How CRASH-2 got it wrong – A response by Dr. Mark Walsh

Some great questions came out of Dr. Mark Walsh’s lecture on TEG, so we decided to go straight to the source and pick the HUMONGOUS brain of Dr. Walsh and get his thoughts on the CRASH-2 trial.  In this 15 minute review, Dr. Walsh will make you reconsider what you once held true about TXA.  Dr. Walsh – we salute you.  An amazing evidence-based rant that should reminds us all about the importance of numeracy as well the need to critically appraise the literature regardless of what “everyone” is saying!

He was also kind enough to solidify his response with a publication worthy, dissertation-esque, written response!!

Listed below is his defense in it’s entirety:


A response to the “passionate” criticism of the CRASH-2 trial. We have been down the road of mass produced “science” before: CRASH-2 is not the child of the original CRASH trial.

Mark Walsh MD 1,2 and Braxton Fritz BS 1,2

1:Memorial Hospital of South Bend, South Bend, IN 46601
2:Indiana University School of Medicine South Bend at the University of Notre Dame campus, South Bend, IN 46617

Derek Sifford is concerned about my “passionate” objection to the Clinical Randomization of an Antifibrinolytic in Significant Haemorrhage (CRASH-2) (1) trial in my pod cast at the University of Maryland Shock Trauma Center on September 5th, 2013. I thank him for the compliment regarding the lecture on the Thromboelastography/Rotational Thromboelastometry (TEG/ROTEM) and for his questions which shed light on the trial known as the CRASH-2 trial.

He accurately identifies one of my concerns as the recruitment of patients based on the “uncertainty principle” whereby the inclusion into a study of patients in mostly low to moderate income developing countries with varying levels of trauma systems depends on the “uncertainty” of the need of tranexamic acid (TXA) by the individual physician caring for the patient. That is, those who really needed the drug received the drug and those who clearly didn’t need it, did not receive it. This left more than 20,000 patients who may or may not have needed an anti-fibrinolytic who were randomized. We feel there are serious problems with the CRASH-2 trial that have been very well delineated by Napolitano et al. in the June 2013 issue of the Journal of Trauma. They have analyzed the CRASH-2 data carefully and have found major problems with this study. (2) These problems were the following:

  1. Only approximately 5% of patients had bleeding as a cause of death.
  2. The CRASH-2 approach to randomization. The CRASH-2 wording is: “Doctor is reasonably certain that antifibrinolytic agents are indicated or contraindicated – Do not randomize”.
  3. Concern regarding selection bias.
  4. No data regarding injury severity of the patient cohort.
  5. No data regarding shock in the patient cohort (i.e. lactate and base deficit) and there was the inability to determine if the cohorts were similar.
  6. Small sample size of hypotensive (SBP < 90 mm Hg) (31.5%) and tachycardic (HR>107) (48%) patients which were the target populations.
  7. No data regarding fibrinolysis on admission and no coagulation testing. The rate of fibrinolysis at admission in North American trauma centers is approximately 5%.
  8. The most common cause of death was traumatic brain injury (TBI).
  9. TXA did not reduce blood transfusions. Only 50% of study cohort received blood transfusions.
  10. No adverse events were regarded as serious, unexpected, or suspected to be related to the study treatment.
  11. Concern about possible inadequate reporting.
  12. Patient follow-up reported as 100% which is difficult to believe.
  13. Effect size was small. This effect was statistically significant but not a clinically meaningful finding. The study determined a 0.8% absolute reduction in “death caused by bleeding”.

Regardless of how large or randomized is this well funded study, we feel that the CRASH-2 is a monumentally flawed study with no specific mechanistic rational.

The more applicable and useful study is the Military Application Of Tranexamic Acid for Trauma Emergency Resuscitation study (MATTERs) (3,4) which evaluated those patients who clearly needed an anti-fibrinolytic. In this study those military trauma patients who needed at least a unit of blood were divided into TXA or no TXA arms. Unlike the subtle CRASH-2 benefit, the MATTERs results were striking. The number needed to treat (NNT) was 1:7 in the MATTERs trial while the NNT was 1:67 in the CRASH-2 trial. The relative reduction in mortality was a very subtle 1.5 percent in the CRASH-2 trial and the mechanism for survival remains unclear in this trial since half of the CRASH-2 trial patients did not receive blood. The relative reduction in mortality was 6.7 % in the MATTERs trial and those who received TXA received less blood products. In the CRASH-2 trial TXA patients received the same amount of blood as those who did not receive the drug. (1) The question remains, why include patients who probably will not need blood products into a trial that tests the ability of TXA to reduce blood product use?

Derek Sifford refers to the Prehospital Treatment of Acute Traumatic Coagulopathy and Hemorrhage (PATCH) trial from the Alfred Trauma Center in Australia as a study that will help clarify practical and mechanistic questions regarding the use of TXA in trauma resuscitation.(5) The PATCH trial from Melbourne uses criteria that select out hypotensive patients who will probably need blood products based on a 7 point numeric system called the COAST criteria. Most of the patients from this trial will have blood pressures less than 100mmHg and have well documented and significant pelvic and or abdominal injury. One of the stated reasons to perform the PATCH trial is that the results of the CRASH-2 trial may not be applicable to the economically developed countries where patients are treated more quickly. Even the authors of the PATCH trial found the CRASH-2 trial problematic and as justification for their study they address the same points that Napolitano et al. delineated. The PATCH authors note that: “Thrombotic complications were reported very rarely in the CRASH-2 study (PE, 0.7% of all patients; DVT, 0.4%),(1) probably because they were not actively sought in many of the participating hospitals. In contrast, the MATTERs study showed that rates of PE and DVT among patients who received TXA were, respectively, 9 and 12 times the rates among those who did not.” (3,5)

Scott Weingart refers to the dilemma of the lack of sensitivity to determine fibrinolysis by the TEG/ROTEM. John Greenwood referred to Larson et al. who pointed out that there is a need for a sensitive test which can detect clinically significant fibrinolysis.(6) Both points are true and are summarized very nicely by Schochl and others review on TEG/ROTEM in trauma in the same June 2013 J of Trauma issue as Napolitano’s excellent review of TXA use in trauma.(7) Currently there is experimental work to predict clinically significant fibrinolysis by plasmin antiplasmin (PAP) levels. (7) Whether the PAP level is another D-Dimer or Fibrin Split Product, both sensitive tests that are of no clinical use in acute trauma, remains to be seen. What still remains is that TEG/ROTEM are bedside point-of-care whole blood coagulation tests that accurately reflect the ability of the blood to clot and there is nothing else currently available that can replace them. At this time there is no adequate definition of clinically significant fibrinolysis. The range of published clinically significant fibrinolysis varies from 15% to 3% of lysis at 30 minutes using TEG/ROTEM. (8-10) Future studies will determine the threshold that is specific and sensitive enough to trigger administration of TXA for trauma resuscitation. For now, the TEG/ROTEM is all we have to rapidly define clinically significant fibrinolysis.

Whether the drug is harmless does not justify its administration to patients who do not need it. There must be a proposed mechanism for the drug’s effect in trauma. The orthopedic surgeons have found TXA useful in limiting blood transfusions in elective hip and knee surgery, but not useful for acute fractures because of thrombotic complications. Why are multiple trauma patients different than the orthopedic patients?

Rather than follow the “Bolshevik” science, which is the word that I used in my talk to describe this massive Government Funded Study in a high income nation that was inflicted on a low to middle income world populace, one might listen to the above mentioned analysis of Lena Napolitano, who trained at Maryland Shock Trauma and is the current Division Chief of Acute Care Surgery, Professor of Surgery and Associate Chair for Critical Care at the University of Michigan.(2)
We have been down this road before with the arbitrary administration of a “harmless” dose of methylprednisolone to patients with cervical spine injury based on an NIH funded National Acute Spinal Cord Injury Study (NASCIS) that was foisted on the emergency physicians with a prepublication printed mailing to most emergency physician’s homes or Emergency Departments in the United States in the early nineties. (11) Subsequently, this unfortunate and massively flawed study put methylprednisolone into countless algorithms and guidelines. Much like the CRASH-2 trial, this study was a well funded government study without mechanistic rational that described a small and equivocal benefit.(12) Those on methylprednisolone administration had increased complication rates of severe sepsis and severe pneumonia. (13) No amount of prepublication “advertisement” of the original flawed study with the subsequent incorporation of methylprednisolone use into the guidelines for treatment of spinal cord injuries could make the original study actually true. It took a generation of trauma physicians pleading for common sense to expunge this unfortunate study from the scientific community and the societies that represent trauma medicine. (12)

Those of us who are driven by a physiologic worldview where a reason must be given for a therapeutic administration of a drug, are concerned by the similar confusion that the CRASH-2 trial has wrought on the trauma community. There is a double irony of the CRASH 2 trial in that the first CRASH trial, Corticosteroid Randomization After Significant Head Injury, exposed the dangers of steroid use in TBI which was encouraged by the NASCIS legitimization of steroid use for trauma related neurologic injuries. (14-15) The second irony is that in the CRASH-2 trial, TXA’s true advantage has been mitigated by the fact that those patients who may have needed the drug were not studied and the true benefit of the drug may have been understated. It took the MATTERs trial to describe the true population that would benefit from TXA. However, the MATTERs study showed that there is a real risk of thrombosis in the bleeding patient who might need the drug.

Finally, regarding my “passionate” objection to the CRASH-2 study, the passion is for science….not Bolshevik like dogma and guidelines imposed on physicians with common sense by government funded researching herds driven by the grant writing bureaucrats who dole out funding to their willing accomplices in return for grants and prestige. Derek Sifford’s response to my concerns about the intellectual inconsistency of the CRASH-2 trial is met with an off handed comment that he does not have a TEG on his helicopter. No one has or needs a TEG on a helicopter. What one needs is the following: common sense and a rationale for action.


Phew…. So after reading through that extremely intelligent and researched rebuttal here is a SUMMARY: 

  • CRASH-2 trial has several limitation, mainly based upon the lack of a defined requirement for the use of TXA (no bleeding) and the small sample size of truly sick (hypotensive, shock-y) patients
  • CRASH-2 also does not adequately show a clinically significant outcome: no transfusion reduction, no clinically relevant mortality benefit (i.e. 0.8% absolute reduction in “death caused by bleeding”….)
  • The MATTERS showed: 1) A true benefit for the use of TXA (NNT was 1:7 in the MATTERs trial while the NNT was 1:67 in the CRASH-2 trial); 2) a relative reduction in mortality of 6.7 %; and 3) a HIGHER risk for thromboembolic events!! (something strangely missing from the CRASH-2 trial…..)
  • The PATCH trial should be able to bridge the gap btw the CRASH-2 limitations and the use in the general trauma population (unlike MATTERS trial and military use)
  • TEG/ROTEM may not be the best at detecting fibrinolysis, but it is certainly the best we have! 
  • It is better to question literature instead of taking a “government funded, highly publicized” study for absolute truth (lessons from NASCIS advocating methylprednisolone for spinal cord injuries)

We welcome any and all responses and debates. BE PREPARED to bring your A-game. You know Mark Walsh will…….


1. CRASH-2 Trial Collaborators. Effects of tranexamic acid on death, vascular occlusive events, and blood transfusion in trauma patients with significant hemorrhage (CRASH-2): a randomized, placebo-controlled trial. Lancet. 2010;376:23-32.

2.  Napolitano LM,  Cohen MJ, Cotton BA, Schreiber MA, Moore EE. Tranexamic acid in trauma: how should we use it? JTrauma Acute Care Surg. 2013; 74:1575-1586.

3. Morrison JJ, Dubose JJ, Rasmussen TE, Midwinter MJ. Military Application of Tranexamic Acid in Trauma Emergency Resuscitation (MATTERs) Study. Arch Surg. 2012; 147:113Y119

4. Inaba K. Antifibrinolytics in trauma patients: does it MATTER? Comment on Military Application of Tranexamic Acid in Trauma Emergency Resuscitation (MATTERs) Study. Invited critique. Arch Surg. 2012;147:119.

5. Gruen RJ, Jacobs IG, Reade MC. Trauma and Tranexamic Acid. Med J Aus. 2013; 199:310-311.

6. Larsen OH, Fenger-Eriksen C, Ingerslev J, Sorensen B.Improved point-of-care identification of hyperfibrinolysis is needed. Throm Res. 2012 Oct;130(4):690-1.

7. Schöchl H, Voelckel W, Grassetto A, Schlimp CJ. Practical application of point-of-care coagulation testing to guide treatment decision in trauma. J Trauma Acute Care Surg. 2013; 74(6):1587-1598.

8. Holcomb JB, Minei KM, Scerbo ML, Radwan ZA, Wade CE, Kozar RA, Gill BS, Albarado R, McNutt MK, Shan S, Adams PR, McCarthy JJ, Cotton BA. Admission Rapid Thromboelastography Can Replace Conventional Coagulation Tests in the Emergency Department Experience With 1974 Consecutive Trauma Patients. Ann Surg 2012; 256:476-486.

9. Cotton BA, Harvin JA, Kostousouv V, Minei KM, Radwan ZA, Schochl H, Wade CE, Holcomb JB, Matijevic N. Hyperfibrinolysis at admission is an uncommon but highly lethal even associated with shock and prehospital fluid administration. J Trauma Acute Care Surg. 2012;73:365-370.

10. Chapman MP, Moore EE, Ghasabyan A, Harr JN, Chin TL, Ramos CR, Stringham JR, Silliman CC, Banerjee A. TEG fibrinolysis above 3% is the critical value for initiation of antifibrinolytic therapy. J Trauma Acute Care Surg. In Press.

11. Coleman WP, Benzel D, Cahill DW, Ducker T, Geisler F, Green B, Gropper MR, Goffin J, Madsen PW 3rd, Maiman DJ, Ondra SL, Rosner M, Sasso RC, Trost GR, Zeidman S.  A critical appraisal of the reporting of the National Acute Spinal Cord Injury Studies (II and III) of methylprednisolone in acute spinal cord injury. J Spinal Disord 2000 Jun;13(3):185-99.

12. Hansebout RR, Kachur E. Acute Traumatic spinal cord injury. Up To Date. 2013.

13. Eck JC, Nachtigall D, Humphreys SC, Hodges SD. Questionnaire survey of spine surgeons on the use of methylprednisolone for acute spinal cord injury. Spine. 2006; 31:E250-253.

14. Geisler FH. Excessively closed science hurts. BMJ. 2008;336:629.

15. Edwards P, Arango M, Balica L, Cottingham R, El-Sayed H, Farrell B, Fernandes J, Gogichaisvili T, Golden N, Hartzenberg B, Husain M, Ulloa MI, Jerbi Z, Khamis H, Komolafe E, Laloe V, Lomas G, Ludwig S, Mazairac G, Munoz Sanchez Mde L, Nasi L, Olldashi F, Plunkett P, Roberts I, Sandercock P, Shakur H, Soler C, Stocker R, Svoboda P, Trenkler S, Venkataramana NK, Wasserberg J, Yates D, Yutthakasemsunt S, et al. Final results of MRC CRASH, a randomized placebo-controlled trial of intravenous corticosteroids in adults with head injury-outcomes at 6 months. Lancet. 2005:365:1957-1959.

Share this Post


  1. Eoin Harty

    Hi Mark & Braxton

    Great article, impeccably researched. You’re completely right, the selection and randomisation in CRASH2 is completely flawed. I hadn’t picked up on the 100% follow up rate either – smells a bit fishy. Hopefully PATCH-Trauma will answer a lot of questions for us. It’s a long way off at the moment though.

    I’ve a few quick points

    1. Can you really compare CRASH2 and MATTERs in terms of drawing conclusions about TxA? Their study designs are about as different as designs can be, a prospective RCT vs a retrospective observational trial, civilians vs military, whole world vs one hospital in Afganistan…

    2. These increased risks of 9 and 12 times chance of VTE. I’ve read through MATTERs, and, while they refer to an increased risk in their discussion, I can’t find them anywhere. I think this might be my reading of the paper – I’ll read it again. They also say that this may be down to their sicker cohort of patients and a survivorship bias – their TxA patients are surviving long enough to get a VTE of some sort.

    3. It can be argued that the true benefit of CRASH2 is that now everyone’s talking about TxA – the orthopods are using it, the WOMAN trial will hopefully tell us how useful it might be in PPH, we’re using it in trauma (for better or for worse) – and people are talking about models of coagulation and fibrinolysis in trauma, which is great. Also we wouldn’t have MATTERs or PATCH-Trauma without it.

    4. Bolshevik science is a bit of a attention-grabbing exaggeration and also it’s totally unfair…
    (a) MATTERs is also 100% government funded (2 governments in fact! is that twice as bad?)
    (b) If we really want genuinely unbiased research, with registered protocols, untainted by the pharma industry, with full access to results positive or negative, then I can’t see any real way of doing this without involvement of the public sector or charities.

    …that’s what I think anyway. Look forward to being torn apart 😉


    (anaesthesia trainee, UK)

  2. rfdsdoc

    thanks folks
    interesting points raised
    You cannot claim that CRASH 2 got it all wrong without some kind of a response!

    1. CRASH2 and MATTERS showed a +ve effect of TXA. a RCT and non RCT study. To me that shows consistency of effect across different populations and methodology. Thats a big tick in my mind

    2. Selection and randomisation was completely flawed? How so?

    3. Several Australian prehospital and retrieval services including my own, RFDS QUeensland have implemented a prehospital TXA protocol based on CRASH 2 criteria ( declaration of possible bias here!). We have given it several times. PATCH trial is being done to try to provide more data on this patient group. thats clinical science at work

    4. The allegation of potential research fraud in relation to 100% followup is unfair. Prove it otherwise that should be retracted or at least taken up with the Lancet editorial office.

    1. John Greenwood


      Thanks for taking the time to contribute to the discussion. I think having people take a stance (right or wrong) is great to stir an academic debate about the strengths & weaknesses of any study – especially one that could potentially change the practice worldwide. It’s clear that many people have many different opinions! Here are some of mine…

      Agreed that MATTERS does support the potential benefit of TXA, excellent point. I’ve always taken the stance that having subjectivity in inclusion criteria does increase risk of bias, but perhaps not “completely flawed” but imperfect. Can’t wait for PATCH. I don’t think Dr. Walsh was saying that “fraud” was committed (that would be pretty harsh), but would guess he wanted more info about the follow-up process. 100% is a huge feat to accomplish – I can barely get a hold of 50% of my patients/emergency contacts. Apparently in Baltimore there are a large number of people with phone numbers that start with (410) 555 – xxxx.

      — John

  3. Jim Lantry

    Thank you all for your comments!

    I tend to agree with the majority opinion: All of these studies have limitations. The days of creating the perfect and flawless study are over. The limitations placed upon us by IRBs, funding, pt compliance to follow-up, etc… have restricted the leaps and bound we could previously make. I feel that CRASH2 did what it was suppose to do, create interest in TXA and lead to the creation of more clinically relevant and sound studies. I welcome the PATCH trial to fill in the missing pieces, just as we await the EOLIA trial to answer the questions left by the CESAR trial. The true mark of a scientist is to question the reality presented to him/her; and if these recent posts are any indication, we have been taking performing extraordinarily well in the “questioning” department!

    Keep the information coming. Dr. Walsh is mounting another wave of defense as we speak. We will post immediately upon receipt. STAY TUNED!!

  4. Stephanie

    Are there any studys out there other than the Crash 2 and the MATTERs ones? That give better clinical signifiacne?

  5. Simon Carley

    You claim ARR for TXA is 0.8% but in study the ARR is 1.5% (NNT of 67). You have mistakenly, perhaps deliberately, only referred to a subgroup outcome.

    The ARR at decision point is 1.5%, but even if you choose 0.8% that is still a clinically significant number and is comparable to many other therapies that we deliver to patients in emergent conditions.

    Patients and physicians should focus on all cause mortality.

    There are many other aspects of your argument that I would disagree with, but at the very least it is vital to get the effect size right.



  6. Heston LaMar

    I agree with the criticism of CRASH-2 that it could have likely provided more valuable information if it had not included just those pts in the “gray area” in terms of “need for TXA”. However, I also agree with others comments here that CRASH-2 has generated strong interest in an inexpensive relatively safe drug that so far seems to save lives especially in the sickest of trauma pts.

    My only true criticism of Dr. Walsh’s commentary is in regards to the MATTERs study and the statement “it showed that rates of PE and DVT among patients who received TXA were, respectively, 9 and 12 times the rates among those who did not”. This is true but that is a RELATIVE number (which can be misleading to some readers) and also has to be analyzed in context. So first, the absolute increase in DVT in the TXA MATTERs pts was 2.2 percentage points (6 patients) and for PE was 2.4 percentage points (also 6 patients) – definitely meaningful but not quite as impressive as “9-12 times”. Secondly and more importantly, the TXA group in MATTERs had 2 key risks factors for development of PE/DVT vs. the non-TXA group: 1) higher injury severity scores (ISS) and 2) a much higher percentage of extremity injuries with an AIS of >=3 (66.6% in TXA group vs. 47.3% in non-TXA group). So, I certainly cannot say that TXA does not contribute to thromboembolic events (don’t think anyone knows this for sure yet & its mechanism of action would logically suggest it would), but I think it is unfair to compare these 2 groups in MATTERs give the statistically significant difference (p<0.001) in both of these key risk factors for DVT/PE.

Leave a Comment