If you have been listening to the Broomeodcs podcast in the last 5 years you will have heard Justin and I discuss a bunch of trials that examine the utility of tranexamic acid (TXA) for indications from epistaxis to intracranial haemorrhage. In summary, since the CRASH-2 trial was published in 2010 there has not been really any robust randomised trial data to support the use of TXA for any indication (depends a bit on which outcomes you care about).

Specifically in the Obstetric literature the WOMAN trial was debated extensively over the ether… my thoughts can be found here. The WOMAN trial did not show a reduction in all cause mortality, but was declared a positive trial as its primary outcome was “risk of death from PPH” was reduced by 0.4% [ note: CI 0·65–1·00; p=0·045… so touching insignificance]. So at best we think that TXA might have a very marginal influence in women who have established post-partum haemorrhage.

The same French group published the original TRAPP trial in 2018 – that trial looked at women who had a vaginal birth and asked if the use of TXA would reduce the rate of PPH.  The TRAPP trial was negative: PPH occured in 8.1% in the tranexamic acid group and 9.8% in the placebo group (relative risk, 0.83; 95% confidence interval [CI], 0.68 to 1.01; P=0.07 [sooo close!]

This is our starting point for the TRAPP2 trial.

The TRAPP2 TRIAL was published on April 29 in the New England Journal of Medicine. This trial is the latest in the TXA trials aimed at reducing bleeding. So let us do a deep dive into the trial to see what they did and what they found….


  • 4551 women recruited from 27 French maternity hospitals .
  • expected to undergo cesarean delivery (they did include elective C-sections and women in labour) 
  • All aged 18 years of age or older
  • completed 34 or more weeks of gestation


  • All women received a dose of uterotonic (either 5 or 10 IU of oxytocin or 100 micrograms  of carbetocin)
  • They then received 1000 mg tranexamic acid as a slow push (30 – 60 secs) within 3 minutes of birth.


  • The control group also received the prophylactic uterotonic as is standard practice in France and placebo in a 10 mL syringe that was identical to the TXA.


  • The primary outcome was postpartum haemorrhage,
    • this was defined as a calculated estimated blood loss greater than 1000 ml [this is a common cutoff used in the Obstetric literature] or
    • a red-cell transfusion within 2 days after delivery
  • The actual volume of blood loss was determined using a somewhat unusual process that seems to be prone to plenty of errors… it goes like this: 
    • estimated blood loss = estimated blood volume × (preoperative hematocrit−postoperative hematocrit) ÷preoperative hematocrit;
    • assuming blood volume = 85 mL / kg
    • * This calculation is not EVER used in practice.  Most centres use Clinician estimates or weigh the blood loss to estimate volumes.
    • ** Haematocrit is prone to serious fluctuations in the post-partum period this therefore introduces serious bias potential

That all seems like a reasonable way to answer the question right? Well, maybe… lets run the quality checklist over the TRAPP2 trial:

Quality Checklist:

1. The study population included patients we want to know about?  Yes, these are the women we see everyday in the Obstetric theatre.

2. The patients were adequately randomized?  Yes, the patients were randomised by a computer generated program in 1:1 ratio.  The patients were randomised in two cohorts – those in labour and those not in labour in order to ensure equal representation in both groups.

3. The randomization process was concealed.   Yes it was.  

4. The patients were analyzed in the groups to which they were randomized.    Yes, there was both an intention-to-treat analysis and a subsequent per-protocol analysis.

5. The study patients were recruited consecutively (i.e. no selection bias). This is unclear.  Women were selected if their clincal provider thought it was likely that they were going to undergo Caesarean section. They were given information late in their pregnancy and those who consented were included.  It is possible that some bias may occur if there was selection of women who believed they were at risk of complications and hence declined to consent.

6. The patients in both groups were similar with respect to prognostic factors.    Yep, the baseline characteristics of both groups were very evenly matched for demographics and the characteristics that might predispose to bleeding.

7. All participants (patients, clinicians, outcome assessors) were unaware of group allocation. Yes, both the clinicians and the patients were blinded by the unlabelled syringe placebo. The same clinicians performed the outcome estimates of blood loss for the primary outcome.

8. All groups were treated equally except for the intervention.Yes, according to the data provided there was no difference in other interventions

9. Follow-up was complete (i.e. at least 80% for both groups).  There were about 2,200 women randomised to each group.  Unfortunately bout 500 in each group did not receive the dose of TXA / placebo within the defined 3 minute window.  Of those who did receive the intervention per protocol there was a good follow-up –  about 90 in each group had data missing and were not able to be included.                                                      

10. All patient-important outcomes were considered.   Hmm, not really.  The primary outcomes were largely numbers… volumes and arbitrary cutoffs.  There were Patient Satisfaction and Edinburgh Depression scores as secondary outcomes.  There was no significant morbidity or mortality to measure – so that is good news.  The need for emergency surgery or adverse effects of TXA were measured but these were rare.

11. The treatment effect was large enough and precise enough to be clinically significant.   Once again – this is a tricky one.  If we take the primary outcome as a dichotomised definiton of PPH ( > 1000 mL) then there was a 4.9% reduction in the TXA group.  The p-value on this was 0.003

However, when you look at the actual average volumes in the two groups there is only a tiny difference – 689 mL vs. 719 ml… this is clearly never going to be clinically significant.  Given the somewhat fuzzy, mathematical formula used to arrive at these numbers I do not really think it can be assumed to be significant in “real life”.  More importantly; this is never going to be a number that a woman cares about.

The most “significant” outcome number in the whole trial was the incidence of vomiting in the post-operative care unit …(43.0%  in the TXA group vs. 36.3% in placebo, adjusted P<0.001).  Thats a NNC (number needed to chuck) of 15!  This was news to me.  I don’t think I would want to use TXA in a previously well woman and have her risk of vomiting at the stage where she is enjoying her first cuddles.  This is not worth the marginal benefits if any.

As always – the risk of thrombo-embolic event was measured though very rare… actually a little too rare only 8 in the TXA group and 2 in the placebo group… so 10 women out of 4100 women having a C-section… a 0.24% VTE rate… the best prior estimate I could find was 0.3% looking at all retrospective data.


Overall this was a well designed and rigorously conducted trial.  There are a few unusual aspects to the design that would limit the generalisability of the outcomes.

Despite the “positive result” I think that this is a negative trial.  Why?

Well, its all about the Bayes.  Given what we know about the prior data from the WOMAN, TRAPP and other smaller trials I would guesstimate the likelihood of prophylactic TXA actually working for women undergoing routine C-section (that is those without a declared bleeding emergency) at about 5%.  I just do not think it will work as the physiology would suggest.  If TXA doesn’t have a major effect even in women who are actively bleeding – then it is asking a lot of a simple amino acid to fix a problem that doesn’t yet exist!

Therefore, in order to say that the TRAPP2 data moves the Bayesian needle to a significant likelihood of effect we would need a really spectacular result.  The Bayes factor here based on a p = 0.003 (being generous with the dubious volume calculations) would get us into the 40% range.  However the practical effect of this is a tiny reduction in blood loss which may be statistically significant, but is clinically meaningless.


The risk here is over-extrapolation of this result.   If you believe it is a real positive trial then logically every woman should receive TXA at elective C-section.  You may want to but some more emesis bags for the recovery room?

However, if like me you are a skeptic and don’t believe the data changes practice you won’t be reaching for TXA…. but remember this is a well, not-bleeding population.  Don’t throw the TXA out with the amniotic fluid…  we still do not have much data to answer the question about the utility of TXA in the very severe bleeds.  The literature described above just does not include enough women with major bleeds to answer that question.

We still need to treat sick, bleeding women on a physiological basis.  We know from ROTEM studies and practice that TXA has a role in treating the hyperfibrinolysis that plagues post-partum bleeding patents. Empirical treatment in this group still has a place.  The risks remain small, where the potential benefits are increased in severe coagulopathy.  So for that group I will continue to use TXA for now until we have more data


Add a Comment

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