Bayesian Design And Analysis Of Bridging Non-inferiority Clinical Trials

A bridging study is defined as a study performed in the new region to provide clinical data on efficacy and safety information in the new region that will allow extrapolation of the original clinical data to the population in the new region. Non-inferiority trial design is one type of clinical study designs. Its aim is to show that any difference between two treatments is small enough to allow a conclusion that the new drug has at least some effect or, in many cases, and effect that is not too much smaller than the active control. Non-inferiority design may be used in a bridging study due to ethical or other reasons.Currently two confidence interval method, especially 95%-95% confidence interval method is most commonly used in the analysis and design of non-inferiority trials. However, one of the concerns of such method is that this may lead to a very large sample size due to its conservativeness in many cases. When applying this method to the bridging non-inferiority trials, this may make a bridging study in a new region impractical to be conducted, and sometimes lead to unnecessary waste of resources and delay of drug availability in a new region. Bayesian method has its characteristics of borrowing strength from the historical data through priors and synthesizing historical data and observed data through posterior or statistical inference. In a bridging study, we can take the drug information for both active control and experimental drug from the original region (historical data) as prior and the information from the new region (observed data) to make statistical inference. Currently, there is no research about using Bayesian method to design and analyze bridging non-inferiority trials.Consequently, the objective of this study is to explore using Bayesian method to design and analyze bridging non-inferiority trials, which is achieved from the following four aspects.Firstly, this study explored operational characteristics of Bayesian method through simulation. When using informative priors from historical data, Bayesian method had greater power compared with two confidence interval method and synthesis method when the historical data supports the non-inferiority, and could control false positive rate (below 0.025) to certain extent. The results were impacted by the choice of priors.Secondly, this study developed the formula of the sample size determination using Bayesian method. The results showed that Bayesian method led to smaller sample size compared with two confidence interval method, while controlling the false positive rate and attaining the intended power when the historical data supports the non-inferiority.Thirdly, we introduced the mixture prior Bayesian model and deducted its posterior distribution. We studied the operational characteristics of this model and conducted sample size determination analysis. The results showed that mixture prior Bayesian model was able to reduce the impact of informative prior from the original region through an appropriate choice of weights, and could provide flexibility in the applicability of Bayesian method.Fourthly, we conducted a case study for bipolar disorder by using Bayesian method to design the trial and analyze the results. The results showed that Bayesian method has its applicability in designing and analyzing bridging non-inferiority trials.In conclusion, this study proposed a new approach for design and analysis of bridging non-inferiority trials, compared to the existing two-confidence interval method, this Bayesian method had the advantages of power increase and sample size reduction when the historical data supports the non-inferiority and this method had real applicability.