Statistical Considerations of RWE/RWD
Organizers: Qi Jiang (Seagen), Jean Pan (Amgen)
Chair: Shuyen Ho (UCB)
Vice Chair: Biao Xing (Seagen)
Martin Ho (FDA)
Ying Yuan (MD Anderson)
Estelle Russek-Cohen (ERCStatLLC / retired FDA)
Title: Landscape of Causal Inference Frameworks of Clinical Studies using RWD/RWE
Speaker: Martin Ho (FDA)
Surging number of available real-world data (RWD) sources (e.g., EHRs, claims data, registries) have attracted many experts from multiple disciplines to actively exploring and developing ways to translate RWD into Real-World Evidence (RWE) to bridge the current gaps in clinical trial enterprise. However, the statistical community can participate more fully in this novel research area. Therefore, the ASA BIOP Section RWE Scientific Working Group (SWG) was chartered in 2018, aiming to use statistics generate RWE designed to inform regulatory decisions of medical products. The SWG has two work streams and they have been focusing on using RWE: (a) to modify existing labels of medical products and (b) to inform better clinical study designs. In this presentation, early study results of the SWG will be presented. This presentation will first discuss the current statistical landscape on using RWD and other external data, followed by some examples, and then concluded with some identified topics where the SWG can make most substantial impacts and future research agendas.
Title: A Bayesian group sequential design for randomized biosimilar clinical trials with adaptive information borrowing from historical data
Speaker: Ying Yuan (MD Anderson)
At the time of developing a biosimilar, the reference product has been on market for years and thus ample data are available on its efficacy and characteristics. We develop a Bayesian adaptive design for randomized biosimilar clinical trials to leverage the rich historical data on the reference product. The design takes a group sequential approach. Patients initially are 1:1 randomized into the test arm and the reference arm. At each interim, we employ the elastic meta-analysis predictive (EMAP) prior methodology to adaptively borrow information from the historical data of the reference product, and calculate the posterior probability that the efficacy of the test product is located within the biosimilar margins, with respect to the reference product. This posterior probability is used to make go/no-go decision. If the decision is “go,” we determine the amount of information that is borrowed from the historical data, measured by the effective sample size (ESS), and adaptively adjust the subsequent randomization ratio with the goal to balance the sample size of the two arms at the end of trials. Simulation study shows that the proposed Bayesian adaptive design can substantially reduce the sample size of the reference arm, while achieving comparable power as the traditional randomized clinical trials that ignore the historical data. We apply our design to a biosimilar trial for treating breast cancer patients. This is a joint work with Wen Zhang and Jean Pan.
Title: Real World Evidence and Rare Disease Drug Development
Speaker: Estelle Russek-Cohen (ERCStatLLC / retired FDA)
I will cover 3 topics in the context of rare disease drug development: 1) The importance of Natural History in rare disease drug development including several examples. 2) When it is appropriate to consider RWE as external controls or possibly a hybrid of external controls augmenting concurrent controls. 3) To utilize RWE in post-market. Along the way I will discuss some relevant FDA guidances and other resources that may be helpful.