Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, UK.
Huntington's Disease Centre, UCL Institute of Neurology, Department of Neurodegenerative Disease, University College London, London, UK.
Mov Disord. 2017 Nov;32(11):1610-1619. doi: 10.1002/mds.27122. Epub 2017 Sep 14.
The purpose of this study was to inform the design of randomized clinical trials in early-stage manifest Huntington's disease through analysis of longitudinal data from TRACK-Huntington's Disease (TRACK-HD), a multicenter observational study.
We compute sample sizes required for trials with candidate clinical, functional, and imaging outcomes, whose aims are to reduce rates of change. The calculations use a 2-stage approach: first using linear mixed models to estimate mean rates of change and components of variability from TRACK-HD data and second using these to predict sample sizes for a range of trial designs.
For each outcome, the primary drivers of the required sample size were the anticipated treatment effect and the duration of treatment. Extending durations from 1 to 2 years yielded large sample size reductions. Including interim visits and incorporating stratified randomization on predictors of outcome together with covariate adjustment gave more modest, but nontrivial, benefits. Caudate atrophy, expressed as a percentage of its baseline, was the outcome that gave smallest required sample sizes.
Here we consider potential required sample sizes for clinical trials estimated from naturalistic observation of longitudinal change. Choice among outcome measures for a trial must additionally consider their relevance to patients and the expected effect of the treatment under study. For all outcomes considered, our results provide compelling arguments for 2-year trials, and we also demonstrate the benefits of incorporating stratified randomization coupled with covariate adjustment, particularly for trials with caudate atrophy as the primary outcome. The benefits of enrichment are more debatable, with statistical benefits offset by potential recruitment difficulties and reduced generalizability. © 2017 International Parkinson and Movement Disorder Society.
本研究旨在通过分析多中心观察性研究 TRACK-Huntington 疾病（TRACK-HD）的纵向数据，为早期显性亨廷顿病的随机临床试验设计提供信息。
我们计算了具有候选临床、功能和影像学结局的试验所需的样本量，这些结局的目的是降低变化率。计算使用两阶段方法：首先使用线性混合模型从 TRACK-HD 数据中估计平均变化率和变异性成分，其次使用这些来预测一系列试验设计的样本量。
对于每个结局，所需样本量的主要驱动因素是预期的治疗效果和治疗持续时间。将持续时间从 1 年延长至 2 年可大大减少样本量。增加中期访问次数，并结合预测结局的分层随机化和协变量调整，可带来适度但并非微不足道的收益。以其基线的百分比表示的尾状核萎缩是所需样本量最小的结局。
在这里，我们考虑了从纵向变化的自然观察中估计的临床试验的潜在所需样本量。试验中选择的结局指标还必须考虑其与患者的相关性以及正在研究的治疗方法的预期效果。对于所有考虑的结局，我们的结果都为 2 年试验提供了强有力的论据，我们还证明了结合分层随机化和协变量调整的优势，特别是对于以尾状核萎缩为主要结局的试验。富集的好处更具争议性，统计上的好处被潜在的招募困难和降低的普遍性所抵消。© 2017 国际帕金森病和运动障碍协会。
