OBJECTIVE We evaluated whether sample size differences between arms of two-arm parallel group randomized controlled trials (RCTs) published in American Psychological Association (APA)-affiliated journals were consistently smaller than expected by chance with simple randomization. METHOD We searched PsycINFO for two-arm parallel group RCTs in APA-affiliated journals published January 2007 to September 2017 that used individual randomization (1:1 allocation ratio), reported the number of participants randomized, and did not describe employing restrictive randomization (e.g., blocking). We queried authors because randomization processes were often not described in articles, and we conducted a post hoc logistic regression analysis to attempt to identify factors associated with overly balanced groups. RESULTS We identified 203 eligible trials, but after the author query, it was determined that only 115 used simple randomization. Among those 115 trials, there was a significantly greater number of trials with smaller sample size differences between trial arms than would be expected by chance (p < .001); 89 of 115 (77%) had differences in trial arm sample sizes smaller than the 50% prediction interval threshold for these differences. Greater proportionate imbalance may be associated with larger trial size (odds ratio of 0.27, 95% CI [0.08, 0.94] for N > 200 vs. N ≤ 100); greater balance may be more common in higher impact journals, though this was not statistically significant. CONCLUSIONS Education is needed to ensure that randomization procedures are implemented as intended and fully and accurately reported and that balanced group sample sizes are not understood as an indicator of trial quality. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

中文翻译：

---

在美国心理学会期刊上发表的随机对照试验中，小组规模的平衡。

目的我们评估了在美国心理学会（APA）附属期刊上发表的两臂平行组随机对照试验（RCT）的两臂之间的样本量差异是否始终小于简单随机化偶然发生的预期值。方法我们在2007年1月至2017年9月出版的APA附属期刊中使用PsycINFO进行了两臂平行组RCT的研究，该期刊使用了个体随机化（1：1分配比例），报告了随机参加的人数，并且没有描述采用限制性随机化（例如，阻止）。我们向作者询问是因为文章中通常没有描述随机化过程，并且我们进行了事后逻辑回归分析，以试图找出与过度均衡的群体相关的因素。结果我们确定了203个合格试验，但是在查询了作者之后，确定只有115个使用了简单随机化。在这115个试验中，试验组之间的样本数量差异较小的试验数量明显多于偶然的预期（p <.001）；115个样本中有89个（77％）的试验组样本大小差异小于这些差异的50％预测间隔阈值。更大比例的失衡可能与更大的试验规模有关（当N> 200与N≤100时，优势比为0.27、95％CI [0.08、0.94]）；在较高影响力的期刊中，更大的余额可能更常见，尽管这在统计上并不显着。结论需要进行教育，以确保按预期方式实施随机化程序，并完整，准确地报告，并且不将平衡的小组样本量理解为试验质量的指标。（PsycInfo数据库记录（c）2020 APA，保留所有权利）。