All Times EDT
Keywords: platform trials, type I error control, false positive decisions, trial adaptations
In platform trials, cohorts of patients treated with different drugs and subject to changing inclusion/exclusion criteria enter and leave the study in a sequential manner. At certain checkpoints, cohorts or treatments within a cohort are stopped for efficacy of the experimental drug or for futility.
As in every clinical trial, these decisions can be wrong, either as false positives or false negatives. Confirmatory trials typically focus on false positive claims of efficacy and require high hurdles to limit the probability of a false positive. In exploratory trials, the hurdles, in particular on false positive claims, can be lower because a subsequent confirmatory phase III trial will usually be required.
For platform trials to be acceptable as confirmatory investigations, the scientific community needs to come to a consensus about the required hurdles for limiting the probability of false positive claims of efficacy. In more traditional phase III trial designs (including seamless phase II/III designs), the prevailing paradigm used to be family-wise error rate (FWER) control. While purely technically, this concept might be expanded to many (but most likely not all) platform trial designs, the requirement seems unreasonably strict for platform trials in rare diseases with many subvariants or in under-researched small subpopulations of patients.
In the talk, we will review some ideas for relaxing the FWER requirement. Various such relaxations can be contemplated, e.g. "population-wise error rates", "intervention-wise error rate" or limits on the effect size probabilities of Bayesian posteriors. A comparison of these concepts will be performed on an example and by simulations which will cover a range of plausible situations.