Abstract:
Modern oncology commonly requires sequential treatment; however, trials for approval typically do not include direct testing of sequences. Adaptive trials and Sequential Multiple Assignment Randomized Trial (SMART) trials are often of approved drugs and are large, slow, complex, expensive and may have high dimensionality making analysis complex. Herein we model SMART trials with streamlined designs, fewer patients and paths, to enhance registration (SMARTer). In the experimental arm an early during-treatment, optimizing diagnostic (ODx), such as ctDNA, is used to optimize each patient’s therapy, increasing the chance of benefit. We modeled six scenarios in which an ODx identified likely nonresponders early in the new therapy (NewRx) first arm and, within that arm, changed their treatment to standard of care (SOC) or, in a second design, re-randomized to SOC or SOC plus NewRx. Scenarios varied ODx performance and relative treatment effectiveness. When NewRx-alone response was modeled at 40%, versus 30% for SOC, and ODx had high sensitivity and specificity (95%), therapy change increased the NewRx-first arm response rate to 55.7% (40% increase) and reduced required sample size from 355 to 57 per arm (84% decrease). A 70% sensitive and specific test reduced size by 50%. Across scenarios, even when NewRx alone was no better than SOC, the SMARTer design yielded clinical benefits while reducing sample size. SMARTer is a streamlined, registration-oriented framework for evaluating new treatment paths. Modeling shows it would improve patient outcome, be more ethical, and reduce costs and time, while potentially providing data to speed regulatory approval.
