The need for evidence we can trust - large international trial finds drug used to treat gastric bleeding does not work

Lead author of the HALT-IT trial Ian Roberts on why he believes the trial findings reinforce why the knowledge system for healthcare is not fit for purpose

For my 40 birthday I got a new brother. Not a scrawny, screaming, baby brother but a corpulent, competent, pillar of the community big brother. An old friend, having drunk his fill, found himself unable to keep his secret for a single second longer and spewed it out at my party.

My parents, teenaged, pregnant and unmarried, had been forced to give him away before I was born. But we found him. We cherished our new brother for five full years before a malign finger of flesh in his stomach rose up to snatch him away for good. He died suddenly and unexpectedly from a gastrointestinal bleed.

Having recently completed a large multicentre randomised trial of tranexamic acid in trauma patients (the CRASH-2 trial) which found that this inexpensive, generic drug cuts deaths by a sixth, I wondered whether tranexamic acid could have extended my brother’s life. Being an evidence enthusiast, I consulted the Cochrane review which showed an astonishing 40% reduction in gastric bleeding deaths.

Once upon a time, I would have found this ‘evidence’ compelling but my love affair with Cochrane ended when I wised up about cheating. Having proudly published a review showing that high dose mannitol reduces head injury deaths, a Colombian colleague questioned whether any of the included trials had actually taken place.

I investigated but could not authenticate any of them. I had taken the published research on trust. When it happened again, this time thanks to the renowned fraudster Joachim Boldt I became bitter. Single centre trials cannot be trusted even when trussed up by Cochrane into a silk purse. No one checks the actual data. Universities don’t and journals don’t. It is easier and cheaper to trust - unless you are a patient when this negligence can cost you dearly. 

To know whether tranexamic acid cuts gastric bleeding deaths - we needed a large trial and with colleagues at the London School of Hygiene & Tropical Medicine clinical trials unit we began building a global team to provide it. With trials you can either play a big part in something trivial or a small part in something important. Large trials are a collaboration with a cast of thousands. But first we needed funding.  We are blessed in Britain with public funding for research that benefits patients (as opposed to shareholders). But even the methodologically sophisticated National Institute for Health Research initially questioned the need for a trial given the Cochrane review.

Even when funded and recruited, the Cochrane review continued to harass. Many UK doctors would recruit bleeding patients, randomly allocate them to tranexamic acid or placebo, but if they continued bleeding, would give tranexamic acid regardless.  We closed trial recruitment at many UK hospitals due to lack of clinical equipoise and without the strong trial enrolment in non-UK hospitals, particularly Pakistan, we would have failed to reach our planned sample size.

Eight years and 12,000 patients later (recruited from 164 hospitals in 15 countries) we have results. Tranexamic acid did not cut bleeding deaths but increased the risk of venous thrombosis and seizures. The discrepancy between the review of small trials and the large multicentre trial is alarming but not unusual. It should prompt a reconsideration of the role of systematic reviews in informing UK health care but the Department of Health is still in love with Cochrane.

The HALT-IT trial is for my dead brother and for the millions of dead brothers and sisters world-wide. It will be published by the Lancet without fanfare because the media is justifiably engrossed with COVID-19. But there are lessons here even for a viral pandemic.

We urgently need effective treatments for COVID but standards of medical research and medical publishing have plummeted. Much of what has been published on treatments is unreliable observational research and some of it is outright fiction. With a few notable exceptions (e.g. the UK Recovery trial), the world has failed to provide large-scale randomised evidence fast enough and this has caused thousands of avoidable deaths. The knowledge system for health care is not fit for purpose.


The HALT-IT Trial Collaborators. Effects of a high-dose 24-h infusion of tranexamic acid on death and thromboembolic events in patients with acute gastrointestinal bleeding (HALT-IT): an international randomised, double-blind, placebo-controlled trial. The Lancet. DOI: 10.1016/S0140-6736(20)30848-5

Short Courses

LSHTM's short courses provide opportunities to study specialised topics across a broad range of public and global health fields. From AMR to vaccines, travel medicine to clinical trials, and modelling to malaria, refresh your skills and join one of our short courses today.