CRASH-2 - TXA in Trauma - EM Final Exams

Instructions

You are the ST4 sitting the Critical Appraisal station. Article: Shakur H et al. 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;376:23-32.

See the structured article summary card overleaf — this is the candidate's 2-week pre-read material. Paper: Shakur H et al.

You are alone in the CLA examination room with one examiner; no actor. You have 8 minutes to critically appraise this paper and discuss it with the examiner. The examiner will ask questions.

Equipment & setting

No equipment notes provided.

Actor / patient brief

No actor brief provided.

Finish your timed attempt before revealing the answer.

The tasks & expected progression

Identify design and EBM hierarchy position
Explain ITT vs PP analysis + why CRASH-2 used ITT
Interpret primary outcome + NNT + time-to-treatment effect
Apply to UK ED major haemorrhage protocol

0-1 min Article identification & design Names Shakur 2010 Lancet CRASH-2; identifies as pragmatic double-blind placebo-controlled multi-centre RCT, 274 hospitals 40 countries, NIHR HTA funded; top of single-trial EBM hierarchy; informed CRASH-3 (TBI) + WOMAN (PPH).
1-3 min PICO + methods critique P: 20211 adults <8h of trauma with/at-risk of significant haemorrhage. I: TXA 1g/10min + 1g/8h IV. C: identical placebo. O: primary 28d all-cause death. Allocation: sequential numbered identical packs (concealment intact). Double-blind. Pre-specified sample size powered for ~1.5% absolute reduction at 80% power.
3-5 min Results interpretation Primary 14.5 vs 16.0% RR 0.91 (95% CI 0.85-0.97, p=0.0035), ARR 1.5%, NNT 67. Bleeding deaths 4.9 vs 5.7% RR 0.85. Crucial time-stratification (BMJ 2011 follow-up paper): benefit 3h - drives current 3h ED window. No excess vascular occlusive events.
5-7 min Validity assessment + GRADE ITT principle preserved despite protocol violations (rate ~3-4%); sensitivity analyses excluding violators gave consistent results - reinforces ITT robustness. ITT preserves randomisation balance and reflects pragmatic 'real-world' effectiveness; PP only acceptable as supportive analysis or for non-inferiority. RoB 2.0 low across all 5 domains. GRADE HIGH. CASP RCT checklist favourable.
7-8 min Application to UK ED + answer examiner Q Yes - TXA 1g IV bolus + 1g infusion within 3h of injury, embedded in NICE NG39, RCEM major haemorrhage SOP, JRCALC pre-hospital; replicated/extended by CRASH-3 (TBI), WOMAN (PPH); signposts RCEM Learning, BestBETs, NHS K+L.

Mark scheme

Clinical reasoning

70%

Excellent Names CRASH-2 design, explains ITT robustly with reference to sensitivity analyses, quotes primary RR with CI and NNT, flags 3h time window from follow-up paper, applies to NICE NG39 + RCEM major haemorrhage SOP, links to CRASH-3 + WOMAN.
Above standard Identifies double-blind RCT and ITT principle; quotes primary effect with CI; mentions time window; applies CASP framework with minor gaps; clear UK ED application.
Standard Recognises pragmatic RCT and ITT in principle; quotes overall mortality finding; basic CASP application; mentions time window when prompted; outlines NICE link.
Borderline Vague on ITT vs PP distinction; quotes numbers without CIs; minimal time-window discussion; surface CASP application.
Below standard Confuses ITT with PP; equates p=0.0035 with clinical significance; ignores time-window; cannot apply to UK guidelines accurately.
Unacceptable Cannot identify design or analyse ITT; misquotes the primary outcome; recommends TXA >6h post-injury without nuance; no understanding of RoB or NNT.

Communication

30%

Excellent Structured CASP-style exposition; explains ITT in plain English; transparent about heterogeneity and time-window; professional academic register; concise answers to examiner probes.
Above standard Logical flow; clear lay translation of statistics; transparent on limitations; responds well to interruption; mostly fluent.
Standard Generally organised; explains key numbers when prompted; acknowledges main limitations; some hesitation but recoverable; professional tone.
Borderline Disorganised; struggles to explain ITT plainly; partial transparency on limitations; jargon-heavy.
Below standard Poorly structured; cannot translate stats; defensive on limitations; jargon-laden; loses examiner.
Unacceptable No structure; misrepresents findings; dismissive of limitations; cannot answer probes; would mislead colleagues.

— Self-mark each domain

Weighted across all domains · Day-1 consultant standard · pass mark 60%.

What good looks like

ITT (intention-to-treat) analyses every patient in the group they were randomised to, regardless of compliance - preserves randomisation balance and gives unbiased estimate of effectiveness.
Per-protocol (PP) analyses only those who completed treatment as planned - introduces selection bias but useful as supportive analysis or in non-inferiority trials.
Pack-based identical-appearance placebo with sequential numbering is the gold standard for allocation concealment in pragmatic multi-centre trials.
CRASH-2 informed CRASH-3 (TBI), WOMAN (PPH) and changed NICE NG39 + RCEM major haemorrhage protocols - landmark UK EM evidence.

Common pitfalls

Switching to PP analysis when ITT result is borderline - introduces post-hoc bias and undermines the randomisation.
Ignoring the time-to-treatment subgroup (benefit 3h) and giving TXA at 6 hours post-injury.
Equating the relatively small absolute reduction (1.5%) with no clinical importance - at population scale (20211 patients) this is highly clinically meaningful.
Calling 274-hospital 40-country generalisability a flaw - heterogeneity strengthens external validity for UK ED practice.