Liberal or restrictive transfusion in brain injury part 2 (The TRAIN trial)

It is pretty rare for the phrase “we need more research” to be immediately followed by more research, but that seems to have happened here.v On October 7 I published my post about the HEMOTION trial, looking at transfusion thresholds in traumatic brain injury, and concluding that despite being a statistically negative trial, the results were promising and required follow-up research. On October the 9th, the TRAIN RCT was published, also looking at transfusion thresholds in patients with acute brain injury, so perhaps we will have a better answer about how to manage these patients.

The paper

Taccone FS, Rynkowski Bittencourt C, Møller K, Lormans P, Quintana-Díaz M, Caricato A, Cardoso Ferreira MA, Badenes R, Kurtz P, Søndergaard CB, Colpaert K, Petterson L, Quintard H, Cinotti R, Gouvêa Bogossian E, Righy C, Silva S, Roman-Pognuz E, Vandewaeter C, Lemke D, Huet O, Mahmoodpoor A, Blandino Ortiz A, van der Jagt M, Chabanne R, Videtta W, Bouzat P, Vincent JL; TRAIN Study Group. Restrictive vs Liberal Transfusion Strategy in Patients With Acute Brain Injury: The TRAIN Randomized Clinical Trial. JAMA. 2024 Oct 9. doi: 10.1001/jama.2024.20424. Epub ahead of print. PMID: 39382241 NCT02968654

The Methods

The TRAIN trial is a multicenter pragmatic open-label RCT from 72 ICUs across 22 countries.

Patients

Adults admitted to the ICU with traumatic brain injury, subarachnoid hemorrhage, or intracerebral hemorrhage, who were within 10 days of their injury, with a GCS of 13 or less, an expected ICU stay of at least 3 days, and a hemoglobin below 90 g/L (or 9 g/dL for Americans). 

Intervention

Restrictive transfusion, with a transfusion threshold of 70 g/L for 28 days or until hospital discharge.

Comparison

Liberal transfusion, with a transfusion threshold of 90 g/L for 28 days or until hospital discharge.

Outcome

The primary outcome was neurologic outcomes at 180 days, measured by the GOS-E scale.

Category number	Category name	Definition
8	Good recovery Upper	no current problems related to the brain injury that affect daily life
7	Good recovery Lower	minor problems that affect daily life; resumes >50% of the pre-injury level of social and leisure activities
6	Moderate disability Upper	reduced work capacity; resumes <50% of the pre-injury level of social and leisure activities
5	Moderate disability Lower	unable to work or only in sheltered workshop
4	Sever disability Upper	can be left alone > 8h during the day, but unable to travel and/or go shopping without assistance
3	Sever disability Lower	requires frequent help of someone to be around at home most of the time every day
2	Persistent vegetative state	unresponsive and speechless
1	Death	death

The Results

They enrolled 850 patients, a median of 3 days after admission to the ICU. The mean age was about 50 with a relatively even split between males and females. The population consisted of about 60% traumatic brain injury and about 20% each of SAH and ICH. Mean hemoglobin on hospital admission was about 118 and at time of enrollment it was 85 (meaning these patients weren’t just anemic, but had had a significant acute hemoglobin drop).

The study protocol was mostly followed, with statistically and seemingly clinically different hemoglobins throughout the trial. The median number of transfusions was 2 in the liberal group and 0 in the conservative group. 

For their primary outcome, 63% of the liberal group and 73% of the conservative group had unfavorable neurologic outcomes (ARR 10%, 95% CI 4-17%). 

Mortality was about the same (20.7% vs 22.5%). There were fewer cerebral ischemic events with liberal transfusion, but the other secondary outcomes were essentially unchanged. 

My thoughts

The population is somewhat different from the HEMOTION trial, and this was entirely an ICU trial with enrollment multiple days into their stay, so you could argue that the results aren’t that important to emergency medicine. However, given the questions that were left after the HEMOTION trial, I think we can probably learn something from this trial. Neither TRAIN nor HEMOTION are perfect trials – I think replication is still necessary – but these are good results and should push people who were on the fence about the HEMOTION results (like me) towards more liberal transfusion. 

Of course, by itself, this trial is far from perfect. It is an unblinded trial, which is a big deal when your primary outcome is highly subjective (neurologic outcomes on the GOSE scale). That being said, unlike some new ‘miracle drug’, I can’t imagine patients or families are all that influenced by whether or not they received a transfusion 180 days later. Conversely, a global sense that the medical team was either good or bad can be influenced by subtle factors (such as ignoring low lab values), and that could easily influence subjective interpretations of long term outcomes.

I talk a lot about the ridiculous sample size calculations that many trials use. I was so pleased to see that this trial used a sample size calculation that was looking for a 5% absolute improvement in neurologic outcomes, because that is semi-realistic. However, the very next sentence is that they decided to abandon this calculation because of slow enrollment, and so the sample is not science based but rather convenience based. Even more concerning, they adjust it twice. First, they doubled the target, to a target 10% absolute improvement, which isn’t ideal but at least could be fair. They subsequently changed it to a target 11% improvement, which clearly is not a number that anyone would choose a priori, and so was obviously cherry picked to suit their data. They ran into some problems because of COVID, but these sample size changes occurred in the final months of the trial, after almost all patients had been enrolled, and after an interim analysis had been completed, which significantly increases researcher degrees of freedom and increases the risk of p-hacking. 

The questionable sample size and unblinded nature of this trial would be enough for me to ignore the results if this trial were asking for a big practice change. However, despite lots of evidence that conservative transfusion strategies are beneficial, more liberal transfusion in patients in brain injuries might be considered current standard practice, and so I think you can look at the combination of HEMOTION and TRAIN and conclude that we should continue with liberal transfusion for now, until we get higher quality data in the future. 

Bottom line

This unblinded RCT demonstrated better neurologic outcomes when anemic patients were more liberally transfused, and although it is not a perfect trial, it will influence my practice, convincing me to be more liberal providing blood transfusions in this population. 

Other FOAMed

Liberal or restrictive transfusion in brain injury (The HEMOTION trial)

Evidence based medicine is easy

The EBM bibliography

Evidence based medicine resources

EBM deep dives

References

Taccone FS, Rynkowski Bittencourt C, Møller K, Lormans P, Quintana-Díaz M, Caricato A, Cardoso Ferreira MA, Badenes R, Kurtz P, Søndergaard CB, Colpaert K, Petterson L, Quintard H, Cinotti R, Gouvêa Bogossian E, Righy C, Silva S, Roman-Pognuz E, Vandewaeter C, Lemke D, Huet O, Mahmoodpoor A, Blandino Ortiz A, van der Jagt M, Chabanne R, Videtta W, Bouzat P, Vincent JL; TRAIN Study Group. Restrictive vs Liberal Transfusion Strategy in Patients With Acute Brain Injury: The TRAIN Randomized Clinical Trial. JAMA. 2024 Oct 9. doi: 10.1001/jama.2024.20424. Epub ahead of print. PMID: 39382241

Turgeon AF, Fergusson DA, Clayton L, et al. Liberal or Restrictive Transfusion Strategy in Patients with Traumatic Brain Injury. N Engl J Med. 2024 Jun 13. doi: 10.1056/NEJMoa2404360. Epub ahead of print. PMID: 38869931