双臂等比例分配试验限制性随机化设计中个体分配的可预测性
本文基于治疗不平衡转移矩阵和条件分配概率,推导了以正确猜测概率和确定性概率衡量的个体治疗分配可预测性及随机化序列平均值。方法适用于满足双臂等比例分配、最大耐受不平衡限制、条件分配概率完全由当前观察到的治疗不平衡决定三项标准的限制性随机化设计。分析表明,双臂等比例分配试验中分配可预测性随治疗分配的奇偶序列顺序交替,且序列平均分配可预测性收敛至渐近值显著慢于个体分配可预测性,基于序列平均值的比较对序列长度敏感,使用序列平均分配可预测性可能低估个体分配的选择偏倚风险。
要点
- 双臂等比例分配试验中,个体分配可预测性随治疗分配的奇偶序列顺序交替。
- 序列平均分配可预测性收敛至渐近值显著慢于个体分配可预测性。
- 基于序列平均值的分配可预测性比较对序列长度敏感,可能低估个体分配的选择偏倚风险。
核心结论
- 双臂等比例分配试验中分配可预测性随治疗分配的奇偶序列顺序交替。
- 序列平均分配可预测性收敛至渐近值显著慢于个体分配可预测性。
- 基于序列平均值的比较对序列长度敏感,使用序列平均分配可预测性可能低估个体分配的选择偏倚风险,在短序列长度时尤为明显,个体分配可预测性可显著高于序列平均值。
研究设计与样本
- 基于治疗不平衡转移矩阵和条件分配概率推导个体治疗分配的可预测性(正确猜测概率和确定性概率)及随机化序列平均值。
- 方法适用于满足以下标准的限制性随机化设计:双臂等比例分配、最大耐受不平衡限制、条件分配概率完全由当前观察到的治疗不平衡决定。
主要发现
- 双臂等比例分配试验中,分配可预测性随治疗分配的奇偶序列顺序交替。
- 序列平均分配可预测性收敛至渐近值显著慢于个体分配可预测性。
- 基于序列平均值的分配可预测性比较对序列长度敏感。
- 使用序列平均分配可预测性可能低估个体分配的选择偏倚风险,在短序列长度时该差异尤为明显。
临床怎么用
- 原文未据此给出推荐。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
This manuscript derives the allocation predictability measured by the correct guess probability and the probability of being deterministic for individual treatment assignments, as well as the averages of a randomization sequence, based on the treatment imbalance transition matrix and the conditional allocation probability. The methods described are applicable to restricted randomization designs that satisfy the following criteria: (1) two-arm equal allocation, (2) restriction of maximum tolerated imbalance, and (3) conditional allocation probability fully determined by the observed current treatment imbalance. Analytical results indicate that, for two-arm equal allocation trials, allocation predictability alternates by the odd/even sequence order of the treatment assignment. Additionally, the sequence average allocation predictability converges to its asymptotic value significantly more slowly than the allocation predictability for individual assignment does. Consequently, comparisons of allocation predictability between different randomization designs based on sequence averages are sensitive to sequence length. Using sequence average allocation predictability may underestimate the risk of selection bias for individual assignment. This discrepancy is particularly pronounced for short sequence lengths, where individual assignment predictability can be substantially higher than the sequence average.
引用
- Zhao W, Livingston S. Allocation Predictability of Individual Assignments in Restricted Randomization Designs for Two-Arm Equal Allocation Trials. Stat Med. 2025. PMID: 39853771. doi:10.1002/sim.10343
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。