Evidence-Based Practice Center, Mayo Clinic, 200, 1st street, Rochester, MN 55905, USA; Knowledge and Evaluation Research Unit, Mayo Clinic, 200, 1st street, Rochester, MN 55905, USA; Division of Preventive Medicine, Mayo Clinic, 200, 1st street, Rochester, MN 55905, USA.
Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario L8S 4L8, Canada.
J Clin Epidemiol. 2017 Feb;82:12-19. doi: 10.1016/j.jclinepi.2016.10.006. Epub 2016 Nov 8.
We explored how investigators of ongoing or planned trials respond to the publication of a trial stopped early for benefit addressing a similar question.
We searched multiple databases from the date of publication of the truncated trial through August, 2015. Independent reviewers selected trials and extracted data.
We identified 207 trials truncated for early benefit; of which 102 (49%) were followed by subsequent trials (262 subsequent trials, median 2 per truncated trial, range 1-13). Only 99 (38%) provided a rationale justifying conducting a trial despite prior stopping. The top reasons were to address different population or setting (33%), skepticism of truncated trials findings because of small sample size (12%), inconsistency with other evidence (11%), or increased risk of bias (7%). We did not identify significant associations between subsequent trials and characteristics of truncated ones (risk of bias, precision, funding, or rigor of stopping decision).
About half of the trials stopped early for benefit were followed by subsequent trials addressing a similar question. This suggests that future trialists may have been skeptic about the decision to stop prior trials. A more rigorous threshold for stopping early for benefit is needed.
我们探究了正在进行或计划中的试验的研究者如何应对因同样问题而提前停止的试验的发表。
我们从截短试验的发表日期到 2015 年 8 月在多个数据库中进行了搜索。独立评审员选择了试验并提取了数据。
我们确定了 207 项因早期获益而截短的试验；其中 102 项（49%）随后进行了后续试验（262 项后续试验，中位数为每个截短试验 2 项，范围 1-13 项）。只有 99 项（38%）提供了一个理由，说明尽管之前已经停止，但仍有必要进行试验。最主要的原因是为了针对不同的人群或环境（33%）、对小样本量的截短试验结果持怀疑态度（12%）、与其他证据不一致（11%）或增加偏倚风险（7%）。我们没有发现后续试验与截短试验的特征（偏倚风险、精度、资金或停止决策的严格性）之间存在显著关联。
约一半因获益而提前停止的试验随后进行了针对类似问题的后续试验。这表明未来的试验设计者可能对先前试验停止的决定持怀疑态度。需要更严格的提前获益停止阈值。
