Strategy for randomised clinical trials in rare cancers

Abstract

Proving that a new treatment is more effective than current treatment can be difficult for rare conditions. Data from small randomised trials could, however, be made more robust by taking other related research into account

Figure 1

Prior and posterior distributions and likelihood “data” for enthusiastic, neutral, and sceptical scenarios. The peak of each distribution corresponds to the most likely value of the true log hazard ratio, as estimated from the prior studies or the “data” or both (see text). The prior distribution in this example assumes that the two treatments are identical (log hazard ratio=0)