倾向指数用于中医药非随机设计临床疗效评价的思考

Propensity Score and the Intervention Effects of Chinese Medicine in Non- randomized Clinical Study

尽管随机对照试验是临床干预研究因果推断的金标准,但是在中医药临床研究的现实中,由于伦理、治疗措施的特征、费用、时间以及外部效度问题,随机对照试验往往无法实施甚至在某些情况下不适用。因果推论的本质是一个反事实问题。随机对照试验通过随机化来平衡可能影响干预分配的特征变量。倾向指数可以应用于非随机设计的观察性研究,如前瞻性队列研究,作为诸多特征变量的一个复合函数,用于均衡暴露组和对照组之间特征变量的分布。在满足一定前提条件的情况下,倾向指数分析法可以提供干预效果的无偏估计。因此,在中医药观察性的临床研究中,倾向指数可能具有广阔的应用价值和潜力。

Randomized controlled trial（RCT） is the golden standard design to make causal inference on intervention effects in clinical trials. While, in the research reality of Chinese medicine, RCTs could not be conducted, even not applicable, due to ethical issue, charac- ters of the intervention, cost, time and external validity. The essence of causal inference is counterfactual framework. All kinds of covariates could be balanced in RCTs via randomization. Propensity score（PS） is a composite score which could balance the distribution of eovariates between intervention and control group. PS could provide unbiased estimate of intervention effect when critical assumptions are met. For the observational studies of Chinese medicine, such as prospective cohort study, PS will have wide application and huge potential.