Layman’s abstract for Statistics in Medicine article on Unit information prior for adaptive information borrowing from multiple historical datasets

Each week, we publish layman’s abstracts of new articles from our prestigious portfolio of journals in statistics. The aim is to highlight the latest research to a broader audience in an accessible format.
 
The article featured today is from Statistics in Medicine with the full article now available to read here.
 
Jin, HYin, GUnit information prior for adaptive information borrowing from multiple historical datasetsStatistics in Medicine20214025): 5657– 5672https://doi.org/10.1002/sim.9146
 
In the development of new drugs, clinical trials play a vital role in helping investigators to understand the characteristics of new treatments, including toxicity, side-effects, and effectiveness etc. However, as clinical trials typically involve human-being participants, ethical concerns often arise in practice. Moreover, with a large number of subjects followed for a long period of time, large-scale clinical trials are often very costly. As a result, improving the efficiency of clinical trials has been an extremely imperative problem. In reality, clinical trials are rarely conducted in isolation. It is common to find some historical clinical trials with similar settings to the current one. Thus, a possible way to improve the efficiency is to incorporate the historical data in the analysis of the current trial. 
 
The main difficulty of incorporating historical data relies on the potential differences between the current and historical trials, e.g., patient populations, types and stages of disease, treatments, dosages, endpoints etc. An ideal method should be able to incorporate the historical data from past studies adaptively while keeping them in a supplemental role to ensure the dominance and accuracy of the current trial data in the analysis. We propose a novel but simple method based on the unit information prior (UIP) to integrate the historical data into the analysis of the current trial in an adaptive manner. Our method is especially useful when multiple historical datasets exist and it has transparent interpretability in the sense that all the parameters in the UIP method have meaningful interpretations. In addition, the proposed method does not require the patient-level historical datasets, which is an important advantage compared with existing approaches, as the patient-level historical data are typically very difficult to obtain.  
More Details