Artikel
Functional approaches for improving sample size recalculation in adaptive study designs
Suche in Medline nach
Autoren
-
Carolin Herrmann
- Charité – Universitätsmedizin Berlin, Berlin, Germany
-
Maximilian Pilz
- Institut für Medizinische Biometrie und Informatik, Universitätsklinikum Heidelberg, Heidelberg, Germany
-
Meinhard Kieser
- Universität Heidelberg, Heidelberg, Germany
-
Geraldine Rauch
- Charité Universitätsmedizin Berlin, Berlin, Germany
| Veröffentlicht: | 26. Februar 2021 |
|---|
Gliederung
Text
Background: Adaptive study designs allow adaptations during an ongoing clinical trial provided that the type I error rate is maintained. Trials can be stopped at an interim stage or in case the trial is continued, adaptations like a sample size modification are possible. Whereas sample size calculations for a clinical trial with a fixed design are often straightforward, there exists no recommended gold standard for adapting the sample size in adaptive clinical trials [1]. Hence, the responsible statistician has to decide on an appropriate sample size recalculation rule for the respective situation at hand. A common approach is to determine the sample size based on conditional power arguments [2], where the conditional power refers to the probability of correctly rejecting the null hypothesis under the knowledge of the observed interim effect. However, the observed interim effect size is most likely not the true effect size. Nevertheless, many sample size recalculation rules do not take the variability of this random variable into account.
Methods: Sample size recalculation rules can be considered as functions of the observed interim effect. We propose two approaches to improve sample size recalculation rules: a) We consider smoothing corrections for sample size increase; b) We examine a polynomial function modelling the sample size decrease with a parameter choice based on pre-defined performance criteria. We apply the location and variation of the conditional sample size as well as the location and variation of the conditional power as performance criteria. These approaches are evaluated and compared with established approaches by means of Monte Carlo simulations.
Results: The new solutions do not overcome the uncertainty of the interim effect size but they address the variability of the resulting estimates. For certain effect sizes, they increase the performance compared to standard sample size recalculation rules.
Conclusion: We present tools for improving sample size recalculation in adaptive study designs, which are easy to apply. The smoothing corrections may be combined with the polynomial or established sample size recalculation rules.
The research was funded by the German Research Foundation (DFG).
The authors declare that an ethics committee vote is not required.
References
- 1.
- Food and Drug Administration. Adaptive designs for clinical trials of drugs and biologics. Guidance for industry. 2019 [accessed March 4 2020]. Available from: https://www.fda.gov/media/78495/download
- 2.
- Lehmacher W, Wassmer G. Adaptive sample size calculations in group sequential trials. Biometrics. 1999;55:1286-1290.
