sided P values.
  • Harms capture: Vascular occlusive events were clinician-diagnosed and rare; the absence of increased events supports safety but does not eliminate under-ascertainment in a global pragmatic trial.

Conclusion on Internal Validity: Strong overall for the primary endpoint (concealed randomisation, robust blinding, near-complete follow-up, and objective mortality outcome), with the main residual uncertainty concentrated in secondary safety and cause-specific endpoints that depend on heterogeneous diagnostic practices.

External Validity

  • Population representativeness: Very broad international trauma population (274 hospitals, 40 countries) with pragmatic clinical inclusion criteria rather than laboratory-confirmed coagulopathy, supporting real-world applicability.
  • Clinical phenotype: Enrolled patients were predominantly young (mean ~35 years), largely blunt trauma (~68%), and commonly in physiological shock strata (e.g., systolic blood pressure ≤75 mm Hg ~15.5–15.9%).
  • System differences: Heterogeneity in trauma system maturity, transfusion practice, access to imaging for thromboembolism, and operative capability may influence baseline risks and measurement of some secondary outcomes.
  • Timing dependence: Because benefit is linked to early administration, generalisability is strongest where TXA can be delivered rapidly (prehospital or early ED pathways) and weakest where delays are common.
  • Applicability to specific subgroups: Evidence supports use in bleeding/risk-of-bleeding trauma broadly; extrapolation to patients without haemorrhage, to late presenters, and to paediatric trauma requires separate evidence.

Conclusion on External Validity: High generalisability for early TXA use in adult trauma with suspected significant haemorrhage, with caution needed when translating safety signals and late-treatment effects across systems with very different diagnostic and resuscitation infrastructure.

Strengths & Limitations

  • Strengths: Extremely large sample; international multicentre pragmatism; concealed randomisation and double blinding; objective primary endpoint; clinically meaningful mortality reduction; low-cost, scalable intervention with clear dosing regimen.
  • Limitations: Secondary safety outcomes dependent on variable ascertainment; cause-of-death classification vulnerable to misclassification; incomplete clarity on infusion completion/adherence; transfusion and procedural outcomes are crude proxies for bleeding control in heterogeneous systems; enrolment window up to 8 hours raises concerns about late administration (addressed further in subsequent analyses).

Interpretation & Why It Matters

  • Clinical practice
    CRASH-2 established TXA as a mortality-reducing therapy in adult trauma patients with suspected significant haemorrhage, forming the evidentiary backbone for early TXA use in damage-control resuscitation pathways.
  • Mechanistic inference
    Mortality benefit occurred without a clear reduction in transfusion or surgery, suggesting that the clinically relevant effect may relate to stabilisation of early clot integrity and prevention of fatal haemorrhage rather than a large, measurable reduction in total blood product exposure.
  • Safety signal
    No statistically significant excess of vascular occlusive events was observed, supporting a favourable benefit–risk profile for early TXA in the studied population.

Controversies & Subsequent Evidence

  • Transfusion “paradox”: The mortality benefit with little to no difference in transfusion or surgery was highlighted as mechanistically and methodologically challenging (pragmatic transfusion measurement, variability in practice, and potential misdiagnosis of haemorrhage at enrolment). 1
  • Time-to-treatment dependence: A subsequent CRASH-2 analysis reported that benefit for death due to bleeding was greatest when TXA was given ≤1 hour (RR 0.68; 95% CI 0.57 to 0.82) and 1–3 hours (RR 0.79; 95% CI 0.64 to 0.97), with apparent harm when given >3 hours (RR 1.44; 95% CI 1.12 to 1.84), underpinning modern “as early as possible, within 3 hours” practice. 2
  • Thromboembolism ascertainment: The low event rates and reliance on clinical diagnosis raised concerns about under-detection of non-fatal venous thromboembolism in a global pragmatic trial; subsequent syntheses generally support no major excess of vascular occlusion with TXA, but diagnostic heterogeneity remains a limitation in interpreting safety across settings. 13
  • Patient selection and fibrinolysis phenotypes: Trauma-associated fibrinolysis is heterogeneous; observational work using viscoelastic phenotyping reported potential harm when TXA is given to patients with “physiologic” fibrinolysis, strengthening arguments for refined selection in some systems (while acknowledging confounding). 4
  • Prehospital RCTs in advanced trauma systems: STAAMP (prehospital TXA) and PATCH (prehospital bolus + in-hospital infusion) tested earlier delivery and different primary endpoints; both inform ongoing debate about best candidates, dosing strategy, and what “clinically meaningful benefit” should be (mortality vs longer-term functional outcomes). 56
  • Guideline incorporation: Major contemporary trauma guidance generally supports early TXA (preferably <1 hour and no later than 3 hours) as part of haemorrhage management, while acknowledging ongoing uncertainty about optimal prehospital selection and dosing in advanced systems. 789

Summary

  • CRASH-2 was a very large, international, double-blind RCT testing early TXA versus placebo in adult trauma patients with significant haemorrhage or at risk of haemorrhage.
  • TXA reduced all-cause mortality (14.5% vs 16.0%; RR 0.91; 95% CI 0.85 to 0.97; P=0.0035) and death due to bleeding (4.9% vs 5.7%; RR 0.85; 95% CI 0.76 to 0.96; P=0.0077).
  • Vascular occlusive events were not significantly increased (1.7% vs 2.0%; RR 0.84; 95% CI 0.68 to 1.02; P=0.084), supporting a favourable safety profile in this pragmatic framework.
  • Transfusion and surgery rates were similar between groups, a key interpretive and mechanistic debate that influenced subsequent research and commentary.
  • Subsequent analyses and trials refined modern practice around very early administration (within 3 hours) and ongoing questions about patient selection in advanced trauma systems.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • CRASH-2 supports TXA for adult trauma with suspected significant haemorrhage; modern practice is strongly shaped by subsequent evidence emphasising administration as early as possible and within 3 hours of injury.

Overall Takeaway

CRASH-2 demonstrated that a simple, inexpensive TXA regimen given early to adult trauma patients with suspected significant haemorrhage reduces mortality, particularly death due to bleeding, without a clear signal of excess vascular occlusive complications in this pragmatic framework. The trial’s scale and feasibility made TXA a cornerstone of modern haemorrhage management, while subsequent work refined the critical importance of very early administration and continues to shape debates about patient selection and endpoints in advanced trauma systems.

Overall Summary

  • Large pragmatic international RCT: TXA reduced all-cause mortality and death due to bleeding in adult trauma with suspected haemorrhage.
  • No statistically significant increase in vascular occlusive events was observed; safety interpretation is limited by pragmatic event ascertainment.
  • Subsequent evidence established time-critical use: as early as possible and within 3 hours.

Bibliography

  • 1Levy JH. Antifibrinolytic therapy: new data and new concepts. Lancet. 2010;376(9734):3-4. Link
  • 2Roberts I, Shakur H, Afolabi A, et al; CRASH-2 Trial Collaborators. The importance of early treatment with tranexamic acid in bleeding trauma patients: an exploratory analysis of the CRASH-2 randomised controlled trial. Lancet. 2011;377(9771):1096-1101. Link
  • 3Fouché PF, et al. Tranexamic Acid for Traumatic Injury in the Emergency Department: A Systematic Review and Bias-Adjusted Meta-analysis. Ann Emerg Med. 2024;83(3):234-246. Link
  • 4Moore HB, Moore EE, Gonzalez E, et al. Tranexamic acid is associated with increased mortality in patients with physiological fibrinolysis. J Surg Res. 2017;220:438-443. Link
  • 5Guyette FX, Brown JB, Zenati MS, et al. Tranexamic Acid During Prehospital Transport in Patients at Risk for Hemorrhage After Injury: A Double-blind, Placebo-Controlled, Randomized Clinical Trial. JAMA Surg. 2021;156(1):11-20. Link
  • 6Gruen RL, Jacobs IG, Reade MC, et al. Prehospital Tranexamic Acid for Severe Trauma. N Engl J Med. 2023;389(2):127-136. Link
  • 7Spahn DR, Bouillon B, Cerny V, et al. The European guideline on management of major bleeding and coagulopathy following trauma: sixth edition. Crit Care. 2023;27:163. Link
  • 8LaGrone LN, et al. American Association for the Surgery of Trauma/American College of Surgeons Committee on Trauma clinical protocol for damage control resuscitation. J Trauma Acute Care Surg. 2024;96(3):e91-e99. Link
  • 9Dumas RP, Succar BE, Vella MA, et al. The use of tranexamic acid in the management of injured patients at risk of hemorrhage: a systematic review and meta-analysis and an Eastern Association for the Surgery of Trauma Practice Management Guideline. J Trauma Acute Care Surg. 2025 Sep 15. Link