The Gist: There's a good bit of chatter in the EM world focused on picking apart literature and trials to ensure that decisions and clinical treatment are up to date and evidence based. This is amazing and something many neglect as they merely peruse a paper's abstract. Might there be a time, however, when one should let some study design technicalities slide? I'm certainly not an expert on this subject, but a few recent debates on tranexamic acid (TXA) and epinephrine (epi) piqued my interest in this subject. Note: This was a behemoth post, so I've split it in two because my attention span is not nearly that long.
Recently, I heard an impromptu debate from a grand rounds lecture on the CRASH 2 study regarding TXA in trauma patients. The trial wasn't well received by the audience because of design issues. I recalled attempting to sell my surgery attending on the idea after I listened to the EMCrit podcast on TXA...he was totally unaware of the drug. This served as a impetus to review research/evidence problems specific to EM - an issue on my mind since the Japanese epi studies were published earlier this year, generating controversy regarding ROSC and neurological outcomes.
The paper: The importance of early treatment with tranexamic acid in bleeding trauma patients: an exploratory analysis of the CRASH-2 randomised controlled trial
What did this RCT show? Dr. Andy Neill of Emergency Medicine Ireland reviews and interprets the study here and the NNT review of TXA is also excellent. Basic rundown:
- Reduction in all cause mortality (RR 0.91, CI 0.85–0.97, p=0.0035) when tranexamic acid was given to trauma patients who were bleeding or at risk for bleeding (2)
- Reduction in the risk of bleeding death (RR 0.85, CI 0.76-0.96, p=0.0077) best benefit in patients who received it within 1 hour of injury (RR 0.68, p<0.0001) (2)
- Subgroup analysis planned a-priori
Ok, now onto some criticisms of the CRASH2 study:
- Size - Some argue that the large scale of the study (n=20211 patients in 40 countries at 274 hospitals) weakened the study results. Size seems to be one of the commonest reasons that a study is cited as weak. In EM, some studies have small numbers because of the infrequency of some presentations (toxic ingestion, difficult airways, etc). These studies are criticized for their lack of generalizability in other populations. CRASH-2, however, received some criticism because the argument that the size created heterogeneity in the population. Instead of believing that this might increase the generalizability across the globe, some believe that the size creates protocol standardization failure, meaning the results aren't actually applicable. Also, it is true that large studies can achieve power more easily than smaller studies, perhaps leading to detection of small clinical benefits.
- Effect - Absolute reduction in bleeding death observed in the treatment group is only 0.8%, so the number needed to treat (NNT) is 125 versus a NNT of 67 in those who received TXA earlier in their course. Thus, although statistically significant, the potential clinical benefit for an individual patient is small. Furthermore, some of the endpoints such as transfusions, were not decreased by TXA.
- Protocol - Patients were entered into the study based on physician's determination that they are "reasonably certain anti-fibrinolytic agents are indicated." This creates a subjective entry point rather than a quantitative, objective point.
- Time - The evidence strongly demonstrates that effect size depends on the time of administration. Harm may exist in administration of TXA after 3 hours. Due to transport time, some doubt their ability to meet this time frame.
So, then, why become a fan of TXA in the bleeding trauma patient?
- TXA is extremely cheap for an intervention that may save lives. The U.S. Department of Defense formulary cost is $39.12 per 10 mL vial containing 1 g of TXA ($80 for the regimen used in the trial). The cost at CVS Pharmacy is $101.99 per 10 mL vial containing 1 g of TXA ($204 for the regimen used in the trial) (1). The drug also demonstrated cost-effectiveness in country-based analysis of the CRASH-2.
- With regard to the size of the trial, this study would suggest that the the benefits of TXA have the potential to be generalized worldwide (as the cohort was rather international), a quality that many studies lack.
- In addition to the subjective entry point of a patient being at risk for significant hemorrhage, there was some objective criteria to guide clinicians. Ex: the patient had to have significant hemorrhage (systolic blood pressure less than 90 mmHg and/or heart rate more than 110 beats per minute), or meet the clinician's gestalt of being at risk of significant hemorrhage within 8 hours of the inciting event.
- The side effect profile was remarkably good, especially in the group that TXA is really indicated for (early post-event, within 3 hours). Compared with pharmaceuticals like Factor VII, which is markedly expensive, there is a paucity of thrombotic events associated with TXA. Also, blood product transfusions are not benign. Early utilization of this drug is likely far less expensive and harmful than massive tranfusions that often ensue.
- The patients that are likely to qualify for TXA are incredibly ill. The reduction in mortality by absolute numbers may be modest, but this cohort doesn't generally do well regardless. In an era when many other standard interventions have questionable efficacy (epinephrine in cardiac arrest, PPIs in UGIB, etc) rejection of the benefits of TXA seems in-congruent.
- The time frame issue seems work-able, especially if TXA were implemented in the pre-hospital setting.
- The findings of CRASH2 have been demonstrated outside of this large trial. The MATTERs study, a retrospective analysis of 293 patients who received TXA in the US and UK military combat setting in Afghanistan demonstrated a statistically significant mortality benefit in patients who received TXA compared with those who did not (17.4% vs 23.9%). The patients that received TXA were sicker than those who didn't (mean Injury Severity Score, 25.2 vs 22.5, P .001) (4).