Abstract 17818: Are Our Clinical Trials Adequately Powered? an Examination of Randomized Controlled Trial Designs in Cardiovascular Medicine
Introduction: A central component of effective clinical trial design is the detailed calculation of sample size. Sample size estimates are often made based upon preliminary data or historical studies, and inaccuracy of anticipated event rates can lead to trials that are underpowered, particularly in an era of increasing cost-consciousness.
Methods: We sought to examine the prevalence of mis-estimation of control group event rates (eCER) in major randomized clinical trials (RCTs) published in major journals of cardiovascular medicine (CV) from 2008-2013. Included trials were those with a superiority design assessing the rate of a primary endpoint in a treatment vs. control group. Abstracted data included positive or negative results, a priori power calculation details, and observed control event rates (oCER). The proportion of eCER mis-estimation was calculated based on the oCER within the completed trials. Mis-estimation was defined as eCER not within 10% of the oCER. The ability of each RCT to detect modest (25%) and large (50%) relative differences between treatment vs. control with adequate power (≥80%) was calculated given the baseline event rate.
Results: Of 321 RCTs, 162 (50.4%) used dichotomous endpoints, with 132 (81.4%) utilizing a superiority design. Of these, 4 did not describe an a priori power calculation. Notably, 104 (78.8%) reported eCER. The baseline rate of eCER mis-estimation across studies was 37.5%; however, this did not differ by whether the trial was positive or negative (p=0.96). The number of negative trials was 80 (60.6%), and the actual mean observed relative effect size between study arms was only 7.4%. As a result, only 31% of RCTs were ≥80% powered to detect a 25% relative difference between treatment groups, while 64% had ≥80% power to detect a 50% relative difference between groups.
Conclusions: CV trials increasingly face unique challenges such as low event rates and small treatment effects. In this setting, accurate sample size estimates are central to feasibility and/or success of a trial. Given the degree of eCER mis-estimation observed in this analysis, novel methods of study design, such as adaptive trial design, should be more commonly considered.
Author Disclosures: R. Easterwood: None. A.J. Kirtane: None.
- © 2014 by American Heart Association, Inc.