Абстрактный

Statistical inference of adaptive randomized clinical trials for personalized medicine

Feifang Hu, Yanqing Hu, Wei Ma, Lixin Zhang & Hongjian Zhu

To design a clinical study of personalized medicine, important covariates (biomarkers) and responses should be incorporated in patient selection. Adaptive designs are suitable for personalized medicine because of their nice properties: efficient; ethical; and incorporating covariates and responses. When covariates or responses are used in the randomization, does this affect the inference procedure? Here we summarize the properties of classical statistical inference in literature: for response-adaptive randomized clinical trials, the classical statistical methods are valid under widely satisfied conditions; for covariate-adaptive randomized clinical trials, the commonly used tests are usually too conservative, some adjustments are necessary; and for covariate-adjusted response adaptive randomized clinical trials, the classical statistical methods are valid under some restricted conditions, further research is needed to address the validness of classical statistical methods under general setting.