纵向测量整群随机试验中比较斜率的样本量测定
针对纵向测量整群随机试验(CRTLM),基于广义估计方程(GEE)在独立工作相关结构下推导出比较两组斜率的闭式样本量公式。该方法可纳入非平衡随机化、任意相关结构、不同缺失情景及整群大小变异,模拟显示经验功效和I类错误率接近名义值,并用真实临床试验示例说明其应用。
要点
- 比较斜率比仅比较均值或终点更能反映纵向趋势。
- GEE独立工作相关下闭式样本量公式灵活适配多种设计。
- 模拟显示经验功效与I类错误率接近名义值。
核心结论
- 基于GEE独立工作相关的闭式样本量公式可用于CRTLM中两组斜率比较,且模拟性能良好。
研究设计与样本
- 方法学研究:推导闭式样本量公式,并通过模拟研究和真实临床试验示例评估。
主要发现
- 模拟研究显示所提方法的经验功效和I类错误率接近名义值。
- 真实临床试验示例展示了该方法在实际实施中的实用性。
关键数值与推荐
- 基于GEE方法在独立工作相关结构下推导闭式样本量公式。
- 方法可纳入非平衡随机化、任意相关结构、通过观测概率和缺失模式描述的各种缺失数据情景以及整群大小变异。
临床怎么用
- 为纵向测量整群随机试验中比较两组斜率提供实用且稳健的样本量计算方案。
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
In cluster randomized trials with longitudinal measurements (CRTLMs), clusters of subjects, rather than individual subjects, are randomly assigned to either control or intervention groups. Measurements are collected from these subjects repeatedly at prespecified times until the end of the study. In clinical research, the focus is typically on investigating trends or progress over time to evaluate the effectiveness of a new treatment or track disease progression, rather than solely analyzing mean values or endpoint measurements. For comparing slopes between two groups, we have derived closed-form sample size formulas based on the generalized estimating equation (GEE) approach under independence working correlation. Our proposed method is highly flexible, allowing for the incorporation of unbalanced randomization, arbitrary correlation structures, various missing data scenarios through observational probabilities and missing patterns, and variability in cluster sizes. This flexibility provides a practical and robust sample size solution for CRTLMs. Simulation studies demonstrate that the proposed method performs well, maintaining empirical power and type I error rate close to their nominal values. Additionally, we illustrate the application of our method using a real clinical trial, showcasing its practical utility in real-world implementation.
引用
- Wang J, Zhang S, Ahn C, et al. Sample Size Determination for Comparing Slopes in Cluster Randomized Trials With Longitudinal Measurements. Stat Med. 2026. PMID: 42082166. doi:10.1002/sim.70576
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。