Adaptive randomization of response can also be used in both umbrella and platform study types. However, it should be noted that there are many debates about adaptive randomization to reaction that are independent of the various problems associated with master`s protocol studies. Previous reports have discussed the designs of platform studies, including adaptive randomization [, , ]. Depending on the purpose of the basket experiment, different multiplicity considerations can be applied. If the goal is to assess each disease population independently without borrowing data between disease populations, each disease population can be considered independent and has its own Type I error rate . Therefore, the same type I error rate (usually bilateral 0.05 in a Phase 3 study) can be used for each sub-study. If pooling rules do not depend on study data, adjustment of multiplicity is not necessary as long as the hypotheses tested after pooling apply to mutually exclusive populations. If, on the other hand, the pooling decision is based on the results of the interim analysis in the study, a multiplicity adjustment is required in the final analysis due to the error inflation caused by the interim analysis. The pruning and pooling mentioned in Supplementary Material 1 uses a Type I error control analytical formula in intermediate analysis and final analysis with estimates of the correlation of test statistics in intermediate and final analysis [34, 43]. For other methods, where Type I error control is not demonstrated analytically, or the Bayesian method, where Type I error is not part of the design features, detailed simulations with complete scenarios are required to assess what the false positive rate (Type I error) would look like.
There is another potential risk when publishing the performance of the common control arm: a future experimental arm added to the study of the platform could use the publicly available common control data to select a subgroup that performed worse than expected and suggest evaluating the new experimental compound in that subgroup. This behavior would affect performance, Type I error, and the interpretation of results for this future comparison. If a poor random set of results occurred in historical patients, inclusion in the analysis has the potential to more strongly separate the control arm from the experimental arm. However, the structure and ongoing nature of the platform experience can mitigate this to some extent. The platform study should include a set of general inclusion/exclusion criteria with limitations for additional inclusion/exclusion criteria for the efficacy of a particular investigational treatment. In this approach, additional inclusion/exclusion criteria should be limited to cases with clear and acceptable justifications, such as safety considerations, in order to ensure homogeneity of the population between several treatment groups. A Weapon Selection Committee (ASC), established for some platform trials, was used to regulate the inclusion and prioritization of new weapons and to transmit the new arm-specific experimental protocol annex with topics such as arm-specific inclusion/exclusion criteria. Secondly, if the study of the platform is open, additional data on the control arm is constantly generated. If historical results turn out to be poor, additional data is likely to bring the results back to the actual underlying rates. For a planned analysis that only includes simultaneous control, it would be very difficult to take advantage of a future randomized comparison if the performance of the controls is already known, since all future comparisons would have the protection of randomization. Where the planned analysis does not include simultaneous checks, predefined sensitivity analyses involving only simultaneous checks should be taken into account.
AcSé (Secure Access to Innovative Targeted Therapies) is an ongoing study of 23 sub-studies evaluating the efficacy of crizotinib as monotherapy in different types of tumors associated with ALK, MET, RON and ROS1 mutations . Menis et al.  reviewed the control protocol study on thoracic malignancies and classified the AcSé study as a basket study. Associated sub-studies were generally based on a two-stage design. The Adjuvant Lung Cancer Enrichment Marker Identification and Sequencing Trial (ALCHEMIST) study is a randomized umbrella study in ALK or EGFR positive patients with elevated lung adenocarcinoma . ALK or EGFR positive patients will be enrolled in a randomized phase III sub-study comparing crizotinib with placebo, or in a randomized sub-trial comparing erlotinib with placebo. The primary endpoint of each study is overall survival and an interim analysis is planned. In patients who are both ALK and EGFR negative, PD-L1 expression is measured and considered for recruitment in a randomised sub-study comparing the nivolumab administration group to an observation group. .