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

20 May
like you really care how it works...

This should surely be kind of a big deal. Given that it seems to be evidence of something cheap, easy to give that seems to make (albeit) a difference in mortality outcomes without any signs of problems.

I had vague memories of tranexamic acid from O&G stuff but wasn’t even on my radar for trauma.

Let me try and summarise the paper


  • the usual good stuff
    • double blind, randomised all that jazz
  • 20000 patients – yes that’s right 20000 patients…
  • included patients with or “at risk of” (that’s the tricksy phrase) major bleeding from trauma
  • outcome was death and subdivided according to cause
  • Pfizer supplied the drug and placebo but nothing else that I can see


  • death 14.5% vs 16% favouring tranexamic acid over placebo

The always inconvenient and tricky details

the inclusion criteria are the real difficulty in this paper. You could get in if you had major bleeding (which seems to be defined as low BP and tachy) OR if someone felt you were at risk for bleeding. Let me quote a bit

Patients were included if the responsible doctor was substantially uncertain about whether or not to treat with tranexamic acid (ie, entry was governed by the uncertainty principle). Patients for whom the responsible doctor considered that there was a clear indication for tranexamic acid were not randomly assigned

There are 2 problems with that statement above
  1. how do i know if i’m substantially uncertain?
  2. what are the clear indications for tranexamic acid – i presume that’s what this trial was to find out
Ultimately in the resus room with an unstable patient it’s hard to know if they’re substantially bleeding or not (unless it’s the type of bleeding that you get wet feet or have to change your scrubs after) which I presume is why they used this definition.
The downside is that they included lots of patients who didn’t seem to actually have major bleeding – only 50% in this study received any blood products at all. Presumably they got randomised, got their treatment and then either stabilised or had a -ve CT or both and had no need for blood.


the reason this trial was so big is that they were looking for a small improvement in mortality and that’s just what they found. It looks like a fairly easy way to get that improvement.

The problems in the inclusion criteria are real but it’s worth remembering that the way they set up the study will tend to include less sick patients and I would imagine that would reduce the observed treatment effect.

The other thing to remember is that we only expect tranexamic acid to reduce death from bleeding not the other causes so note that bleeding was only cause of death in about 1/3 of the patients who died. Again this should reduce not increase the treatment effect.

So in the sicker, truly bleeding patients (say with mortality rates of 50%) then you’d expect the improvement to be greater than 1.5%.

They had a 70/30 split between blunt and penetrating injury which is probably more than us but I have to say I found this fairly compelling

Any thoughts? Anybody used this yet? 

For the ultimate brief-take on this check out the always excellent NNT

5 Replies to “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

  1. Pingback: The LITFL Review 020 - Life in the FastLane Medical Blog

  2. Pingback: CASE #6 UPDATE – FIND THE BLEEDING, STOP THE BLEEDING « Rural Doctors Net

  3. Pingback: Trauma! Massive Transfusion • LITFL • Trauma Tribulation

  4. Pingback: Managing the Critical Bleeder!! • LITFL • Hematology Hoodwinker

  5. Pingback: R&R in the FASTLANE 015 • LITFL

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.