Start Your Search
PC01 - Clinical Trial Design for Drug Development (ID 70)
- Event: WCLC 2013
- Type: Pro/Con Session
- Track: Statistics
- Presentations: 1
- Moderators:N. Yamaguchi, M. Stockler
- Coordinates: 10/28/2013, 14:00 - 15:30, Parkside Ballroom B, Level 1
PC01.3 - Adaptive Designs: Against (ID 626)
14:00 - 15:30 | Author(s): M. Buyse
Adaptive clinical trials designs are sometimes claimed to have many benefits over conventional group sequential designs. Such claims are generally unfounded, though adaptive designs may be quite useful in specific situations. Adaptive designs can be quite useful in early phase dose finding trials, where the purpose is to minimize the number of patients exposed to inefficacious or overly toxic doses. Adaptive designs can also be useful in later phase trials when a biomarker (or baseline patient characteristic) is suspected to be strongly predictive of the effect of the treatment under investigation; in this case, an adaptive enrichment of the trial can greatly improve its power. Adaptive designs can also prove useful when the assumptions underlying a trial design are subsequently found to be grossly inadequate, but this is an undesirable as well as uncommon situation. Adaptive randomization, which consists of allocating more patients to the treatment that appears to have more efficacy ("play-the-winner"), is justified neither statistically nor ethically. This strategy may produce slight reductions in the number of patients exposed to the inferior treatment, but it may increase the total sample size of the trial as compared to using a fixed allocation ratio (e.g. 1:1). More importantly, this adaptive strategy conveys the misealding impression that one treatment is known to be better than the other, a situation in which equipoise is not maintained and randomization no longer ethical.
Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.