Walters Corbin, Meyer Chase, Fladie Ian, Wayant Cole, Vassar Matt
Department of Institutional Research, Oklahoma State University Center for Health Sciences, 1111 West 17th Street, Tulsa, OK, 74107, USA.
Indian J Gastroenterol. 2020 Feb;39(1):92-96. doi: 10.1007/s12664-019-01007-9. Epub 2020 Mar 4.
A multidisciplinary international cohort of 72 expert statisticians and researchers recently proposed lowering the p value threshold from 0.05 to 0.005 to mitigate distortion of trial results and decrease bias. We hereby explored how a change to the p value threshold may alter the statistical significance of primary endpoints in gastroenterology (GE) randomized control trials (RCTs). We analyzed RCTs published in the 20 highest ranked GE and medicine journals. For each trial, we extracted the p values for the corresponding primary endpoints. We retrieved 233 RCTs, of which 159 were included in the final analysis yielding 202 primary endpoints. Of these endpoints, 60% had a p value less than 0.05 and when a threshold of less than 0.005 was applied, approximately 50% retained significance. We endorse a lower p value threshold as an actionable, provisional measure for improving statistical inference in GE RCTs until more long-term solutions become available.
一个由72位专家统计学家和研究人员组成的多学科国际团队最近提议将p值阈值从0.05降至0.005，以减轻试验结果的扭曲并减少偏差。我们在此探讨了p值阈值的变化如何改变胃肠病学（GE）随机对照试验（RCT）中主要终点的统计学显著性。我们分析了发表在排名前20的GE和医学期刊上的RCT。对于每项试验，我们提取了相应主要终点的p值。我们检索到233项RCT，其中159项纳入最终分析，产生了202个主要终点。在这些终点中，60%的p值小于0.05，当应用小于0.005的阈值时，约50%仍具有显著性。我们支持较低的p值阈值，作为在GE RCT中改善统计推断的一项可操作的临时措施，直到有更多长期解决方案可用。
