Outcome selection in clinical trials of pharmacologic haemostatic agents in surgery: systematic review
Yasaman Mohammadi Kamalabadi, Guangyong Zou, Atefeh Jafari, Adam G R Klotz, Pavel S RoshanovIntroduction
Perioperative bleeding is an important contributor to morbidity and mortality after surgery, and numerous randomised trials have evaluated pharmacologic interventions to control its occurrence. However, outcomes used in these trials vary widely, which may affect trial design, interpretation, and clinical applicability. We aimed to summarise primary and non-primary outcomes used in randomised controlled trials evaluating pharmacologic haemostatic interventions in surgery.
Research design and methods
This systematic review included trials published between 1 January 2015 and 26 December 2025 that enrolled at least 200 patients undergoing surgery and evaluated a pharmacologic haemostatic intervention. Descriptive analyses summarised the types and frequency of primary and non-primary outcomes, the distribution of sample sizes by primary outcome type, and risk of bias related to outcome selection.
Results
Of 19 032 records screened, 87 trials (61 666 participants) met inclusion criteria. Most trials (66/87, 76%) used surrogate primary outcomes, including blood loss volume (46/87, 53%), haemostatic function (13/87, 15%) and haemoglobin/haematocrit change (7/87, 8%). Only 21 trials (24%) used clinical or management outcomes as the primary endpoint, including transfusion-related outcomes (11/87, 13%), composite outcomes (3/87, 3%) and other clinical outcomes (7/87, 8%) such as venous thromboembolism. Trials with clinical primary outcomes were larger than those with surrogate outcomes (median 657 (IQR 353–1880) vs 230 (210–325)). Common non-primary outcomes included blood transfusion (56/87, 64%), adverse events or complications (50/87, 57%) and haemoglobin/haematocrit change (45/87, 52%), whereas use of standardised bleeding definitions (4/87, 5%) was uncommon. Selective reporting was the most frequent source of high risk of bias (30/87, 34%).
Conclusions
Most trials of pharmacologic haemostatic agents rely on surrogate primary outcomes, while clinically important outcomes are usually secondary. Greater use of patient-relevant endpoints may improve clinical applicability.