A systematic review and meta-analysis of sample size methodology for traumatic hemorrhage trials

Abstract

BACKGROUND Trauma hemorrhage remains the most common cause of preventable mortality in trauma. To guide clinical practice, RCTs provide high-quality evidence to inform clinical decision making. The clinical relevance and inferences made by RCTs are dependent on assumptions made during sample size calculation. METHODS To describe the quality of methodology for sample size determination, we conducted a systemic review RCTs evaluating interventions that aim to improve survival in adults with trauma-related hemorrhage. Estimated and actual outcome data are compared, including components of sample size determination. RESULTS A total of 13 RCTs were included. We noted a high rate of negative trial results (11 of 13 studies). Most studies were multi-center and conducted in North America, evaluating patients with blunt and penetrating injuries. The criteria for hemorrhagic shock varied across studies. All studies did not accurately estimate the mortality rate during sample size calculation. All but one study overestimated the mortality reduction during sample size calculation; the median absolute mortality reduction was 3%, compared with a target of 10%. Only the CRASH-2 study used a minimal clinically important different for treatment effect target. No RCTs employed prognostic enrichment. Most studies were terminated (8 of 13), mainly for futility. CONCLUSION Taken together, this review highlights that current clinical trial methodology is limited by imprecise control group risk estimates, overly optimistic treatment effect estimates, and lack of transparent justification for such targets. These limitations result in studies at high risk for futility and potentially premature abandonment of promising therapies. Given the high morbidity and mortality of trauma-related hemorrhage, we recommend that future conduct of trauma RCTs incorporate (1) prognostic enrichment to inform baseline risk, (2) justify target treatment differences based on clinical importance and realistic estimates of feasibility, and (3) be transparent and provide justification for the assumptions made. LEVEL OF EVIDENCE Systematic Review/Meta-Analysis; Level III.