Adaptive Pathways: A Statistical View on the Decisions Involved in Conditional Licensing

Abstract

Adaptive Pathways (AP) has been proposed as a exible regularization approach for the Con- ditional Approval (CA) of drugs targeting an unmet medical need, especially rare diseases. AP allows for early approval based on a more limited amount of data, on the condition that addi- tional post-approval evidence is collected by the sponsor. However, CA is likely to be accompa- nied by increased uncertainty regarding a drugs bene?t-risks, leading to the inferential question of whether the initial licensing decision was correct. So far, common methodological implications regarding the decision criteria that govern a potential AP approach have not been discussed. This paper investigates three straightforward statistical methods that may serve as a basis for reevaluation of the initial licensing decision in a simple AP model, in the context of rare diseases: a trial-by-treatment interaction test, Conditional Power, and AP framed as an adaptive design are investigated. Results from a simulation study indicate that the interaction test is not a valid measure for identifying a false-positive conditionally approved drug. Conditional power applied to whom correspondence should be addressed. Adaptive Pathways 3 to a post-CA trial and the adaptive design approach are more promising, but can lead to to large post-CA sample sizes, which may not be feasible for drugs targeting rare diseases. Hence, these results suggest that reconsideration of the initial licensing decision through these simple methods is not straightforward. We suggest that further research into the methodology and underlying decision criteria is required, for AP to be a consistent and viable licensing approach.