Article
Adaptive designs for diagnostic studies
Search Medline for
Authors
-
Antonia Zapf
- Institute of Medical Biometry and Epidemiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
| Published: | August 19, 2022 |
|---|
Outline
Text
Background: The aim of diagnostic studies is to investigate how reliably a diagnostic test can detect a condition, e.g. the presence of a disease. As in therapeutic studies, careful planning including sample size planning is essential to obtain valid results. In the planning of the sample size, various assumptions are made that may turn out to be incorrect in retrospect. While adaptive designs, which allow pre-specified design modifications and sample size re-estimation during the course of the study, are already well established in therapeutic studies, the development of these approaches for diagnostic studies is still in its early stages [1], [2], [3].
Methods: In this talk, I will present several methods for adaptive designs with blinded and unblinded interim analyses for sample size reestimation and for adaptive seamless designs. Assumptions regarding prevalence, sensitivity (true positive fraction) and specificity (true negative fraction), the proportion of discordant results, and the proportion of missing values can be corrected and studies of different phases can be combined [4], [5], [6], [7], [8]. In addition, the designs allow for early stopping due to futility and some even for efficacy. While adjustment of the type I error is not necessary in blinded interim analyses, the conditional error function approach is used for adjustment in unblinded interim analyses. The statistical properties (bias, type I error, and power) are investigated in simulation studies and the methods are applied to an example study [9].
Results: While bias and type I error are comparable in the adaptive and fixed design, the power in the adaptive design is equal to the theoretical one, while the fixed design often results in an overpowered or underpowered study and thus in too large or too small sample size. In addition, the example study shows the flexibility that these adaptive designs allow.
Discussion: Adaptive designs have the potential to make studies more efficient, especially to save time and money. However, these designs also have limitations, such as higher complexity and possibly larger sample size.
The authors declare that they have no competing interests.
The authors declare that an ethics committee vote is not required.
This contribution has already been published: ICB 2022 Riga
References
- 1.
- Zapf A, Stark M, Gerke O, Ehret C, Benda N, Bossuyt P, et al. Adaptive trial designs in diagnostic accuracy research. Stat Med. 2020;39(5):591-601.
- 2.
- Hot A, Bossuyt PM, Gerke O, Wahl S, Vach W, Zapf A. Randomized test-treatment studies with an outlook on adaptive designs. BMC Med Res Methodol. 2021;21(1):110.
- 3.
- Vach W, Bibiza E, Gerke O, Bossuyt PM, Friede T, Zapf A. A potential for seamless designs in diagnostic research could be identified. J Clin Epidemiol. 2021;129:51-59.
- 4.
- Stark M, Zapf A. Sample size calculation and re-estimation based on the prevalence in a single-arm confirmatory diagnostic accuracy study. Stat Methods Med Res. 2020;29(10):2958-2971.
- 5.
- Stark M, Hesse M, Brannath W, Zapf A. Blinded sample size re-estimation in a comparative diagnostic accuracy study. BMC Med Res Methodol. 2022 Apr 19;22(1):115.
- 6.
- Hot A, Benda N, Bossuyt PM, Gerke O, Vach W, Zapf A. Sample size recalculation based on the prevalence in a randomized test-treatment study. BMC Med Res Methodol. 2022 Jul 25;22(1):205.
- 7.
- Blohm C, Schlattmann P, Zapf A. Sample size estimation and blinded re-estimation for diagnostic studies with single-imputed missing values. Under review.
- 8.
- Hoyer A, Köster D, Brannath W, Zapf A. Unblinded reestimation of sample size indiagnostic test accuracy studies. In preparation.
- 9.
- Hot A, Stark M, Friede T, Zapf A. A two-part diagnostic seamless design to investigate the diagnostic accuracy and clinical effectiveness of a diagnostic test – applied to the HEDOS study. In preparation.
