Maturation of effect size during enrollment of prospective randomized trials
Maturation of effect size during enrollment of prospective randomized trials
复制标题
DOI：
10.1016/j.jss.2017.06.082
复制
发表时间：
2018-03-01
影响因子：
2.2
通讯作者：
St Peter, Shawn D.
中科院分区：
文献类型：
--
作者：
Poola, Ashwini S.;Oyetunji, Tolulope A.;St Peter, Shawn D.
背景：随机临床试验是通过计算达到统计显著性所需的最小样本量来支持的，给定估计的效应量（ES）。ES是两个治疗组之间的原始差异。ES量化了队列间临床差异的实际程度，通常反映了试验的真实意义，而不考虑统计学意义。在固定的方案下，我们假设ES可以在比预先设计的更小的样本中获得。为了调查入组期间ES的模式，我们分析了在我们机构完成的已完成的试验。方法：回顾我院11项前瞻性随机临床试验的结果。ES在每次试验中每隔一段时间计算一次，以确定治疗组之间在哪一点实现了稳定的临床差异。结果：ES稳定在64%的中位入组率。在我们最小的研究中，所有患者都需要满足精确的ES，这表明需要在较小的研究中进行全入组。否则，我们50%的试验需要48%到76%的患者入组来满足ES。在比较临床结果时，12人中有9人发现了最终的差异，几乎与可以更早确定的差异相同。分类结果达到稳定ES的51%，连续结果达到68%。结论：我们的大多数研究在完成入组前就获得了ES和最终临床结果。这表明，通过随机化检测到的临床差异可能不一定需要建立统计显著性所需的强大样本量。这在同质人群的固定方案干预性试验中尤为重要。(C) 2017爱思唯尔公司版权所有。
Background: Randomized clinical trials are powered by calculating the minimum sample size required to achieve statistical significance, given an estimated effect size (ES). The ES is the raw difference between two treatment arms. ES quantifies the actual magnitude of clinical differences between cohorts and is usually reflective of the true meaning of the trial, regardless of statistical significance. Under a fixed protocol, we hypothesize that the ES may be attained at a smaller sample than predesigned. To investigate patterns of ES during enrollment, we analyzed completed trials that were completed at our institution.Methods: Outcomes of 11 prospective randomized clinical trials from our institution were reviewed. ES was calculated at intervals throughout each trial to determine at which point a steady clinical difference was achieved between treatment cohorts.Results: ES stabilized at a median of 64% enrollment. All patients were needed to meet the precise ES in our smallest study, indicating the need for full enrollment in smaller studies. Otherwise, 50% of our trials required between 48% and 76% of patient enrollment to meet ES. In comparing clinical outcomes, 9 of 12 found a final difference that was nearly identical to the difference that could have been determined much earlier. Categorical outcomes met stabilized ES at 51% enrollment and continuous outcomes at 68%.Conclusions: ES and final clinical outcomes were achieved before the completion of enrollment for most of our studies. This suggests that clinical differences detected by randomization may not necessarily require the robust sample size often needed to establish statistical significance. This is particularly relevant in fixed-protocol interventional trials of homogenous populations. (C) 2017 Elsevier Inc. All rights reserved.