Article
Comparison of the efficacy of Bayesian and frequentist designs for oncological phase II basket trials
Search Medline for
Authors
-
Maja Krajewska
- Institut für Biometrie und Klinische Epidemiologie, Charité Universitätsmedizin Berlin, Berlin, Germany
-
Geraldine Rauch
- Institut für Biometrie und Klinische Epidemiologie, Charité Universitätsmedizin Berlin, Berlin, Germany
| Published: | September 6, 2019 |
|---|
Outline
Text
Basket trials allow for the examination of a treatment in multiple patient subgroups and, in oncological settings, are based on the accrual of patients exhibiting tumors in different anatomic locations but the same genetic mutation. These patients are then allocated into subgroups based on the anatomical location of their tumor. Propositions for the evaluation of this type of trials have been mainly based on the concept of hierarchical Bayesian modeling [1], [2], [3], which allow for the exchange of information among patient subgroups exhibiting homogeneous treatment effects. However, frequentist approaches with the option to account of treatment homogeneity among subgroups have also been proposed [4], mainly based on the Simon two-stage design [5].
In this work, we aim to compare the efficiency of such frequentist designs for oncological phase II basket trials to a number of Bayesian designs.
We perform Monte Carlo simulations of a basket trial examining the response to a treatment in R [6] for both Bayesian and frequentist designs. We consider all possible scenarios of effect homogeneity and heterogeneity among patient subgroups while assuming the true underlying response rate to be either at an ineffective null response rate θ0 or at an effective alternative response rate θa.
We present newest simulation results and we discuss the advantages and disadvantages of each design. The efficiency of the designs will be examined by comparing them based on Specificity, Sensitivity, Type I error rates, Type II error rates as well as expected trial sample size. Additionally, a special focus will be set on the consequences of various prior choices for the discussed Bayesian designs.
The authors declare that they have no competing interests.
The authors declare that an ethics committee vote is not required.
References
- 1.
- Berry SM, Broglio KR, Groshen S, Berry DA. Bayesian hierarchical modeling of patient subpopulations: efficient designs of phase II oncology clinical trials. Clinical Trials. 2013 Oct;10(5):720-34.
- 2.
- Neuenschwander B, Wandel S, Roychoudhury S, Bailey S. Robust exchangeability designs for early phase clinical trials with multiple strata. Pharmaceutical statistics. 2016 Mar;15(2):123-34.
- 3.
- Simon R, Geyer S, Subramanian J, Roychowdhury S. The Bayesian basket design for genomic variant driven phase II trials. Seminars in Oncology. 2016: 43(1):1-6.
- 4.
- Cunanan KM, Iasonos A, Shen R, Begg CB, Gönen M. An efficient basket trial design. Statistics in medicine. 2017 May 10;36(10):1568-79.
- 5.
- Simon R. Optimal two-stage designs for phase II clinical trials. Controlled clinical trials. 1989 Mar 1;10(1):1-10.
- 6.
- R Foundation for Statistical Computing. R: A language and environment for statistical computing. 2017 [Accessed 16 July 2019]. Available from: https://www.R-project.org/
