Keywords: Borrowing historical control,type I error, operating characteristics.
Room: Madison A
Borrowing historical control to current control in trial design and analysis is certainly appealing since the historical data can provide information on the control subjects so that more subjects and trial resources can be allocated to the active treatment arm to efficiently learn the safety and efficacy profiles of the novel treatment while retaining accurate estimates of the current control arm parameters. Similarly, borrowing historical adult data to pediatrics can be applied to facilitate trial design and analysis in rare disease setting. If historical information is sufficiently similar to the current control data, using historical control data can result in more accurate point estimates, power gain, and reducing type I error in clinical trials. If the similarity assumption is not satisfied, however, using historical control data can bias estimates and may inflate type I error.
This roundtable aims to provide an opportunity for statisticians in industry and regulatory agencies to share experience and ideas regarding how much historical data to borrow and the appropriate metrics to use for calibrating the design parameters to achieve desirable operating characteristics.