The caption's POISE-3 numbers are exact. Its opening sentence, that TXA reduces bleeding "without promoting clot formation," is contradicted by the same trial's safety result two paragraphs later.
9,535 noncardiac surgery patients; 1 g IV at start and end of surgery vs placebo.
| Outcome at 30 days | TXA | Placebo | Effect |
|---|---|---|---|
| Composite bleeding (primary efficacy) | 9.1% | 11.7% | HR 0.76 (0.67–0.87); −2.6 points |
| Composite cardiovascular (primary safety) | 14.2% | 13.9% | HR 1.02 (0.92–1.14) |
To claim safety, the upper bound of the one-sided 97.5% confidence interval had to fall below 1.125. It came in at 1.14, so noninferiority was not shown (one-sided P = 0.04 against a 0.025 threshold). The safety composite included myocardial injury after noncardiac surgery, nonhemorrhagic stroke, peripheral arterial thrombosis and symptomatic proximal VTE. "Not established" is not the same as "harmful." The difference was 0.3 points, but it means the trial can't support a no-clot claim.
The landmark trials by indication.
| Trial | Population | Result | Verdict |
|---|---|---|---|
| CRASH-2 (2010, 2011) | 20,211 bleeding trauma patients, enrolled within 8 h | All-cause mortality 14.5% vs 16.0% (RR 0.91). Exploratory timing analysis: death due to bleeding fell with treatment ≤1 h (RR 0.68) and 1–3 h (RR 0.79), but rose with treatment after 3 h (RR 1.44) | Works early |
| WOMAN (2017) | 20,060 women with postpartum hemorrhage | Death from bleeding 1.5% vs 1.9% (RR 0.81, 0.65–1.00; P = 0.045), larger within 3 h of birth (RR 0.69). The original composite primary outcome (death or hysterectomy) was not reduced. Thromboembolic events did not differ | Works early |
| CRASH-3 (2019) | 12,737 adults with TBI; primary analysis in 9,202 treated within 3 h | Head-injury death 18.5% vs 19.8%, RR 0.94 (0.86–1.02), not significant. Mild-to-moderate TBI RR 0.78 (0.64–0.95); severe TBI RR 0.99. Vascular occlusive events and seizures similar. The authors' conclusion is more positive than the primary result | Selected patients |
| HALT-IT (2020) | 12,009 patients with significant GI bleeding; 1 g load then 3 g over 24 h | Death from bleeding 4% vs 4% (RR 0.99). Venous thromboembolism 0.8% vs 0.4% (RR 1.85, 1.15–2.98); arterial events similar | Doesn't work |
| POISE-3 (2022) | 9,535 noncardiac surgery patients | Less bleeding; cardiovascular safety not established | Works, safety open |
The pattern: TXA helps when started early in hemorrhage where fibrinolysis drives bleeding, and the benefit fades or reverses with delay (CRASH-2 after 3 hours) or in bleeding where fibrinolysis isn't the main problem (HALT-IT, with a high-dose 24-hour infusion). Note also that FDA labeling covers only short-term use in hemophilia patients around tooth extraction; every use in this table is off-label, supported by trial evidence. CV's whole-blood post covers how TXA fits into early trauma resuscitation.
Mechanism and data, side by side.
The mechanism is right: TXA is a lysine analogue that blocks plasminogen's lysine-binding sites, so plasmin can't dock on fibrin and break down clot that has already formed. It doesn't activate coagulation. That's the basis for "doesn't promote clot formation." The trouble is that a drug that stabilizes clots can still tip the balance toward thrombosis in some patients, and the outcome data haven't fully excluded that. A safer caption: TXA reduces surgical bleeding; whether it adds thrombotic risk is not settled, and in GI bleeding it did.
The caption names these considerations; here are the specifics from the trials.
| Setting | Regimen used in the trial |
|---|---|
| Trauma (CRASH-2) | 1 g IV over 10 minutes, then 1 g over 8 hours, within 3 hours of injury |
| PPH (WOMAN) | 1 g IV; second 1 g if bleeding continues after 30 minutes or restarts within 24 hours |
| Noncardiac surgery (POISE-3) | 1 g IV bolus at start and at end of surgery |
From the FDA label (Cyklokapron): infuse undiluted drug no faster than 1 mL/min (100 mg/min) to avoid hypotension, which is why the trial boluses run over 10 minutes; reduce the dose in renal impairment, using the label's creatinine-based table. Seizures are most often reported in cardiovascular surgery, at doses up to tenfold higher than labeled. For anesthesia, the label warning that matters most: inadvertent neuraxial injection has caused seizures, arrhythmias and deaths. The ampules can be confused with local anesthetics, so the label calls for clearly marking syringes with the IV route.