Publication

• Title: Effect of Out-of-Hospital Tranexamic Acid vs Placebo on 6-Month Functional Neurologic Outcomes in Patients With Moderate or Severe Traumatic Brain Injury
• Acronym: ROC TXA
• Year: 2020
• Journal published in: JAMA
• Citation: Rowell SE, Meier EN, McKnight B, et al. Effect of out-of-hospital tranexamic acid vs placebo on 6-month functional neurologic outcomes in patients with moderate or severe traumatic brain injury. JAMA. 2020;324(10):961-974.

Context & Rationale

• Background

  • Traumatic brain injury (TBI) is a leading cause of death and long-term disability, with few proven pharmacologic therapies that improve patient-centred outcomes.
  • Early post-traumatic intracranial haemorrhage expansion is common and is biologically plausible as a modifiable contributor to secondary brain injury.
  • Trauma-induced coagulopathy and (context-dependent) hyperfibrinolysis can occur after TBI; tranexamic acid (TXA) inhibits fibrinolysis and could, in theory, reduce haemorrhage progression.
  • Pre-trial evidence for TXA in TBI suggested potential time-dependent benefit but left major uncertainty regarding ultra-early (prehospital) administration, optimal dosing strategy, and safety (particularly seizures and thromboembolic complications).
• Research Question/Hypothesis

  • Does initiating TXA in the prehospital phase of care improve 6-month functional neurological outcome in patients with suspected moderate-to-severe TBI who are not in shock?
  • Does the effect (and safety profile) differ between a 2 g bolus-only strategy and a 1 g bolus plus 1 g maintenance infusion strategy?
• Why This Matters

  • Prehospital protocols for TXA were expanding internationally, and clinicians needed high-grade evidence specific to isolated/primary TBI (not simply haemorrhagic shock).
  • Functional outcomes (not just mortality) are central to TBI therapeutics; a prehospital RCT with 6-month follow-up directly addresses this clinical priority.
  • Dosing strategy matters in TXA (efficacy vs adverse events); testing two regimens in the same platform provides pragmatic and mechanistic insight.

Design & Methods

• Research Question: In prehospital patients with suspected moderate-to-severe TBI (prehospital GCS ≤12) without shock (SBP ≥90 mmHg), does prehospital TXA (two dosing strategies) compared with placebo improve 6-month functional neurological outcome?
• Study Type: Randomised, multicentre, double-blind, placebo-controlled, three-arm trial (Resuscitation Outcomes Consortium); prehospital enrolment with in-hospital continuation infusion; group sequential design with prespecified interim monitoring; conducted across 12 ROC sites (20 trauma centres; 39 EMS agencies) in the United States and Canada.
• Population:
  • Setting: Out-of-hospital EMS care with subsequent trauma-centre management.
  • Key inclusion: Age ≥15 years (or estimated weight >50 kg if age unknown); blunt or penetrating trauma with suspected TBI; prehospital GCS ≤12 (before sedatives/paralytics); prehospital SBP ≥90 mmHg; at least 1 reactive pupil (or both pupils if GCS 4–12); able to start the out-of-hospital infusion within 2 hours of injury.
  • Key exclusions: GCS 3 with neither pupil reactive; CPR prior to randomisation; estimated time from injury to infusion start >2 hours or unknown; suspected seizure activity; acute myocardial infarction or stroke; known history of seizures, thromboembolic disorder, or renal dialysis; pregnancy; incarcerated status; opt-out/declined participation (where applicable).
• Intervention:
  • Bolus-maintenance TXA strategy: 1 g TXA in a 250 mL out-of-hospital infusion, followed by 1 g TXA in a 1 g/8 h in-hospital infusion.
  • Bolus-only TXA strategy: 2 g TXA in a 250 mL out-of-hospital infusion, followed by placebo in-hospital infusion over 8 h.
  • Delivery: Out-of-hospital infusion initiated after essential life-saving interventions; in-hospital infusion initiated as soon as feasible after hospital arrival.
• Comparison:
  • Placebo out-of-hospital infusion (250 mL), followed by placebo in-hospital infusion over 8 h.
• Blinding: Double-blind (prehospital clinicians, in-hospital clinicians, patients, and outcome assessors); matched placebo bolus and infusion using identical study kits.
• Statistics: Planned 1200 randomised participants to yield ~963 treated participants for analysis; designed to detect an absolute 7% difference in favourable 6-month outcome (combined TXA strategies vs placebo) with 80% power using an asymmetric group sequential design (one-sided α=0.10 for benefit; one-sided α=0.025 for harm); primary analysis was a modified intention-to-treat approach including randomised participants who received the out-of-hospital infusion, with multiple imputation for missing 6-month outcomes.
• Follow-Up Period: 28 days (mortality and early clinical events); 6 months (GOSE and Disability Rating Scale).

Key Results

This trial was not stopped early. Enrolment and follow-up were completed to the prespecified analysis population (966 treated participants analysed).

• The primary endpoint was not statistically different: the site-adjusted estimate was compatible with a small benefit or no effect, and did not exclude clinically meaningful harm.
• Mortality and intracranial haemorrhage progression were numerically lower with TXA overall, but effect estimates were imprecise; the CT-positive exploratory mortality signal requires external confirmation.
• Regimen-level divergence mattered: seizure/seizure-like activity was more frequent with the 2 g bolus-only strategy, while thromboembolic outcomes differed between regimens without a consistent class-wide signal.

Internal Validity

• Randomisation and allocation: Computer-generated randomisation using permuted blocks (sizes 3 and 6); allocation concealed via sequentially used, identical study drug kits placed on EMS vehicles.
• Drop out/exclusions: 1063 participants were randomised; 97 randomised participants did not receive the out-of-hospital infusion and were excluded from analysis; 1 incarcerated participant was excluded; the primary analysis therefore used a modified intention-to-treat population (966 treated participants).
• Performance/detection bias: Double blinding was maintained with matched placebo bolus and infusion; emergency unblinding was uncommon (treatment arm unblinded in 7/312, 4/345, and 11/309 participants), with open-label TXA administered in 53% of those unblinded, limiting but not eliminating contamination risk.
• Protocol adherence: Out-of-hospital infusion completion was high: 291/312 (93%), 327/345 (95%), and 290/309 (94%); in-hospital infusion was initiated in 253/312 (81%), 298/345 (86%), and 243/309 (79%), with complete in-hospital infusion in 227/312 (73%), 264/345 (77%), and 214/309 (69%).
• Baseline characteristics: Groups were broadly comparable for demographics, injury severity, and CT findings; penetrating mechanism differed (4% bolus-maintenance, 1% bolus-only, 5% placebo), a potential prognostic imbalance in a trial with modest event rates.
• Timing: TXA was administered very early: median time from injury to start of out-of-hospital infusion was 43 (IQR 31–59) minutes (bolus-maintenance), 40 (28–59) minutes (bolus-only), and 41 (29–59) minutes (placebo); median time to in-hospital infusion initiation was 146 (119–187), 150 (120–189), and 150 (121–194) minutes, respectively.
• Dose: The trial tested two distinct strategies (1 g + 1 g maintenance vs 2 g bolus only); regimen-level divergence in effect estimates and adverse events complicates inference about a single “TXA effect” and raises the possibility that dose and/or delivery profile influences both efficacy and harms.
• Separation of the variable of interest: Clear pharmacologic separation was achieved (TXA vs placebo; and two TXA strategies), but incomplete initiation/completion of the in-hospital infusion (started in 79–86%; completed in 69–77%) likely reduced separation between strategies.
• Outcome assessment: GOSE and Disability Rating Scale are validated, patient-centred functional endpoints; 6-month outcomes were missing for a minority and addressed by multiple imputation.
• Statistical rigour: Prespecified asymmetric group sequential design (benefit tested at one-sided α=0.10; harm at one-sided α=0.025) and a prespecified primary comparison (combined TXA vs placebo) were methodologically coherent for safety monitoring, but increase interpretive complexity for clinicians accustomed to two-sided α=0.05.

Conclusion on Internal Validity: Overall, internal validity is moderate: randomisation, concealment, blinding, and early treatment delivery were strong, but post-randomisation exclusions and missing long-term outcomes (handled by imputation) introduce plausible bias and reduce confidence in small effect estimates.

External Validity

• Population representativeness: Best reflects haemodynamically stable (SBP ≥90 mmHg), suspected moderate-to-severe TBI patients treated very early in high-resource EMS systems; findings are less applicable to hypotensive/polytrauma patients (excluded), delayed presentations (>2 h; excluded), and children <15 years.
• Applicability: Translation is most direct to EMS systems with capacity for rapid IV infusion and coordinated trauma-centre continuation; generalisability to low-resource settings or systems with longer prehospital times is uncertain.
• Clinical phenotype: The pragmatic prehospital enrolment approach (GCS/pupils/time window) inevitably includes heterogeneous injury patterns and may enrol some patients without significant intracranial injury, which may attenuate apparent treatment effects in real-world implementation.

Conclusion on External Validity: External validity is moderate: results are most generalisable to early-treated, haemodynamically stable suspected moderate–severe TBI within organised trauma systems, and should not be extrapolated to shock/polytrauma or delayed presentations without caution.

Strengths & Limitations

• Strengths:
  • Large, multicentre, double-blind prehospital RCT with clinically meaningful 6-month functional outcomes.
  • Very early drug administration (median ~40 minutes from injury), directly testing a time-sensitive biological hypothesis.
  • Two distinct TXA strategies evaluated, addressing real-world uncertainty about dose and infusion necessity.
  • Prespecified analytic framework with explicit harm monitoring and long-term follow-up.
• Limitations:
  • Modified intention-to-treat analysis excluded 97 randomised but untreated participants (post-randomisation exclusions), risking bias and reducing power.
  • Incomplete initiation and completion of the in-hospital infusion may have reduced regimen separation and diluted efficacy signals.
  • Missing 6-month outcomes required multiple imputation; those lost to follow-up were less severely injured, raising the possibility of informative missingness.
  • Primary analysis combined two TXA regimens despite divergent regimen-level point estimates, complicating clinical translation of the “TXA effect”.
  • Population restricted to haemodynamically stable patients; does not address TXA effects in shock/polytrauma where TXA is commonly used for haemorrhage control.

Interpretation & Why It Matters

• Clinical practice

  Prehospital TXA for haemodynamically stable suspected moderate–severe TBI did not improve 6-month functional outcome versus placebo; routine use specifically for neuroprotection in this phenotype is not supported by the primary result.
• Dosing nuance

  Regimen-level divergence (2 g bolus-only vs 1 g + 1 g maintenance) suggests that “TXA in TBI” is not a single, dose-agnostic intervention; seizure risk signals and differing vascular occlusive event patterns should inform future protocol design.
• Trial design implications

  Field-based enrolment criteria create heterogeneity; future trials likely require biological enrichment (e.g., imaging-confirmed intracranial haemorrhage) and/or adaptive designs to detect modest, phenotype-specific benefits.

Controversies & Other Evidence

• The accompanying editorial emphasised that pragmatic prehospital enrolment can overinclude patients without the intended phenotype (≈20% had in-hospital GCS >12; ≈24% had no head injury by AIS), plausibly diluting any treatment effect on TBI-specific outcomes.¹
• The editorial highlighted interpretive complexity from regimen-level divergence (bolus-only vs bolus-maintenance) and the likelihood that any true benefit (if present) is modest and would require much larger trials to confirm using patient-centred outcomes.¹
• A post-publication correction amended a thromboelastography table cell (>3% clot lysis at 30 minutes after maximum amplitude) from 31 to 3 in the bolus-only group, underscoring data curation challenges in complex multicentre trials without changing the primary endpoint interpretation.²
• The large in-hospital CRASH-3 trial found no overall reduction in head-injury death with TXA (RR 0.94; 95% CI 0.86 to 1.02) but suggested time- and severity-dependent effects; ROC TXA tested a different (prehospital, very early, haemodynamically stable, moderate–severe) phenotype and did not demonstrate a functional benefit.³
• A systematic review and meta-analysis of RCTs in acute TBI found no mortality benefit (RR 0.95; 95% CI 0.88 to 1.02) and no effect on disability (MD -0.18; 95% CI -0.39 to 0.03), while suggesting reduced haematoma expansion (RR 0.77; 95% CI 0.63 to 0.93); this places the ROC TXA neutral functional result within a broader pattern of small, uncertain clinical effects.⁴
• An accompanying Intensive Care Medicine commentary argued that, when integrated with the totality of evidence, regimen-specific signals (including subgroup effects) should be treated as hypothesis-generating until replicated, and that meta-analytic synthesis remains essential for interpreting large but single-platform trials in heterogeneous syndromes like TBI.⁵
• Neurosurgical correspondence challenged causal inference from exploratory subgroup findings and highlighted the risk of overinterpreting chance imbalances in modestly sized strata, particularly when biological plausibility coexists with statistical fragility.⁶
• In an observational prehospital cohort (BRAIN-PROTECT), confounder-adjusted analyses found no association between TXA and 30-day mortality in the full cohort (OR 1.17; 95% CI 0.84 to 1.65), but higher adjusted odds in isolated severe TBI (OR 2.05; 95% CI 1.22 to 3.45), underscoring residual confounding risk and the need for phenotype-specific randomised evidence.⁷
• The European trauma coagulopathy guideline continues to recommend early TXA for bleeding trauma, but this evidence base is anchored in haemorrhage control rather than isolated TBI neuroprotection; extrapolation of “trauma TXA” guidance to haemodynamically stable isolated TBI remains uncertain.⁸
• Updated Brain Trauma Foundation prehospital TBI guidelines incorporated TXA as a dedicated evidence topic, reflecting ongoing uncertainty and signalling that TXA use in suspected moderate–severe TBI remains an evidence-informed but contested domain rather than settled standard of care.⁹

Summary

• In 1063 randomised prehospital patients with suspected moderate–severe TBI and SBP ≥90 mmHg, TXA (two strategies) did not significantly improve 6-month favourable functional outcome versus placebo (65% vs 62%; adjusted difference 3.5%; 95% CI -0.9 to 10.2).
• TXA did not significantly reduce 28-day mortality overall (14% vs 17%; adjusted difference -2.9%; 95% CI -7.9 to 2.1), and did not significantly reduce intracranial haemorrhage progression in those with paired CTs.
• An exploratory subgroup with intracranial haemorrhage on initial CT showed lower 28-day mortality with the 2 g bolus-only strategy (18% vs 27%; adjusted difference -8.2%; 95% CI -16.6 to -0.8), requiring confirmation.
• Safety signals differed by regimen: seizure/seizure-like activity occurred in 5% with bolus-only TXA vs 2% with placebo, and thromboembolic event patterns varied across strategies.
• Methodological strengths were substantial (blinding, early delivery, long-term outcomes), but post-randomisation exclusions, incomplete in-hospital infusion, and missing 6-month outcomes (addressed by imputation) temper confidence in small effect estimates.

Overall Takeaway

The ROC TXA trial is a landmark prehospital, double-blind pharmacologic RCT in suspected moderate–severe TBI with 6-month functional follow-up, rigorously testing the ultra-early antifibrinolysis hypothesis. Its neutral primary outcome argues against routine TXA use purely for neuroprotection in haemodynamically stable suspected TBI, while regimen-level divergence and an exploratory CT-positive mortality signal point to the need for targeted, biologically enriched confirmatory trials.

Bibliography

• Cone DC, Spaite DW, Coats TJ. Out-of-hospital tranexamic acid for traumatic brain injury. JAMA. 2020;324(10):946-947. Link
• Error in Table. JAMA. 2020;324(16):1683. Link
• CRASH-3 trial collaborators. Effects of tranexamic acid on death, vascular occlusive events, and other morbidities in patients with acute traumatic brain injury (CRASH-3): randomised, placebo-controlled trial. Lancet. 2019;394(10210):1713-1723. Link
• Al Lawati KA, Sharif S, Maqbali SA, et al. Efficacy and safety of tranexamic acid in acute traumatic brain injury: a systematic review and meta-analysis. Intensive Care Med. 2021;47(1):14-27. Link
• Maas AIR, Steyerberg EW, Citerio G. Tranexamic acid in traumatic brain injury: systematic review and meta-analysis trumps a large clinical trial? Intensive Care Med. 2021;47(1):74-76. Link
• Volovici V, Haitsma IK. Tranexamic acid and severe traumatic brain injury: the futile search for causality? Neurosurgery. 2021;88(5):E484-E485. Link
• Bossers SM, Loer SA, Bloemers FW, et al. Association between prehospital tranexamic acid administration and outcomes of severe traumatic brain injury. JAMA Neurol. 2021;78(3):338-345. Link
• Rossaint R, Afshari A, Bouillon B, et al. The European guideline on management of major bleeding and coagulopathy following trauma: sixth edition. Crit Care. 2023;27:80. Link
• Hawryluk GWJ, Lulla A, Bell R, Jagoda A, Mangat HS, Bobrow BJ, Ghajar J. Guidelines for prehospital management of traumatic brain injury 3rd edition: executive summary. Neurosurgery. 2023;93(6):e159-e169. Link