Abstract
BACKGROUND: Many published accounts of clinical trials report no differences between the treatment arms, while being underpowered to find differences. This study determined how the authors of these reports interpreted their findings. STUDY DESIGN: We examined 54 reports of surgical trials chosen randomly from a database of 110 influential trials conducted in 2008. Seven that reported having adequate statistical power (β ≥ 0.9) were excluded from further analysis, as were the 32 that reported significant differences between the treatment arms. We examined the remaining 15 to see whether the authors interpreted their negative findings appropriately. Appropriate interpretations discussed the lack of power and/or called for larger studies. RESULTS: Three of the 7 trials that did not report an a priori power calculation offered inappropriate interpretations, as did 3 of the 8 trials that reported an a priori power < 0.90. However, we examined only a modest number of trial reports from 1 year. CONCLUSIONS: Negative findings in underpowered trials were often interpreted as showing the equivalence of the treatment arms with no discussion of the issue of being underpowered. This may lead clinicians to accept new treatments that have not been validated.