Evaluating the efficacy and safety of novel therapies in rare indications can be challenging due to the difficulty of recruiting enough patients to conduct a well-powered clinical trial. To address these challenges there has been growing interest in the use of complex and innovative trial designs, which allow for borrowing of information from external data sources to supplement limited trial sample sizes1 or reduce control group allocation in a hybrid-controlled trial2 — often through the use of Bayesian methods. The U.S. Food and Drug Administration (FDA) has provided guidance on considerations for the appropriate use of Bayesian borrowing, including the importance of selecting a suitable external data source and evaluating the the type-I and type-II error operating characteristics under the proposed borrowing approach.3,4,5
Bayesian borrowing methods such as the use of power priors6 or robust meta-analytic predictive priors (robust MAP)7 provide methods for partial information borrowing from external data sources to augment a concurrent control arm or pediatric trial using trial data in adult populations. These methods allow for the external data to be down-weighted either statically, or dynamically based on the degree of similarity in outcomes between data sources.1 There has been a growing precedent for the use of Bayesian borrowing approaches, including a recent application in a post-marketing required pediatric study for treatment of pediatric systemic lupus erythematosus (SLE), which resulted in an FDA approval recommendation.8 Applications of prospective Bayesian borrowing designs have also seen uptake, for example, in a recent trial of a PDE4 inhibitor for idiopathic pulmonary fibrosis (IPF), which used Bayesian borrowing to reduce the control group allocation.9
In an upcoming contributed poster presentation at the American Statistical Association’s JSM 2023 conference in Toronto, my colleague Aaron Springford and I will be presenting recent research on the impact of hyperprior choice for Bayesian dynamic borrowing via a normalized power prior (NPP). The amount of borrowing from the external data via an NPP can be sensitive to the choice of hyperprior for the NPP discount parameter, so care needs to be taken to select an appropriate distribution. The goal of this research is to provide insight to practitioners implementing NPP Bayesian borrowing designs on how their hyperprior choices may influence trial operating characteristics.
As new drugs are developed to target a variety of rare indications, we will likely see increased use of complex and innovative designs to address the key challenges of evaluating efficacy and safety in these settings. Bayesian borrowing methods are poised to see increased uptake due to their promise for augmenting limited sample sizes using external data.
Want to learn more about Bayesian methods? Download our complimentary ebook, Topics in Bayesian Methods, below:
 Viele K, Berry S, Neuenschwander B, Amzal B, Chen F, Enas N, Hobbs B, Ibrahim JG, Kinnersley N, Lindborg S, Micallef S. Use of historical control data for assessing treatment effects in clinical trials. Pharmaceutical statistics. 2014 Jan;13(1):41-54.
 Dron L, Golchi S, Hsu G, Thorlund K. Minimizing control group allocation in randomized trials using dynamic borrowing of external control data–an application to second line therapy for non-small cell lung cancer. Contemporary Clinical Trials Communications. 2019 Dec 1;16:100446.
 Richeldi L, Azuma A, Cottin V, Hesslinger C, Stowasser S, Valenzuela C, Wijsenbeek MS, Zoz DF, Voss F, Maher TM. Trial of a preferential phosphodiesterase 4B inhibitor for idiopathic pulmonary fibrosis. New England Journal of Medicine. 2022 Jun 9;386(23):2178-87.
Read more from Perspectives on Enquiry & Evidence:
Sorry no results please clear the filters and try again