PSI Training Course: Use of Historical Data in Clinical Trials: An Evidence Synthesis Approach
Date: Monday 24th - Thursday 27th January 2022
Time: Lectures 09:00-12:00 on the 24th & 27th. Separate forum setup for practical exercises between the 24th-26th.
Speakers: Sebastian Weber and Gaëlle Saint-Hilary
Who is this event intended for? All statisticians working on clinical trials, with a basic knowledge of R. Some notions of Bayesian statistics could be helpful.
What is the benefit of attending? Attendees will have the chance to cover; Bayesian Dynamic Borrowing designs based on the meta-analytic predictive (MAP) model; Design planning, operating characteristics, statistical analysis; and Applications using the R package RBesT.
Regular Members = £340+VAT
Regular Non-Members = £465*+VAT
*Please note: Non-Member rates include PSI membership until 31 Dec. 2022.
To book your space, please click here.
There is an intrinsic interest of leveraging all available information for an efficient design and analysis of clinical trials. Including external data in the analysis of clinical trials may increase study power or allow for the reduction of (usually) the control group sample size. The use of external data in trials are nowadays used in earlier phases of drug development (Trippa, Rosnerand Muller, 2012; French, Thomas and Wang, 2012; Hueber et al., 2012; Smith et al., 2019), occasionally in phase III trials (French et al., 2012; Viele et al., 2018), and also in special areas such as medical devices (FDA, 2010a), orphan indications (Dupont and Van Wilder, 2011) and extrapolation in pediatric studies (Berry, 1989; Best et al., 2019). This allows adequately powered trials at smaller sample sizes leading to faster trial conduct and exposure of fewer patients to a potentially in-effective control treatment.
In this short course, we will provide a statistical framework to incorporate external information into a trial. During the first part of the course, we will introduce Bayesian Dynamic Borrowing designs based on the meta-analytic predictive (MAP) model (Neuenschwander et al., 2010). The MAP model is a Bayesian hierarchical model, which combines the evidence from different potentially heterogeneous sources. Dynamic borrowing permits to limit the use of historical data when it is incompatible with the data observed within the trial.
In the second part of the course, we will propose a practice session with applications using the R package RBesT, the R Bayesian evidence synthesis tools, which are freely available from CRAN. These exercises will enable participants to apply the presented approach themselves. During third and last part of the course, more advanced topics will be detailed such as effective and maximum sample sizes, advanced operating characteristics, and probability of success.
Lecture session 1
(morning of Monday 24th January (09:00-12:00))
• Meta-analytic predictive (MAP) model
• Robustification for dynamic borrowing
• Design planning, operating characteristics
• Final analysis
(afternoon of Monday 24th, all day Tuesday 25th, all day Wednesday 26th)
• Supervised homework, set-up in a separate forum
Lecture session 2
(morning of Thursday 27th January (09:00-12:00))
• Effective sample size, maximum sample size
• Advanced operating characteristics
• Equivalence between MAP and MAC (Meta-analysis combined)
• Probability of success, decision rule
Gaëlle Saint-Hilary is Statistical Methodologist, CEO and founder of the consulting company Saryga (France). Prior to this role, Gaëlle was Statistical Methodologist at Servier until December 2021. With more than 15 years of experience in the pharmaceutical industry (Servier, Novartis) and a strong and long-lasting collaboration with academia, Gaëlle Saint-Hilary is an expert in Bayesian statistics and decision-making support. She has worked at developing novel approaches to improve drug development’s performances, and her main scientific interests are quantitative decision-making, benefit-risk assessment, innovative study designs and historical data.
Sebastian Weber is working as Director in the Department of Advanced Methodology and Data Science at Novartis. He holds a PhD in Physics from the TU Darmstadt and joined Novartis 8+ years ago. He has worked extensively on enabling the use of historical (control) information in clinical trials through consulting and working on tools to facilitate the application of historical control information from trial design to analysis. Furthermore, Sebastian has experience in designing Oncology phase I dose-escalation trails and is also involved in pediatric drug development programs, where he applies extrapolation concepts. His research interests include the application of pharmacometrics in statistics, model-based drug development and application of Bayesian methods for drug development.
PSI is a non-profit organisation run by volunteers. Many of the event organisers and presenters donate their time, while the majority of the event registration cost is spent on administrative support, venue rental / online conferencing, travel costs for the presenter, software licences, and general running of the society. PSI strives to offer high quality courses, but cannot offer a guarantee that the content presented is accurate or fit for your particular needs. Please check if any software is required for this course and ensure you are able to run it prior to registering.
Cancellation and Moderation Terms
For cancellations received up to two weeks prior to a PSI event start-date, the event registration fee will be refunded less 25% administrative charge. After this date, no refunds will be possible. A handling fee of 20 GBP per registration will be charged for every registration modification received two weeks prior or less, including a delegate name change.