Department of Oncology and Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, Ontario, Canada.
Clin Trials. 2010 Jun;7(3):227-34. doi: 10.1177/1740774510368301. Epub 2010 Apr 20.
Dynamic allocation (DA) methods which attempt to balance baseline prognostic factors between treatment arms, can be used in multi-arm clinical trials to sequentially allocate patients to treatment. Although some experts express concern regarding the validity of inference from trials using DA, others believe DA methods produce more credible results.
A review of published multi-arm cancer clinical trials was conducted to explore the frequency of DA use in oncology.
Multi-arm phase III clinical trials of at least 100 patients per arm, published in 13 major oncology journals from 1995-2005 were manually reviewed. Information about reported use of DA methods, or randomization via random permuted blocks (PB), was extracted along with trial characteristics.
Of 476 published clinical trials, 112 (23.5%) reported using some form of DA method, while 103 (21.6%) reported using PB methods. Most trials (403 or 84.7%) reported stratifying on at least one baseline factor. The mean number of stratification factors was 2.70 per trial, and 78.6% of DA trials reported 3 or more stratification factors compared with 30.2% of non-DA trials (p < 0.001). The frequency of DA use increased over time, with 20.2%, 21.3%, 25.8%, 28.8% and 38.9% of trials reported use in 1995-2001, 2002, 2003, 2004, and 2005, respectively. Use of DA methods was more frequently reported in trials involving an academic co-operative group (28.4% vs. 13.8%), however, no difference was observed between industry-funded and other-funded trials (24.0% vs. 23.2%) or geographical region (19.7% of North American trials, 26.2% of European trials and 21.7% of multinational/other trials).
As a retrospective analysis, the true frequency of DA use is likely underreported. Few trials gave complete details of the allocation method used, thus it is possible some manuscripts reported incorrect allocation methods. Journals were selected which were assumed to publish most large, multi-arm clinical trials in cancer from 1995-2005, however, some trials were likely reported in journals other than what was reviewed.
DA methods are frequently used in multi-arm cancer clinical trials. The use of DA appears to becoming more common over time and are used more frequently when an academic cooperative group is involved. No relationship between industry funded trials or geographic region and allocation method was observed. Clinical Trials 2010; 7: 227-234. http://ctj.sagepub.com.
动态分配（DA）方法旨在平衡治疗组之间的基线预后因素，可以用于多臂临床试验中，以便按顺序为患者分配治疗方法。尽管一些专家对使用 DA 进行试验的推论的有效性表示担忧，但也有其他专家认为 DA 方法可产生更可信的结果。
对已发表的多臂癌症临床研究进行综述，以探究肿瘤学中 DA 使用的频率。
对 1995 年至 2005 年间在 13 种主要肿瘤学杂志上发表的，至少每臂 100 例患者的多臂 III 期临床试验进行了手工综述。提取了关于报告使用 DA 方法或通过随机置换块（PB）进行随机分组的信息，以及试验特征。
在 476 项已发表的临床试验中，112 项（23.5%）报告使用了某种形式的 DA 方法，103 项（21.6%）报告使用了 PB 方法。大多数试验（403 项或 84.7%）报告至少在一个基线因素上进行了分层。每个试验的分层因素平均为 2.70 个，78.6%的 DA 试验报告有 3 个或更多的分层因素，而非 DA 试验的比例为 30.2%（p < 0.001）。DA 的使用频率随时间增加，1995-2001 年、2002 年、2003 年、2004 年和 2005 年分别有 20.2%、21.3%、25.8%、28.8%和 38.9%的试验报告使用了 DA 方法。在涉及学术合作组的试验中，DA 方法的报告频率更高（28.4% vs. 13.8%），但是，在工业资助和其他资助的试验（24.0% vs. 23.2%）或地理区域（北美试验的 19.7%、欧洲试验的 26.2%和跨国/其他试验的 21.7%）之间没有差异。
作为回顾性分析，DA 使用的真实频率可能被低估。很少有试验提供了所用分配方法的完整详细信息，因此，可能有些文献报告了不正确的分配方法。选择了这些期刊，假设它们在 1995-2005 年期间发表了癌症的大多数大型多臂临床试验，但可能有一些试验发表在未被综述的期刊上。
DA 方法常用于多臂癌症临床试验中。DA 的使用似乎随着时间的推移而变得越来越普遍，当涉及到学术合作组时，使用的频率更高。未观察到工业资助的试验或地理区域与分配方法之间存在关系。临床试验 2010；7：227-234。http://ctj.sagepub.com。
