All Times EDT
Keywords: External Data Leveraging and Borrowing, Real World Data and Evidence, Small Population, Propensity Score Method, Missing data Impact, Bayesian Borrowing
Nowadays, with the development of registries, EHR, post-marketing studies, more and more reliable, and relevant external data are available for medical research and development, particularly, in medical device area. These Real-World Data are utilized to serve as external control, augment the sample size in single arm trial etc., through various data leveraging methods such as Propensity Score (PS) Method. However, majority of such leveraging exercises are demonstrated with large or even abundant external data source. In reality, the size of available external data is often moderate. For example, in Orthopedic and Surgical medical device areas, usually, less than hundred subject-level external data might be available, due to the evolution of surgery/operation techniques and post-operation care. This would be also true for rare diseases and pediatric patient population. The limited size of external data source brings us different challenges and issues when we leverage them into clinical studies. For example, the unbalanced sample size issue between treatment groups. Plus, the model fitting issue and missing data issue in external data will be more noticeable and pronounced. Hence, in this roundtable, we will discuss these issues from both regulatory and industry perspectives. The discussion will start with Propensity Score (PS) method, the most used method for leveraging external data in medical device area, and then expand to other alternative approaches such as Bayesian Power prior.
Topics to be discussed: • Covariates selection and Modeling fitting issues such as Collinearity and identifiability • Unbalance Issues and Impact on the Estimated • Missing Data issue in clinically important covariates and primary outcomes • Alternative Methods to leverage moderate sized external data