摘要
目的介绍一种新的Ⅱ/Ⅲ期无缝设计期中分析策略,应用短期结局预测概率进行剂量筛选,并对其相应统计学性质进行评价。方法以二分类数据为例,通过Monte Carlo模拟方法探讨短期结局有效剂量组同对照组率差变化、长短期结局间不同相关关系及给定不同先验信息对Ⅱ/Ⅲ期无缝设计I类错误、检验效能的影响。结果期中应用短期结局剂量筛选的II/III期无缝设计可以控制总的I类错误在检验水准以内,但是当相关关系很弱时结果偏保守。其检验效能一般低于长期结局剂量筛选结果,即期望检验效能,但是当短期结局有效剂量组同对照组率差增加,两指标相关关系较强或者短期结局有较强的先验信息时,应用短期结局剂量筛选可以大大提高检验效能。结论在Ⅱ/Ⅲ期无缝设计临床试验研究中,长期结局在短时间内无法获得,此时期中分析可以选用临床公认的、有文献支持的短期结局指导期中分析决策。
Objective We describe a new interim analysis strategy with treatment selection based on predictive power of the short-term endpoint in adaptive seamless phase Ⅱ/Ⅲ design and corresponding statistical properties are evaluated. Meth- ods Based on Monte Carlo simulations, we evaluate the influence of parameter variation on the type I error and power. Resuits The familywise typeIerror rate can be controlled in the strong sense, especially when the short-term endpoint is weekly correlated with the long-term endpoint. Furthermore, we cannot get a higher power by using the short-term endpoint for subgroup selection at interim analysis than by using the long-term endpoint. However, when the rate difference increases ,the short-term/ long-term endpoint correlation is not week or the prior information is strong ,the power of the new interim analysis will be greatly improved and close to the expected value. Conclusion As few final outcomes can be observed at the time of interim analysis, the short-term endpoint, which has been proved effective in papers and clinical practice, can be considered as the surrogate of the long-term endpoint to save development cost and to shorten R&D( research and development)cycle in the study of adaptive seamless Ⅱ/Ⅲ design.
出处
《中国卫生统计》
CSCD
北大核心
2015年第2期243-247,共5页
Chinese Journal of Health Statistics
基金
国家自然科学基金(81273184)
国家自然科学青年基金(81302512)