We performed a post hoc secondary analysis of the STAAMP prehospital clinical trial. The STAAMP trial was a pragmatic, phase 3, multicenter, double-blind, placebo-controlled, randomized clinical trial that examined outcomes in injured patients at risk for hemorrhage who received prehospital TXA during air medical or ground transport. The study included injured patients at risk for hemorrhage transported from the scene or transferred from an outside emergency department to a participating trauma center within 2 hours of injury. Patients were primarily randomized to receive TXA (1 g bolus over 10 minutes en route to hospital) versus placebo in the prehospital phase. Those in the treatment arm were subsequently randomized to 1 of 3 in-hospital phase TXA dosing regimens (no additional TXA, 1 g of TXA infused over 8 hours, or bolus of 1 g TXA followed by 1 g TXA infusion over 8 hours). The study was approved by the US Food and Drug Administration (Investigational New Drug 121102), the Human Research Protection Offices of the US Department of Defense, and institutional review boards of all participating sites as previously described. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.
Eligibility criteria for the current secondary analysis mirrored that of the STAAMP trial and included those patients with hypotension (systolic blood pressure ≤90 mm Hg) and/or tachycardia (heart rate ≥110 per minute) during the prehospital phase of care before trauma center arrival. Additional exclusion criteria for the current analysis included those patients with a qualifying vital sign shock index (heart rate/systolic blood pressure) ≤0.9 to minimize inclusion of those patients with negligible risk of hemorrhage. This was to focus the analysis on those patients with a greater risk of hemorrhage and shock where TXA may have a beneficial effect.
For the purposes of this secondary analysis, patients from the primary STAAMP trial were stratified into TXA (all 3 in-hospital dosing regimens) versus placebo and comparisons were limited to prehospital randomization arms.
The primary outcome for this secondary analysis was 30-day mortality. Secondary outcomes included 24-hour mortality, multiple organ failure (MOF) as defined by the Denver Multiple Organ Dysfunction scoring system, nosocomial infection (NI), acute lung injury (ALI), and 6- and 24-hour in-hospital blood component transfusion. We also analyzed presenting measures of coagulation [international normalized ratio (INR) and thromboelastography (TEG)], and measures of safety including pulmonary embolism (PE), deep venous thrombosis (DVT), and seizures. All analyses were performed using intention-to-treat randomized assignment.
Time of injury in the primary STAAMP trial was estimated using data from the corresponding Emergency Communications Center and Emergency Medical Service. Timing of prehospital TXA administration for the current secondary analysis was defined as EARLY (within 1 hour from estimated time of injury during the prehospital phase of care) and DELAYED (>1 hour from estimated time of injury during the prehospital phase of care).
We first examined whether the treatment effect of prehospital TXA versus placebo on 30-day mortality varied across EARLY and DELAYED subgroups. The interaction between prehospital TXA treatment and timing of administration was assessed for statistical significance, accounting for clustering by study site. We then evaluated 30-day mortality across treatment arms in both EARLY and DELAYED subgroups using chi-square test.
Kaplan-Meier survival analysis on 24-hour and 30-day mortality was performed to compare TXA treatment versus placebo patients across EARLY and DELAYED subgroups using log-rank testing. A multivariate regression analysis of survival using Cox proportional-hazard modeling clustered on study site with adjustment for confounders was performed to verify these unadjusted findings. The model was built on the primary outcome of 30-day mortality for EARLY patients. We selected covariates based on biological plausibility and strong univariate associations for the model. To prevent overfitting, we only included covariates with a Pvalue <0.1 and confounding covariates that changed the hazard ratio of TXA treatment by >10% in the final model. All covariates included in the final model passed linearity assumption and proportional-hazards assumption based on Schoenfeld residuals and Grambsch and Therneau's method.[23,24] The final model had adequate goodness-of-fit based on Gronnesby and Borgan test.
Secondary outcomes were tested using chi-square test for binary variables and non-parametric Wilcoxon rank-sum test for continuous variables. Patient demographics, injuries, and outcomes of interest were represented using descriptive statistics. Categorical variables were presented as frequencies (%) and were compared using chi-square or fisher exact test. Continuous variables were presented as means and standard deviations (SD) or medians and interquartile ranges (IQRs) and were compared using t test or Wilcoxon rank-sum test. Statistical significance was defined at a 2-sided P value <0.05. Stata SE 15.1 (College Station, TX) was used to perform all statistical analyses.
Annals of Surgery. 2021;274(3):419-426. © 2021 Lippincott Williams & Wilkins