Every year, millions of people are treated for head injury. A substantial proportion die or are permanently disabled. The theory goes that aside from the direct damage at the time of the injury, the inflammation that follows leads to further problems. Fortunately, corticosteroids can reduce this inflammation so have been used to treat head injury for more than 30 years. This makes perfect sense, right?
The inflammation theory in head injury has been around for a while. The doctors who routinely gave steroids to their patients probably based this upon the theory that these drugs worked, maybe learning it in medical school. But, not all doctors used steroids, with surveys in the mid-90s showing that anything from 15% to 65% of trauma units gave steroids after major head injury. Given this uncertainty, which is not uncommon in medicine, randomised trials were conducted, the first ones in the 1970s, and here is what they showed:
Note: if you want an explanation of what this type of plot shows, it can be found here.
The long horizontal lines show the 95% confidence intervals for each trial. These tell us that we can be 95% certain, that is pretty certain, that the actual result lies somewhere on this line, but as you can see, the problem with these early trials was that the lines all crossed the value of one on the vertical scale. This means that these trials could not reliably tell us whether giving corticosteroids after head injury had any benefit or caused any harm. Even when the results of all of these trials were combined, using the statistical technique of meta-analysis, the overall result had a range that went from corticosteroids having a moderate benefit (15% less people died) through to a moderate harm (8% more people died). What often then happens is that this confusion about whether or not to treat gets compounded by all sorts of theories to do with nuances of treatment, such as a benefit would have been seen if the steroids were given sooner, or, steroids only work on this particular type of head injury. These theories often don’t stand up to proper scrutiny.
This confusion led Ian Roberts, who is a supporter of MoreTrials, and his colleagues at the London School of Hygiene and Tropical Medicine to establish a large simple randomised trial to compare corticosteroids after head injury to placebo. The MRC CRASH trial started recruitment in 1999 and was a massive undertaking involving 239 hospitals from 49 countries. The original goal was to recruit around 20,000 people, but the trial was stopped early, when half of this target was recruited, as the independent data monitoring committee, who regularly reviewed the unblinded results, concluded that the results seen so far proved beyond reasonable doubt that one treatment was better than the other (something all of the previous trials had failed to do).
Following the line of conventional thinking about inflammation outlined above, was the trial stopped early because the clear result that had emerged was that steroids significantly reduced death and disability after head injury?
No, the surprising result, but clear result, was exactly the opposite of this.
Steroids caused more harm than good with around 15% more deaths at 6 months in those randomised to steroids compared to placebo. It also clearly showed that this harm was seen in all of the different subgroups studied, so a lot of the different pet theories around at the time were all disproved.
As for the earlier article on the treatment of heart attacks, size matters. All of the previous trials before CRASH studied less than 2,500 patients in total, with most of them randomising less than 100 patients. CRASH, by contrast, randomised more than 10,000 patients, gave a crystal clear answer that steroids given after head injury caused more harm than good and, as a result, doctors stopped giving them and today, as a result, many more survive people survive after a major head injury.
The numbers above are worthy of a pause for reflection. What the CRASH showed is that in the group randomised to corticosteroids there were 1,248 deaths at 6 months from a total of 4,854 patients given the treatment. In the placebo group, by comparison, there were 1,075 deaths in a group of 4,819 patients. Remember, also the trial was double-blind so neither the patient or the medical team caring for them knew which group they were in.
This landmark trial also had other benefits. It created a worldwide network of trialists that more than 10 years on are still working together to reliably answer questions in emergency medicine.
So, yes, the MRC CRASH trial turned conventional wisdom on its head, but if that turning results in a major head injury at least you won’t now get a drug that might reduce your chances of surviving it. That’s what we mean when we highlight here randomised trials that matter.