新药研发论坛.doc

确证性临床试验中数据缺失的处理指南 陈渊成(, 张 菁 (复旦大学附属华山医院抗生素研究所, 上海 200040) [摘要] 数据缺失是确证性临床试验研究中不可避免的问题，如何正确处理数据缺失是影响临床试验结论客观性的重要因素。临床试验中的数据缺失主要表现为病例脱落，根据其机制可分为完全随机缺失、随机缺失和非随机缺失。通过完善病例分析、谨慎选择缺失值处理方法(如填补法和混合效应模型法)和进行敏感性分析等尽量减少数据缺失所致的试验结论偏倚。欧洲医药产品管理局就临床试验中的数据缺失问题制订了指导原则，现已正式发布并实施。本文以该指导原则为核心阐述临床试验中数据缺失的主要处理方法，旨在为国内新药研发和临床研究提供有益参考。 [关键词] 数据缺失；确证性临床试验；指导原则；填补法；混合效应模型；敏感性分析 [中图分类号] R95 [文献标志码] C [文章编号]1003-3734 Guideline on missing data in confirmatory clinical trials CHEN Yuan-cheng, ZHANG Jing (Institute of Antibiotics, Huashan Hospital, Fudan University, Shanghai 200040, China) [Abstract] Missing data is an unavoidable problem in the research of confirmatory clinical trials, and the approach for handling missing data is an important factor data affecting the objectivity of the trial conclusions. The data missing in the clinical trial is mainly induced by drop-out, which could be classified as missing completely at random (MCAR), missing at random (MAR) and missing not at random (MNAR) according to the mechanism. The bias of trial conclusion could be reduced through the complete case analysis, careful selection of the method handling for missing data (e.g., imputation and mixed models) and the sensitivity analysis. To address this issue, European Medicines Evaluation Agency (EMEA) published a guideline on the missing data in the confirmatory clinical trial, and now it had come into effect. With focus on this guideline, this paper presents the main methods for handling missing in clinical trials with the aim providing information for the drug development and clinical research in China. [Key words] missing data; confirmatory clinical trial; guideline; imputation; mixed model; sensitivity analysis 确证性临床试验是在探索性临床试验基础上，通过开展多中心临床试验，进一步确证药物临床疗效，为获得上市许可获得充足证据。该阶段试验通常要求纳入较多的样本例数，设计随机双盲对照试验，并在此基础上进行数据统计以客观评价新药疗效。然而，受各种因素影响，在确证性临床试验中出现数据缺失是不可避免的[1]，例如: 患者拒绝继续研究、治疗失败/成功或出现不良事件导致退出试验、患者移居，以及观测指标的基线水平测量出现空缺等。数据缺失是导致临床试验统计分析结果出现偏倚的重要因素，当数据缺失比例较可观时，对临床试验结果的解释以及结论可信度会受到明显影响[2]。因此，如何减少数据缺失，以及在统计分析