Bariani Giovanni M, de Celis Ferrari Anezka C R, Precivale Maristela, Arai Roberto, Saad Everardo D, Riechelmann Rachel P
*Discipline of Radiology and Oncology, Instituto do Câncer do Estado de São Paulo, Universidade de São Paulo †PGS Statistics ‡Clinical Research Unit, Instituto do Câncer do Estado de São Paulo §Dendrix Research, São Paulo, Brazil.
Am J Clin Oncol. 2015 Dec;38(6):570-4. doi: 10.1097/01.coc.0000436085.23342.2d.
Sample size calculation (SSC) is a pivotal step in clinical trial conception and design. Herein, we describe the frequency with which oncology phase III trials report the parameters required for SSC.
We systematically searched for phase III trials published in 6 leading journals, which were accompanied by editorials from January 2008 to October 2011. Two blinded investigators extracted required and optional parameters for SSC according to the primary endpoint.
We retrieved 140 eligible phase III trials. The median target sample size was 596 subjects (50 to 40,000); in 66.4% of cases, the number of enrolled subjects was at least 90% of the target. The primary endpoint was a continuous variable in 5.7%, categorical in 30.0%, and a time-to-event variable in 64.3% of phase III trials. Although nearly 80% reported a target sample size, only 27.9% of the trials provided all the required parameters for proper SSC. The most commonly reported parameters for sample size computation were α (93.6%) and β (90.7%) errors. The parameters least reported were the expected outcomes in the control or experimental groups, each provided in only 57.9% of trials.
The quality of SSC reporting in phase III cancer trials is poor. Such incomplete reporting may compromise future study designs, pooling of data, and interpretation of results. Lack of transparency in SSC reporting may also have ethical implications.
样本量计算是临床试验构思与设计中的关键步骤。在此，我们描述肿瘤学III期试验报告样本量计算所需参数的频率。
我们系统检索了2008年1月至2011年10月在6种主要期刊上发表的III期试验及其社论。两名盲法研究者根据主要终点提取样本量计算所需和可选参数。
我们检索到140项符合条件的III期试验。目标样本量中位数为596名受试者（50至40000）；在66.4%的情况下，入组受试者数量至少为目标的90%。在III期试验中，5.7%的主要终点为连续变量，30.0%为分类变量，64.3%为事件发生时间变量。尽管近80%的试验报告了目标样本量，但只有27.9%的试验提供了样本量正确计算所需的所有参数。样本量计算最常报告的参数是α（93.6%）和β（90.7%）错误。报告最少的参数是对照组或试验组的预期结果，各仅在57.9%的试验中提供。
癌症III期试验中样本量计算报告的质量较差。这种不完整的报告可能会影响未来的研究设计、数据汇总和结果解释。样本量计算报告缺乏透明度也可能具有伦理意义。
