Cargando…

An Adaptive Information Borrowing Platform Design for Testing Drug Candidates of COVID-19

BACKGROUND: There have been thousands of clinical trials for COVID-19 to target effective treatments. However, quite a few of them are traditional randomized controlled trials with low efficiency. Considering the three particularities of pandemic disease: timeliness, repurposing, and case spike, new...

Descripción completa

Detalles Bibliográficos
Autores principales: Su, Liwen, Zhang, Jingyi, Yan, Fangrong
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Hindawi 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9029212/
https://www.ncbi.nlm.nih.gov/pubmed/35462681
http://dx.doi.org/10.1155/2022/9293681
Descripción
Sumario:BACKGROUND: There have been thousands of clinical trials for COVID-19 to target effective treatments. However, quite a few of them are traditional randomized controlled trials with low efficiency. Considering the three particularities of pandemic disease: timeliness, repurposing, and case spike, new trial designs need to be developed to accelerate drug discovery. METHODS: We propose an adaptive information borrowing platform design that can sequentially test drug candidates under a unified framework with early efficacy/futility stopping. Power prior is used to borrow information from previous stages and the time trend calibration method deals with the baseline effectiveness drift. Two drug development strategies are applied: the comprehensive screening strategy and the optimal screening strategy. At the same time, we adopt adaptive randomization to set a higher allocation ratio to the experimental arms for ethical considerations, which can help more patients to receive the latest treatments and shorten the trial duration. RESULTS: Simulation shows that in general, our method has great operating characteristics with type I error controlled and power increased, which can select effective/optimal drugs with a high probability. The early stopping rules can be successfully triggered to stop the trial when drugs are either truly effective or not optimal, and the time trend calibration performs consistently well with regard to different baseline drifts. Compared with the nonborrowing method, borrowing information in the design substantially improves the probability of screening promising drugs and saves the sample size. Sensitivity analysis shows that our design is robust to different design parameters. CONCLUSIONS: Our proposed design achieves the goal of gaining efficiency, saving sample size, meeting ethical requirements, and speeding up the trial process and is suitable and well performed for COVID-19 clinical trials to screen promising treatments or target optimal therapies.