Our post on the CRASH-3 trial, an RCT examining the use of TXA in head injury, was arguably our most controversial of 2019 (1). Our view was that the evidence was not entirely definitive, but on the balance of probabilities TXA should be given to patients with moderate head injury, although it was probably of little use amongst those who had severe injury (1).
We are of course always open to the opportunity to change our views and therefore it’s good to see another RCT of TXA in head injury published in JAMA. The abstract is below but, as ever, we strongly recommend you read the full paper and accompanying editorial (2,3).
What kind of study is this?
It’s a three-arm (sort of) randomised controlled trial which is exactly the design type we want to see for a therapeutic trial.
Tell me about the patients
The study recruited patients across US and Canadian emergency medical services between 2015 and 2017. They only recruited adults with a head injury and with a GCS of 12 or less and with a systolic BP of 90mmHg or more. In CRASH 3 patients with mild head injury (GCS 13-14) were also recruited, but the focus in this paper is on the more severe end of the spectrum. TXA was given in the prehospital setting which is ideal as its purported mechanism is time critical.
There were three arms to the trial. In group 1 placebo doses were given. In group 2 the dose of TXA used was the same as in CRASH2 and CRASH3. 1g given as a bolus and then 1g over 8 hours, although the logic of this in head injury is something we questioned when reviewing CRASH2 (1). In group 3 the patients were given a 2g bolus of TXA and then a placebo infusion. In non-bleeding patients the utility of the infusion is in question and so it does make sense to test the 2g bolus. However, the initial plan to compare bolus against infusion was changed during the trial and both TXA groups were subsequently analysed together.
They recruited 1063 patients, but 96 in the TXA arm did not get the drug. 966 were therefore included. Of these, 819 were available for analysis for the primary outcome at 6 months. They did not follow up patients who were randomised but did not receive the drug. In this respect this means that this is not an intention to treat analysis which would have been preferable.
Most patients were male and the average GCS at enrolment was 8.
What about the outcomes
Whilst other trials have looked at mortality, this trial examined the neurological outcome at 6 months which we would agree is incredibly important. The extended Glasgow Outcome Scale(4) was used, which is a bit of a blunt tool, but is widely used in trials of this type (and in post cardiac arrest trials for example).
A favourable outcome was considered to be >4. At that level patients would be expected to be able to work in sheltered jobs, to have ongoing psychological issues and to have limited social/leisure activities. (5)
There were an additional 18 secondary outcomes (!).
Tell me about the main results.
As previously mentioned the main outcome was tested amongst 819 (85%) of the 1063 patients recruited. Interestingly the authors claim that those lost to follow up were likely to have better outcomes. 91 patients had their outcomes imputed as they were not able to be obtained directly. The imputation process is described in the supplementary material, and appears to be fiendishly complex. However, the bottom line is that primary data is always going to be better than imputed data, no matter how complex the imputation method.
The authors report their results as follows. Remember that their primary analysis combines the two groups who received TXA against placebo.
The site-adjusted absolute difference in the primary outcome of favorable neurologic outcome (GOSE score >4) between the combined tranexamic acid group and the placebo group was −3.5% (65% vs 62%; [90% 1-sided confidence limit for benefit, −0.9%]; P = .16; [97.5% 1-sided confidence limit for harm, 10.2%]; P = .84)
That’s not easy to follow, and I’m struggling to find what is meant by site adjusted analysis. I’ve looked in the statistical analysis plan available as a supplement and in the main paper, but I can’t see it so far. In terms of the raw data we can look at what I think are the main outcomes.
For a GOS of >4 at 6 months there are few differences.
- TXA bolus + placebo 153/312 (58%)
- TXA bolus 178/345 (62%)
- Placebo 163/309 (60%)
In terms of mortality at 28 days
- TXA bolus + placebo 55/312 (21%)
- TXA bolus 46/345 (16%)
- Placebo 54/309 (20%)
Amongst the numerous secondary outcomes the authors have found little difference between the treatment groups.
So what does this mean for TXA in head injury?
This trial concludes that there is no benefit to TXA in head injury, but I have doubts about the certainty of that conclusion, a number of which are acknowledged by the authors and reiterated online.
- There is a significant amount of missing data and it’s not really an intention to treat analysis.
- Although the statistical plan states that there may be an analysis by prehospital GCS score, I cannot find it here. Yet we know from CRASH3 that there is a highly likely association between initial GCS and the potential to benefit from TXA. Patients in CRASH3 with a GCS of 3 or fixed dilated pupils did not benefit from treatment. In this paper I don’t know whether that is a factor, and I would love to see a re-analysis based on this. I think that such a re-analysis would be justified based on the fact that the results of CRASH3 were not available when this trial was designed. My calculation from the paper is that 532 patients in this trial had a GCS of 8 or less. Pupillary responses were not given. If that’s the case then amongst patients in whom we might be expected to see a benefit from TXA based on CRASH3 results then there are only 287 patients in this analysis which makes it severely underpowered to detect a difference in that group.
- Survivor bias may be a factor here as only those who survive to 6 months can have an outcome recorded.
- Overall mortality is better with TXA, though admittedly this was not the primary outcome.
The bottom line.
The evidence for or against TXA in head injury remains uncertain. On the balance of probabilities my view remains unchanged from our previous CRASH3 conclusions.
- TXA has been shown to reduce bleeding related deaths in a variety of settings suggesting that it fundamentally improves outcomes for patients with life threatening bleeding.
- TXA should be given to all mild/moderate head injured patients with evidence of bleeding on CT scan.
- TXA should be given to all patients who have bilaterally reactive pupils and blood on CT scan.
- Patients with head injury PLUS extra-cranial injury TXA should be given to patients within 3 hours.
- Out-of-Hospital Tranexamic Acid for Traumatic Brain Injury https://jamanetwork.com/journals/jama/article-abstract/2770388
- Rowell SE, Meier EN, McKnight B, et al. Effect of out-of-hospital tranexamic acid vs placebo on 6-month functional neurologic outcomes in patients with moderate or severe traumatic brain injury. JAMA. Published September 8, 2020. doi:10.1001/jama.2020.8958
- The Extended Glasgow Outcome Scale (GOSE) http://www.gicu.sgul.ac.uk/resources-for-current-staff/rehabilitation-during-and-after-a-critical-illness/12_F_01_GOSE.pdf/view
- Trauma and Transfer https://criticalcarenorthampton.com/trauma-stuff/
- Understanding the Intention-to-treat Principle in Randomized Controlled Trials https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5654877/