Episode 22, “The Time Machine”
Episode 22 of "In the Interim…" Dr. Scott Berry and Dr. Kert Viele examine the “statistical time machine,” a modeling approach popularized first in sports applications and applied with impact in platform trials such as I-SPY 2. By exploiting the bridging of arms in platform trials—where multiple treatment arms share control data across different periods—models can estimate the effect of “era” in a platform trial, providing better estimates of the relative treatment effects.
Key points:
● Overlapping enrollment periods allow statistical linking (bridging) between non-concurrent arms, much like methods used to compare athletes from different eras using overlapping careers.
● In I-SPY 2, this approach minimized unnecessary control enrollment and empirical evidence supported stable effects across a decade.
● The method relies on the assumption of treatment effect additivity; it can falter in settings with strong era-by-treatment interactions – which are rare.
● There is an industry continued skepticism about the use of non-concurrent controls, which is surprising given they may represent the single best case scenario for “historical” data use.
Listen to the full episode today: https://www.berryconsultants.com/resource/22-the-time-machine