Estimating Treatment Effects in Three-arm Non-inferiority Clinical Trials based on Compliance of Active Treatments
Objective Non-compliance of active treatments occurred in clinical trials is usually unavoidable,and improper use of standard approaches may lead to biased results,especially for non-inferiority trials.Thus,for three-arm non-inferiority clinical trials,we established a Bayes causal model to estimate causal effects in the presence of non-compliance.Methods Based on the framework of principal stratification,population was stratified according to types of compliance,and the issue was transformed into mixed-distribution identification.Bayes causal model was constructed and data augmentation(DA)algorithm were employed to calculate the posterior distribution of parameters of interest and complete statistical inference.Through simulation,we evaluated performances of our approach,compared with traditional methods including ITT,PP and AT.Results The method of ITT,PP,and AT all had a significant bias when the type of compliance associated with outcomes.The method proposed in this study both had a good performance whether the type of compliance associated with outcomes or not.Conclusion For non-inferiority trials with a high proportion of non-compliance,the method in this article has a better control of the bias.
NETL
NSTL
万方数据
