{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"In the Interim...","title":"Digital Googols","html":"","width":"100%","height":180,"duration":2294,"description":"In this episode of \"In the Interim…\", Dr. Scott Berry examines the concept of “digital twins” in clinical trials. He details how simulation of clinical trials is a direct analog of digital twin methodology, allowing for the in-silico modeling of the physical trial conduct, enrollment, dropouts, and patient outcomes under varied assumptions. Scott discusses model-based patient prediction and highlights scenarios where prediction of counterfactual outcomes can increase efficiency, particularly in rare disease or limited-data settings. He provides a systematic comparison of Unlearn’s PROCOVA neural network approach with traditional covariate adjustment, noting that proprietary models must demonstrate clear improvement over standard methods, which is unlikely. There is great potential in the simulation of many digital twins for a patient as a potential augmentation or substitute for controls.\nKey Highlights\nDefines digital twins using NASA history and Wikipedia.\nDescribes clinical trial simulation as a digital twin methodology.\nExamines patient-level model-based prediction and covariate adjustment.\nCompares Unlearn’s PROCOVA with traditional approaches.\nHighlights transparency and reproducibility concerns with proprietary algorithms.\nAsserts that future trial efficiency demands integration of predictive modeling with randomization and large external datasets.\nFor more, visit: https://www.berryconsultants.com/","thumbnail_url":"https://img.transistorcdn.com/toFATB0JXvqQXRqzUqrK1EWCJCLOA_Qvjkb5ml3q-s8/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9jM2M0/ZDE2YzA1N2FhNjkx/NDk1NDczNjYzM2E5/NjlmYS5wbmc.webp","thumbnail_width":300,"thumbnail_height":300}