Presentation (CTI)

OBJECTIVES: The U.S. Right-to-Try Act allows patients with life-threatening conditions to access investigational medicines outside clinical trials. While this accelerates access, treatment choices are made independently by physicians and patients, often without coordinated evidence generation. Consequently, real-world data accumulate slowly and unevenly, limiting the ability of regulators, payers, and manufacturers to assess comparative effectiveness. This study proposes a framework to accelerate learning about treatment efficacy from real-world use while preserving patient-first decision making. We examine two use cases: (1) identifying the most effective treatment and (2) determining whether any treatment exceeds a predefined efficacy benchmark.
METHODS: We model this setting as a distributed decision environment in which many physicians repeatedly choose among several treatment options, each with uncertain effectiveness, based on limited information and the best interest of the current patient (formalized using a multi-armed bandit framework with myopic agents). In this context, physicians are not incentivized to experiment for future learning. A manufacturer periodically receives aggregated outcome data across physicians. We propose an honest information-sharing policy: upon identifying an inferior treatment with sufficient statistical confidence, this information is transparently communicated to physicians, thereby shifting utilization and enabling further evidence generation.
RESULTS: We derive the optimal information-sharing policy and quantify the number of patients and physicians required to achieve statistically significant conclusions. For example, with manufacturer communication, achieving a 90% significance guarantee requires approximately 560 patients per treatment, contributed by at least 40 physicians. In contrast, without manufacturer communication, the number of required patients and physicians increases prohibitively—by a factor proportional to the number of treatments.
CONCLUSIONS: Evidence-triggered, transparent information sharing can substantially accelerate real-world efficacy learning without requiring randomization or deviation from patient-centered care. This approach provides HEOR and market access teams with a practical strategy for improving the speed, efficiency, and reliability of evidence generation in Right-to-Try and other early-access settings.