Adaptive hybrid control design for comparative clinical trials with historical control data
提出一种自适应混合控制因果设计，利用历史对照数据减少随机对照试验所需样本量，通过多阶段调整随机化比例，在保留统计功效的同时节省样本。
Abstract We propose an adaptive hybrid control causal (AHCC) design to leverage historical control data to reduce the sample size demanded by standard randomised controlled trials (RCT). Under the causal inference framework, we define the causal estimand of the average treatment effect and derive the corresponding estimator based on the trial data and historical control data. The AHCC design takes a multistage or group sequential approach. The number of patients randomised to the concurrent control is adaptively adjusted based on the amount of information borrowed from the historical control data. At each stage, based on the interim data, the contribution of the historical control data, quantified by the effective sample size, is updated and used to determine the randomisation ratio between the treatment and control arms for the next stage, with the goal to resemble a standard RCT upon the completion of the trial. Simulation studies show that the AHCC design has desirable operating characteristics. For example, it saves on sample size when substantial information can be borrowed from the historical control, and it maintains power when little information can be borrowed from the historical control.