Artikel
Optimal designs for multi-arm phase II/III drug development programs
Suche in Medline nach
Autoren
Veröffentlicht: | 27. August 2018 |
---|
Gliederung
Text
Introduction: In drug development programs, phase II trials are often conducted as multi-arm studies with the aim of identifying the dose(s) with the best benefit-to-risk profile(s). Subsequent confirmatory phase III trials then attempt to demonstrate efficacy and safety of the selected dose(s). In the context of optimal planning of phase II/III with regard to sample size allocation and go/no-go decision rules (applied to decide whether to stop or to proceed to phase III), this leads to more complex go/no-go decision rules as compared to two-arm studies. For example, one has to decide whether to conduct the phase III trial with a single (the most promising, compare e.g. [1]) dose only, or with multiple doses (if sufficiently promising, compare e.g. [2]). The former strategy is less expensive as fewer patients are included in the trial. However, lower success probabilities are to be expected. Furthermore, it is well-known that selection of the most promising dose based on maximum treatment effect leads to overestimation of the effect which has to be taken into account. Thus, it is not clear which strategy results in a higher expected benefit. Therefore, it is worthwhile to examine different go/no-go decision rules and program strategies.
Methods: A utility function based on the concept of assurance [3], was recently proposed [4]. Here, optimal designs of phase II/III drug development programs in terms of sample size allocation and go/no-go decision rules are accomplished by considering fixed and variable per-patient costs of the program and prospective gains after a successful product placement. However, this framework is limited to drug development programs with two-arm trials. As dose selection is an important part of the go/no-go decision to proceed to phase III, we introduce a generalization of this method to drug development programs with multi-arm trials. Optimal sample size allocations and go/no-go decision rules are presented for time-to-event outcomes and scenarios, where at least one dose needs to show efficacy in the final stage of the drug development program.
Results: Generally, scenarios with larger presumed benefits and/or optimistic opinions about the true treatment effect lead to higher phase II investments in terms of sample size and to more liberal go/no-go decision rules for the optimal drug development design. These scenarios, therefore, favor strategy 1 ("most promising dose only"), as overestimation of the treatment effect is reduced in larger phase II trials with more liberal go/no-go decision rules, while the costs for phase III are lower compared to a multi-arm phase III trial.
Discussion: The proposed method takes into account costs of the program, expected benefit when launching the product successfully on the market, and development risk (success probability) to deliver insights on how optimal design parameters (e.g. sample size allocation, go/no-go decision rule regarding efficacy, number of arms in phase III) change with varying benefit, program costs and assumptions about the true treatment effect. In future work we will consider efficacy and safety of the doses as success criterion, leading to even more complex go/no-go decision rules and definitions of a successful program.
The authors declare that they have no competing interests.
The authors declare that an ethics committee vote is not required.
References
- 1.
- Stallard N, Todd S. Sequential designs for phase III clinical trials incorporating treatment selection. Statistics in medicine. 2003;22(5):689-703.
- 2.
- Magirr D, Jaki T, Whitehead J. A generalized Dunnett test for multi-arm multi-stage clinical studies with treatment selection. Biometrika. 2012;99(2):494-501.
- 3.
- O'Hagan A, Stevens JW, Campbell MJ. Assurance in clinical trial design. Pharmaceutical Statistics. 2005;4(3):187-201.
- 4.
- Kirchner M, Kieser M, Götte H, Schüler A. Utility-based optimization of phase II/III programs. Statistics in medicine. 2016;35(2):305-16.