观察性疗效比较分析中的混杂因素控制统计学方法研究

Study on good design can control the confounders from the beginning, such as randomized controlled trial, is obviously a good choice. But when RCT is not available, or when the analysis is required based on the real world data, increasingly importance has been attached to the application of observational comparative effectiveness research, but confounder is the common bias which is not easy to control..The existing confounding adjusted methods have many choices, but they are not perfect. The main issues are：The matching methods for measured confounders have lost data representativeness by re-selecting samples, there is no satisfactory method for controlling unmeasured confounders, how to evaluate the uncertainty of confounders, etc. .This study will establish a pre-post with comparison design concept, which can eliminate the influence of confounders by pre-post and concurrent comparison, will investigate the multiple instrumental variables method based on Mendelian randomization theory to overcome the weakness of single instrumental variable, will define confounder as fuzzy random variable and research sensitivity analysis method and strategy based on uncertainty theory, and will establish observational comparative effectiveness research statistical application guideline to serve clinical research practice. .This study will enrich and standardize the statistical analysis methods of real world data with crucial scientific significance and practical value.

控制混杂最理想的方法是在设计时充分考虑，如采用随机对照研究（RCT）。但在无法采用RCT时，或面对现实世界数据分析时，观察性疗效比较研究(CER)就成为重要补充，并越来越受到重视。然而在观察性CER中混杂因素不可避免。混杂因素控制的统计学分析方法尚不完善，主要表现在：针对已测量混杂因素的一些方法，如匹配法，因重新选择了分析样本，失去了原资料的代表性的优势；未测量混杂因素的控制尚无满意方法；如何衡量混杂因素的不确定性等。本研究拟建立有对照的前后比较设计，通过前后和平行比较消除混杂因素的影响；借助孟德尔随机化法，开展多工具变量的方法学研究，以克服单一工具变量的弱效性；另辟蹊径，将混杂因素看作模糊随机变量，基于不确定性理论对混杂因素的干扰进行敏感性分析和评价；建立观察性CER统计分析应用指南，服务于临床研究实际。本研究将丰富并规范现实世界数据的统计学分析方法，具有重要科学意义和应用价值。

观察性疗效比较研究应用广泛，但由于缺乏随机化技术，混杂因素控制尤为重要，然而混杂因素控制的统计分析方法并不完善。本研究按照混杂因素的不同类型，对混杂因素控制的统计学分析方法进行了比较和评价；聚焦备受关注的敏感性分析“E-value”法进行探讨，提出该方法的实际应用问题和注意事项；针对广泛应用的“倾向性评分法”及“孟德尔随机化法”进行了充分讨论，并形成了专家共识。本研究依托复旦大学附属中山医院国家主动脉专病队列对B型主动脉夹层覆膜支架腔内修复术手术病人特征及早期结局进行了评价，并进行了B型主动脉夹层早发基因突变关联性研究。