Key points are not available for this paper at this time.
BACKGROUND: We performed a review of acute stroke trials to determine features associated with premature termination of trial enrollment, defined by the authors as not meeting preplanned sample size. METHODS AND RESULTS: MEDLINE was searched for randomized clinical stroke trials published in 9 major clinical journals between 2013 and 2022. We included randomized clinical trials that were phase 2 or 3 with a preplanned sample size ≥100 and a time-to-treatment within 24 hours of onset for transient ischemic attack, ischemic stroke, or intracerebral hemorrhage. Data were abstracted on trial features including trial design, inclusion criteria, imaging, location and number of sites, masking, treatment complexity, control group (standard therapy, placebo), industry involvement, and preplanned stopping rules (futility and efficacy). Least absolute shrinkage and selection operator regression was used to select the most important factors associated with premature termination; then, a multivariable logistic regression was fit including only the least absolute shrinkage and selection operator selected variables. Of 1475 studies assessed, 98 trials met eligibility criteria. Forty-five (46%) trials were prematurely terminated, of which 27% were stopped for benefit/efficacy, 20% for lack of money/slow enrollment, 18% for futility, 16% for newly available evidence, 17% for other reasons, and 4% due to harm. Complex trials (adjusted odds ratio aOR, 2.76 95% CI, 1.13-7.49), presence of a futility rule (aOR, 4.43 95% CI, 1.62-17.91), and exclusion of prestroke dependency (none/slight disability only; aOR, 2.19 95% CI, 0.84-6.72 versus dependency allowed) were identified as the strongest predictors. CONCLUSIONS: Nearly half of acute stroke trials were terminated prematurely. Broadening inclusion criteria and simplifying trial design may decrease the likelihood of unplanned termination, whereas planned futility analyses may appropriately terminate trials early, saving money and resources.
Aziz et al. (Fri,) studied this question.