CRASH-3 Results

The presentation of CRASH-3 is perplexingly more of the same. Although CRASH-2 investigated TXA’s use for major trauma and extracranial bleeding, CRASH-3 evaluated its use for intracranial hemorrhage. Originally designed to detect overall mortality within 28 days, the study was changed following publication of CRASH-2 to focus on head injury–related death, and then was expanded to ensure an adequate sample could be enrolled and treated within the crucial first three hours. However, like the WOMAN trial before it, neither the results for the original primary outcome nor the modified primary outcome reached statistical significance.

Again, however, the authors expend but a few words acknowledging their original primary outcome. Nor do they pour much mention upon the observed increase in non-head injury–related deaths associated with TXA treatment. Instead, as with the WOMAN trial, the authors dive into the subgroups for head injury–related death, further stratifying those results by Glasgow Coma Scale (CGS).

Within the 12,000 patients randomized, only 4,500 had GCS scores of 9–15 and were treated within three hours. In this subgroup alone, the authors report beneficial effects from TXA administration, with a reduction in head injury–related death from 7.5 percent to 5.8 percent. This subgroup alone forms the foundation of the authors’ discussion and the subsequent coverage in the popular press.

And what of adverse events? In all these studies, the authors confidently state no differences were observed between TXA and placebo. Although technically true, these trials were not specifically powered to detect differences in many infrequent outcomes. For example, in both the WOMAN and CRASH-2 trials, there were more deaths in the group of patients treated with TXA after the three-hour time window, indicating there are clearly some difficult-to-detect harms associated with the use of TXA. When only small benefits in a subgroup are touted as the positive outcome, even a handful of excess adverse events may be important signals.

CRASH-3 is not the only recent trial that assessed the utility of TXA for intracerebral hemorrhage. A smaller trial, Tranexamic Acid for Hyperacute Primary Intracerebral Hemorrhage (TICH-2), randomized 2,325 patients with spontaneous intracerebral hemorrhage to either TXA or placebo.⁶ This trial was unable to reliably detect a difference favoring TXA. However, as with WOMAN and CRASH-3, overall trends and secondary outcome measures generally favored the TXA cohort, albeit the true absolute difference is likely to remain vanishingly small.

So clearly the data do not fully jibe with the findings promoted in the lay press. Where does that leave us when making a bedside decision on TXA use? The overall effect of TXA administration in a subset of patients with head injury is most likely to be positive with respect to head injury–related mortality. This is not inconsequential. However, there are wide confidence intervals and uncertainty resulting from the overall survival observations, and the number of patients needed to treat to save one life may be hundreds of patients or more. Meanwhile, the points the various authors of these studies make are true: This treatment is relatively inexpensive, and the rate of serious adverse events is low enough as to be challenging to detect.

The bottom line, unsatisfyingly enough, is uncertainty. There is clearly room for individual practice variation with no obviously right answer. When considering the use of TXA for intracerebral hemorrhage, at the least it must be given as early as possible and within three hours, targeted at those with mild to moderate head injury, and only after the other higher-yield clinical interventions have been performed.

The opinions expressed herein are solely those of Dr. Radecki and do not necessarily reflect those of his employer or academic affiliates.