Comparative Study of Bayesian Information Borrowing Methods in Oncology Clinical Trials

Abstract

PURPOSE With deeper insight into precision medicine, more innovative oncology trial designs have been proposed to contribute to the characteristics of novel antitumor drugs. Bayesian information borrowing is an indispensable part of these designs, which shows great advantages in improving the efficiency of clinical trials. Bayesian methods provide an effective framework when incorporating information. However, the key point lies in how to choose an appropriate method for complex oncology clinical trials. METHODS We divided the borrowing information scenarios into concurrent and nonconcurrent scenarios according to whether the data to be borrowed are observed at the same time as in the current trial or not. Then, we provided an overview of the methods in each scenario. Performance comparison of different methods is carried out with regard to the type I error and power. RESULTS As demonstrated by the simulation results in each borrowing scenario, the Bayesian hierarchical model and its extensions are more appropriate for concurrent borrowing. The simulation results demonstrate that the Bayesian hierarchical model shows great advantages when the arms are homogeneous. However, such a method should be adopted with caution when heterogeneity exists. We recommend the other methods, considering heterogeneity. Borrow information from informative priors is more suggested for nonconcurrent borrowing scenarios. Multisource exchangeability models are more suitable for multiple historical trials, while meta-analytic-predictive prior should be carefully applied. CONCLUSION Bayesian information borrowing is useful and can improve the efficiency of clinical trial designs. However, we should carefully choose an appropriate information borrowing method when facing a practical innovative oncology trial, as an appropriate method is essential to provide ideal design performance.

Enhancing generalizability and efficiency in clinical trials through dynamic information borrowing for both experimental and control arms: A simulation study

“…This figure reached 1.70 for the NPP approach. This result tied well with previous studies 7,19 that information borrowing was considered for control group only. When comparing the performance of dynamic informationborrowing approaches with the traditional approach of no borrowing in scenarios where heterogeneity existed between historical and current trials, it must be pointed out that the mean type I error rates for NPP, CPP and ELP exceeded 20%, while the mean power decreased to around 80%.…”

“…When comparing the performance of dynamic information‐borrowing approaches with the traditional approach of no borrowing in scenarios where heterogeneity existed between historical and current trials, it must be pointed out that the mean type I error rates for NPP, CPP and ELP exceeded 20%, while the mean power decreased to around 80%. The above methods underperformed the findings of studies where information borrowing was performed for the control group only 7,19,25,26 . A possible reason was that the definition of homogeneous datasets (e.g., datasets with response rates of 0.6 and 0.63) that should be borrowed was relatively close to the definition of heterogeneous datasets (e.g., datasets with response rates of 0.6 and 0.66) that should not be borrowed in our simulation, while the three dynamic information‐borrowing approaches based on Bayesian theory were insensitive to the relatively small differences 7 .…”

“…The above methods underperformed the findings of studies where information borrowing was performed for the control group only. 7,19,25,26 A possible reason was that the definition of homogeneous datasets (e.g., datasets with response rates of 0.6 and 0.63) that should be borrowed was relatively close to the definition of heterogeneous datasets (e.g., datasets with response rates of 0.6 and 0.66) that should not be borrowed in our simulation, while the three dynamic information-borrowing approaches based on Bayesian theory were insensitive to the relatively small differences. 7 For one reason, we designed the simulation parameters this way because head-to-head design or add-on design were often employed for trials with RWE in both groups, in which case [27][28][29] the differences between the experimental and control groups were relatively small.…”

“…Bayesian information borrowing is increasingly recognized as a key tool in clinical trials, particularly for bolstering study robustness through the integration of existing real-world evidence (RWE) in the face of insufficient data and limited sample size. 1,2 The current focus predominantly lies on external information borrowing for the control group, [3][4][5][6][7] but the expanding domain of real-world studies 8 points to the necessity of leveraging RWE for the experimental group. For instance, in-hospital compounded drugs 9 and traditional therapies 10 like traditional Chinese medicine (TCM) have an extensive experience in human use, offering substantial RWE that can inform both arms of a trial.…”

A frequentist approach to dynamic borrowing

“…The robust MAP prior approach constructs a mixture prior for modeling external controls from multiple sources. Despite their differences, the performance characteristics of aforementioned methods are similar in practice: the type I error is controlled in the absence of any systematic difference between the external and the internal controls; however, when the outcomes from external controls are worse than those from internal controls, these methods may have a substantially inflated type I error rate due to biased information borrowed from the external controls (Su et al., 2022; van Rosmalen et al., 2018). To improve the type I error control, Psioda and Ibrahim proposed a fixed‐borrowing adaptive design through a power prior approach.…”