post

Thoughts on the WOMAN Trial

Postpartum haemorrhage is really quite a broad term. I frequently manage women with retained placental tissue who require manual removal in theatre to halt persistent bleeding. This is routine, mundane Obstetric care. We follow well-trodden protocols. Most of these women do well. We sometimes need to resort to steps 3 or 4 in our algorithm but we know that once we get the tissue out, things will likely settle. Unfortunately, sometimes the bleeding doesn’t stop. These are the scary ones…

Massive PPH is a very difficult scenario. We empty our blood bank of products, we labour late into the night. This is a complex team resuscitation involving multiple surgical, pharmacological and “old-fashioned, elbow grease” techniques. I work in a remote hospital with limited staff and resources – this scenario depletes our reserves completely.

Twice in my career I have had to walk into a room and tell a new father that his partner has died within hours of birth. I have had to grasp a babe from that young man’s arms as they have fallen limp in anguish. These are the worst moments of my professional life.

The sudden death of a healthy, young woman at a time of such great joy and expectation is the most tragic event I can imagine. The death of a mother can devastate a family, destroy a team and evaporate the community’s confidence in our care. I desperately want to do whatever I can to never see that again. I am hopelessly biased towards anything that may save that precious life. Conflict declared!

The WOMAN Trial was published in the Lancet last week. This is a massive, international trial that sought to examine the effect of IV tranexamic acid (TXA) on maternal mortality in post-partum haemorrhage. Those of us who work in Obstetrics have been eagerly awaiting the outcomes of this trial since its methodology was published. This is high quality information, similar to the CRASH-2 data in trauma. 20,000 women with rigorous randomisation, protocol adherence and follow up.

Importantly this trial was conducted in multiple countries including those with both the highest burden from maternal death and with the least resources. So, can a simple, cheap and readily-available medicine save the lives of women giving birth? The answer….. umm. MAYBE…

As is often the case when landmark trials such as this (think: IST-3, CRASH2, PROPPR, PESIT) are published there is a wave of euphoria as the results are discussed in mainstream media. There is then the inevitable realisation that maybe the data is not so clear. Maybe the conclusions are overstated or understated, depending on one’s prior biases and beliefs. After a period of sober reflection we can all take time to think about what we have really learned from this new information.

We cannot see where we are going with this new data until we know where we are now. So Iets see where TXA fits prior to the WOMAN trial. Then I will discuss how I believe this new data can be applied.

First, some basic science. Late pregnancy is a time of physiological oddity. The coagulation system goes into overdrive, there is an increase in clot formation and fibrinolytic activity to match. This is why the normal range for D-dimer is elevated as the pregnancy advances. The peripartum period has an incredibly high incidence of thromboembolic disease. PE is, at times, the number 1 killer of women in the first world (sepsis and PPH battle it out for places). As such we are very cautious in Obstetric care – bleeding is a big risk, but also is clotting. ‘Tis a tightrope of thrombin our patients walk! If we are going to interfere with this delicate balance then we need to be safe.

Let us talk about the safety that WOMAN demonstrates. Tranexamic acid is a very weak tilter of the thrombosis balance. TXA doesn’t magically turn off the bleeding tap, it has subtle effects. One would not anticipate a large effect on mortality. It is not like a thoracotomy, this is not the stuff of heroics. TXA is cheap and can be given by any nurse, midwife or any clinician who can place a drip. So why are we not doing this already?

Easy, yes. Safe? Until now we have been unsure. Previously, I have been reluctant to use it empirically (i.e. In the absence of documented hyperfibrinolysis ). The benefits were unclear, but the safety in parturient women was also unknown. In my mind it was restricted to cases where we had our backs to the wall or had thromboelastographic data showing hyperfibrinolysis.

The WOMAN trial gives us the best data to date ( unlikely to be repeated soon ) that this is a safe intervention. So if we were to roll it out in our standard empirical PPH Resus protocols then at least we are very unlikely to do any harm. Agree? I think this is a reasonable interpretation of the WOMAN data on safety. Prima non nocere… CHECK.

So now onto efficacy. Where do we currently stand with TXA in PPH? What is it’s current role in modern Obstetric / Obs Anaesthetic practice?

The experience from centres that routinely use ROTEM is that almost all women with a decent PPH demonstrate rapid clot lysis and that TXA can correct this to some extent. In these centres TXA has been effectively been a “standard of care”, based on ROTEM, for some years. This is critical to understand before we start discussing p-values and statistical significance.

The next point to consider is the potential diluting effect that “minor PPH” has on the WOMAN data. More than half of the women had an estimated blood loss (EBL) less than 1000 mL. These women rarely feature in mortality data in my experience. Approximately 20% of those recruited had a “massive PPH”, with EBL over 1500 mL. I could not find a sub-group analysis of the mortality for this group in the paper. I would love to see it. Did TXA have a more potent effect in those at higher risk of death? My guess is that the benefit is higher in this group. This remains a guess… if you know, please tell me. So should we reserve TXA for women with ‘big bleeds’?

No, that is not how PPH works. We cannot predict which woman will settle, and which will continue to bleed to become a “massive PPH” with much accuracy. The strategy is to anticipate and prevent further bleeding. We treat empirically and aggressively with oxytocics and manual compression in order to avoid falling into the spiral of coagulopathy and more bleeding. If we were to use TXA, it would be early in the algorithm. My imaginary protocol would involve empirical TXA bolus at the time that a PPH Emergency is declared, as the theatre staff are being alerted. We typically give antibiotics, antiemetic and acid-suppressant at this point. TXA would fit in here.

Some commentators on the paper have suggested that in the absence of a clear cut benefit ( based on p-values… more on that later ) makes TXA a useless distraction in the time of intense activity. Well… yes and no. If given early as part of a PPH package of care it would take only minutes for one team member to complete the injection. Our Labour Ward teams are well drilled in these scenarios. In my experience there is always an awkward pause in the transition from the labour ward to operating room. There is usually ample time to load our patients with all the drugs we need to optimise the upcoming anaesthetic.

Now, let us discuss statistics- that’s why you started reading this post! The raw numbers do not lie. The primary end-point, “death from all causes or hysterectomy”, was not altered by tranexamic acid use.

Instead we see a small reduction in “bleeding death” and secondary outcomes like “need for laparotomy”. The reduction in laparotomy is particularly interesting when you consider that much of the world’s population is born in places where surgery is not accessible. The remainder of the analysed outcomes were unchanged.

The study had a few technical / statistical problems. There was a change in the calculations to power the data from 15,000 to 20,000. The expected mortality of 2.5% was higher than that observed, at ~ 1.7%. As such the trial was less likely to find a benefit as it was structured.

There was also the odd use of a composite primary outcome “death or hysterectomy at 42 days”. The authors appear to have thrown this out as it became clear that using hysterectomy muddied the waters. Many hysterectomies were done early- so any benefit from TXA would be removed along with the uterus. I think it would be better to look at a hard, patient-oriented, singular outcome eg. all cause mortality alone.

The authors’ discussion of the data is very enthusiastic. They focus on the reduction in “bleeding death” from 1.9% to 1.5% – this was statistically significant. That translates to an NNT of 250 for bleeding death. As in CRASH2 trauma, this effect was only seen if TXA was given prior to the 3 hour mark. If you only include the women who got TXA < 3 hours post partum the numbers improve to a 0.5% absolute risk reduction or an NNT of 200. (Yes, I realise that these are all secondary outcomes, and that it is statistically naughty to calculate NNTs for these.)

Now about p-values and statistical significance. I prefer to take a Bayesian approach to application of p-values ( yes, that is a thing). Basically, you have a prior estimate of the likelihood that TXA is useful, you can convert p-value into a “Bayes factor” akin to a likelihood ratio and you arrive at a new likelihood that TXA helps. I won’t go into detail here as I have a super nerdy guest coming up later into the year who will explain it all in gory detail!

For me, TXA is already an accepted part of tertiary Obstetric PPH care in hospitals who use ROTEM in massive transfusion. My personal estimate of it’s likely utility would be in the 50 – 66% range. As such we really only need a weak signal from our p-value to improve this estimate.

Another way of looking at this is to consider TXA to be an existing ‘gold standard’, then imagine if this new data would make it redundant. I think that the weak signal towards benefit would maintain the status quo.

So here is the summarised version in bullet points:
– TXA is cheap, easy to use and widely available.
– we now have clear data from a large Obstetric-specific context of the safety of TXA
– it fits easily into our existing protocols and systems of care
– TXA needs to be given early (< 3 hrs)
– TXA is already widely used in many centres to treat hyperfibrinolysis in major PPH
– we are not great at guessing who will develop major bleeds
– we should use it empirically and early if we do so
– the WOMAN trial did not demonstrate a clear mortality benefit.
– it remains unclear if there was more benefit in the more severe PPH cohort
– there is a suggestion from this data that bleeding-death is reduced
Ok a summary of that summary:
– we already thought it helped, it doesn’t do any harm. Carry on.

I copped a bit of criticism online when I described the WOMAN trial as a “Game changer”. Based on the numbers, that is probably fair criticism. This trial doesn’t really change the game, though it depends on how you were playing the game before you read it.

If PPH management were a game – then it would be chess and having TXA in your protocol is a bit like having an extra pawn at the outset. Chess is complicated and predicting what will occur in 20 moves is impossible. Your TXA-pawn is a weak player, it probably won’t change the outcome very often. However, occasionally that small, cheap piece can make the difference between victory and defeat.

At a population level, an NNT of 250 and a cost of $20 a dose equates to about 5 grand to save a life, a wife or a family. This seems like a good investment to me, but as stated, I am hopelessly biased.

The late Dr John Hinds coined the term #ResusWankers to describe folk who are unhelpful in moments like these, or in the dissection at the aftermath. In his plenary talk in Chicago ( SMACCUS) he described a hypothetical man who dies of a traumatic chest injury, but no thoracotomy was performed. To the grieving family the Doc says: “well, there was one thing we could have tried, but it is only successful occasionally…”
Is this a talk you want to deliver?

Comments

  1. Dyfrig Hughes says:

    As one of those who questioned the ‘game-changer’ comment (sorry!) I really appreciate this detailed evaluation by someone who is much more versed in statistical analysis than I am, so thank you.

    Your observation that “it depends how you were playing the game” is really important. I have never, mercifully, had to inform a husband of this worst of all news, as you have, and I never wish to, least of all without being able to say “we did everything we could”.

    • Twitter is not a good forum for this debate ( or any really). Agree “game changer” is a useless description- not really a way to analyse complex data
      More gold on the application of Bayesian data interpretation coming soon… just for you mate

      • A Bayesian interpreation of WOMAN trial results. (Fuller treatment than bayes factor)

        For death due to bleeding outcome :

        Using a weak informative prior .This prior has location of zero and scale of 1. is recommended by Gelman and pedroza et al. for odds ratio in particular. This prior assumes that in general clinical studies have mean odds ratio of 1.000 with lower and higher 95% credible limits of odds ratio from 0.141 to 7.099 respectively . Combining study results with this prior information . The study has mean odds ratio of 0.806 with lower and higher 95% credible limits of odds ratio from 0.652 to 0.997 respectively . The probability that treatment is better than control is 97.641 percent. The probability that treatment causes 10%,20% and 30% Risk reduction is 83.357, 43.556, 7.523 percent respectively . The probability that treatment causes specified 12 % reduction is 77.405 percent .The probability that treatment effect lies between 2.5 and 0 percent reduction is 1.734 percent

  2. Thanks Casey, for breaking it down for us all.

    I’ve been involved in a lot of nasty PPH cases with the retrieval service and often late in the piece after the massive transfusion protocol has been activated and Bakri balloons inserted. I’ve always recommended TXA, essentially under the “It can’t do any harm approach” and I don’t think this trial will change my approach. What it does do, though is give us a minor pause for thought about those more ‘minor’ PPH’s (if there is such a thing).

    Does it potentially reduce transfusion requirements? I think that is a valid endpoint given that blood supplies are rapidly depleted in most rural obstetric practices?

  3. Thankfully never had to tell a new Dad anything that bad. Shit position to be in for you. Good we have one more tool in the armamentarium to prevent that outcome.

  4. Roger Browning says:

    Great analysis Casey,

    This is not a straight forward study to interpret and like CRASH2 I anticipate there will be a lot of different opinions and discussion on this!

    My take is, like yours Casey also inherently biased by my preconceptions and current practice (pretty liberal use for the last 5-6years)!
    Points to note:
    1 -- No evidence of harm, and a sample size of 20000, is very reassuring.

    2 -- Slightly surprising that there was no overall decrease in mortality from all causes, but reassuring that it decreased death from bleeding. (Fibrinolysis and TXA is only a small part of the overall scenario).

    3 -- Probably better to give TXA earlier rather than later.

    We have learnt from our use of ROTEM at KEMH over the last 5-6 years that a small number of women develop severe hyperfibrinolysis -- and this process unchecked can rapidly lead to extremely rapid depletion of fibrinogen (and platelets). If you miss this diagnosis and don’t treat it you can have real trouble….. Sooo if you can’t test for this (you need TEG or ROTEM) and the patient is still bleeding then you probably should just give TXA to make sure you have it covered… my humble personal opinion.

    So how will WOMAN change my practice ?

    -- I will continue to do what I already have been doing -- pretty much give TXA 1g to everyone who is having a significant PPH (both at the big tertiary hospital with ROTEM and the smaller peripheral hospital without an onsite lab where I also work). I think it decreases overall blood loss significantly (not in this study perhaps but evidence form other studies in the developed world) is compelling and there is good evidence it causes no harm.

    Remember though the most important treatment for PPH (and all haemorrhage) is to STOP the bleeding -- it’s not TXA or the blood products!

    Like Casey said a large number or even the majority of PPHs are well semi -- routine almost but every now and then you have one where you crap your pants (pardon the French Casey!) If you want to save someone’s life and I had to recommend 3 treatments which can be instituted within minutes pretty much anywhere :

    1 -- Learn how to perform manual aortic compression -- https://www.youtube.com/watch?v=rc9BYcIhamA

    2 -- Tranexamic acid

    3 -- Fibrinogen concentrate

    (4 -- O negative / blood would be good too -- but not always immediately available)…

    Here’s the link to our brief discussion of hyperfibrinolysis from our ROTEM teaching package.
    https://www.dropbox.com/s/vwn4hn3ijheneq7/2%20Fibrinolysis.mp4?dl=0

    Or the link to the whole ROTEM learning folder in dropbox:
    https://www.dropbox.com/sh/t054a1p7v1mopc4/AAAdWoCurMd3vflNPs7psPnHa?dl=0

  5. Excellent work Casey. I think this is one of the benefits of #FOAMed in nuancing the discussion around evidence. You and people like Ken Milne do this so well. EBM is not just about the numbers. but also about what they ‘mean’ to my practice and patients. The best clinicians are those that understand this.

    As for this trial I agree with your thoughts. Interestingly, I often challenge colleagues to give me the NNT of many of our treatments. Usually they are unknown or wildly optimistic. An NNT of 200 for something this cheap is really not that bad.

  6. Thanks Casey -- a really useful discussion.

    As the person who is usually at the business end of the PPH (thank goodness for waterproof shoes) I feel that any harmless intervention that might potentially tip the odds in our favour is worth doing.

    It’s certainly not going to stop us from clipping spurting arterial bleeders in the perineum and bimanually compressing the uterus and giving synto and ergo and IDC and all the other 20 things we always do. As you say, it’s a team sport -- every individual intervention helps swing the balance in favour of a healthy mum.

    As others have suggested, the effect on blood transfusion volumes would be interesting, as would data around other soft complications of PPH like low milk supply, neonatal weight etc.

    Thanks again.

  7. Glenn Barker says:

    Hi Casey
    Can I ask how you and others are administering the TXA in this context.
    The WOMAN trial methods section states a 10mg/ml solution given at a rate of 1ml/min, which would be 1000mg (in100 mls) over 100mins.
    (Although the actual participants protocol (http://womantrial.lshtm.ac.uk/wp-content/uploads/2015/10/WOMAN-Trial-Protocol-v1.1.pdf ) implies the gram is made up into 10mls ( and therefore given over 10mins) without explicitly saying that)

    This is obviously different from the loading dose in trauma (1g over 10mins)

    • After tweeting through the trial’s official twitters feed I have found the answer.
      Strangely the actual protocol was quite unclear on the exact dilution and delivery in the text!

      They essentially did the same as the initial bolus phase of the CRASH2 trial
      Simply:
      2x500mg ampoules ( we have 1000mg amps in Australia)
      1000 mg diluted to 10 ml
      Equivalent to 100mg per ml
      Injected at 1 ml/ min over 10 minutes

      Hope that is clear
      As an Anesthetiser- any slow push drug that takes longer than 2 minutes requires a syringe driver. Nobody has the patience for 10 minutes!!
      Casey

  8. Nicholas Adams says:

    From and evidence-appraisal perspective:
    1. The trial was unable to demonstrate a survival benefit overall (the primary outcome)
    2. Sub-group analyses are notoriously unreliable (whether post-hoc or pre-hoc) and should be ignored
    3. If there is a mortality benefit in the death due to bleeding group but no benefit overall there must have been an excess of mortality NOT due to bleeding in the treatment group (n=379 in TXA group, n=355 in placebo group)
    4. Bayesian priors are entirely subjective and thus essentially useless

    Summary -- this trial provides no evidence to support the use of TXA in this situation. This cost of TXA is irrelevant. The distressing nature of PPH is irrelevant.

    • Thanks Nicholas
      Fair comments
      I would point out these arguments can be applied to much of what we consider “standards of care” in much of medical care.
      I think time will tell. Safety is demonstrated- so as it gets used more we will see more data

      I’m really interested to see data on the > 1500 ml PPH- that would be the place to look for benefits that are more tangible
      Casey

      • Nicholas Adams says:

        Safe but ineffective = homeopathy.
        I would think one would have better things to do during a massive PPH than administer an ineffective drug.
        Beware commission bias!

        • I believe I addressed this concern in the above post.
          Women routinely get litres of IV crystalloid which is not helping / likely harmful in PPH.
          Should we keep doing this until an RCT shows benefits?
          This is more complex than yes/ no
          If TXA is used it must be as part of a comprehensive protocol-
          C

    • Rory Miller says:

      These comments are true with ready access to theatre and massive transfusion products/protocols.
      In the Glorious world outside of the city this is not (always) true.

      TXA is a very relavent drug and this trial for me (like CRASH2) just reinforces my use of this.

    • not sure where you got n = 379 and n = 355 from. what data are you referring to?

Leave a Reply to Roger Browning Cancel reply

*