Speaker

Biography

Abstract

Satrajit Roychoudhury, Pfizer

Dr. Satrajit Roychoudhury is an Executive Director and Head of the Statistical Research and Innovation group in Pfizer Inc. He has 18+ years of extensive experience in working with different phases of clinical trials for drug and vaccine. His research interest includes survival analysis, use of model-based approaches and Bayesian methods in clinical trials. Satrajit is an elected Fellow of the American Statistical Association and recipient of Royal Statistical Society(RSS)/Statisticians in the Pharmaceutical Industry (PSI) Statistical Excellence in the Pharmaceutical Industry Award in 2023 and Young Statistical Scientist Award from the International Indian Statistical Association in 2019.

The recent 21st Century Cures Act propagates innovations to accelerate the discovery, development, and delivery of 21st century cures. It includes the broader application of Bayesian statistics and the use of evidence from clinical expertise. An example of the latter is the use of trial-external (or historical) data, which promises more efficient or ethical trial designs. The FDA draft guidance “Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products” emphasized that sponsors must include patient-level data in the market-application. Though there are considerable literature discussing Bayesian methods to include summary-level external control data in design and analysis, the literature handling individual-level external control data is still small. We provided a robust meta-analytic framework when individual patient-level data is available and analyze individual patient data from various sources. We suggest a two-step approach. In the first step, the data from each source is analyzed in isolation, resulting in an estimate (and standard error) for the main parameter of the target population, taking account of the covariate information from each patient. In the second step, these adjusted data are then analyzed using a robust Bayesian hierarchical model. The utility of the method is illustrated using simulation and a case study from Alopecia Arrieta area.

Isaac Gravestock PhD is a statistician and software developer who leads the Statistical Engineering team at Roche. His interests include Bayesian methodology for external and hybrid controls, causal inference for observation data using target trial emulation and statistical computing. Previously, he was part of the Real World Data Science group at Roche for 5 years designing and analysing studies using real-world and observational data.

While the randomized controlled trial (RCT) comparing experimental and control arms remains the gold standard for evaluating the efficacy of a novel therapy, there are settings where integrating historical or external data into the control arm of an analysis, a so-called hybrid control, is essential. While the methods literature for hybrid controls is growing rapidly, the adoption of these methods is lagging. One obstacle is the lack of easy-to-use software implementations and the requirement for MCMC sampling. Therefore, we have developed psborrow2, an open-source R package which implements a selection of hybrid control borrowing methods for time to event, binary, and continuous outcomes.

The package has two main goals: First, to provide a user-friendly interface for analyzing data without the need for the user to have MCMC programming knowledge. Second, to provide a framework for conducting simulation studies to evaluate the impact of different trial and borrowing parameters (e.g., sample size, covariates) on study power, type I error, and other operating characteristics.

We demonstrate the package features and workflow with a recent example in oncology.

Christina Fillmore formally trained as a statistician, but over her 8 years at GSK transitioned into a data scientist. With a keen interest in R, she's co-lead of the PHUSE CAMIS project and R/pharma Diversity Alliance. She maintains metacore, metatools, and beastt R packages. Her focus is on creating open-source packages that enable others to deliver studies faster.

The {beastt} (Bayesian Evaluation, Analysis, and Simulation Software Tools for Trials) package offers an innovative solution to this challenge through the use of inverse probability weighted Bayesian dynamic borrowing. This methodology provides a robust framework for effectively integrating historical information while mitigating the associated risks.

The {beastt} package, developed by in partnership with the methodology and data science groups at GSK, leverages advanced Bayesian methodologies to create adaptive trial designs that are both efficient and reliable. A key feature of the package is its ability to implement an inverse probability weighted robust mixture prior, which dynamically adjusts the influence of historical data based on its compatibility with the current study population, thereby reducing potential biases and improving the integrity of trial outcomes.

In our session, we will delve into the mechanics of inverse probability weighted Bayesian dynamic borrowing as implemented in the {beastt} package. We will highlight how r packages can make it easier for methodology groups to get their innovation into the hand of the wider clinical statistics audience and how interdisciplinary partnerships are critical to advancing statistical innovations. Participants will leave with an enriched understanding of how the {beastt} package can be applied to manage the complexities of historical data integration, thereby supporting more reliable decision-making in clinical research.

Dr Antonio Remiro-Azócar is a statistical methodologist within the Methods and Outreach department in Novo Nordisk. His expertise lies in quantitative evidence synthesis, health technology assessment and data fusion. Prior to his current role, Antonio was lead statistician for health technology assessment at Bayer and an independent contractor providing statistical support to contract research organizations. Antonio holds a PhD in Statistical Science and an MSc in Machine Learning from University College London.

Integrating clinical trial data with external data requires adjusting for imbalances in baseline covariates between the data sources. The most widely used technique for this adjustment are propensity score-based weighting and outcome modelling-based approaches. We present alternative weighting methods based on entropy balancing, which directly enforce covariate balance and are generally more stable, precise and robust to model misspecification than the standard “modelling” approaches to weighting. We develop doubly robust estimators that augment the entropy balancing approaches by fitting a model for the conditional outcome expectation, then combining the predictions of the outcome model with the entropy balancing weights. The application of the methods is illustrated in an example analysis and simulations with binary outcomes and a logistic outcome model.