PSI Events

PSI One Day Meeting: Bayesian Methods for Dose Finding and Biomarkers

  • Dates: 28 – 28 Feb, 2018

Timings: 09:30 - 17:00 UK Time
Location: Royal Statistical Society
Address: 12 Errol Street, London

Bayesian methods are being increasingly used in the design and analysis of trials through all stages of clinical drug development. This meeting will include talks on two areas of application commonly used in early phase trials – dose selection and biomarkers. The meeting will include a range of case studies presented from different therapeutic areas. More details to follow. 

Registration is now closed.

Agenda

Time

Agenda

09.30 - 09.55 Registration

09.55 - 10.00

Welcome and Introduction

10.00 - 10.40

First-in-Human Studies: Practice or Principle
Andy Grieve, UCB Pharma

10.40 - 11.20

Model based dose escalation: in practice
Phil Overend, AstraZeneca

11.20 - 11.40

Break

11.40 - 12.20

A Bayesian information theoretic design for phase I dose finding trials without monotonicity assumption
Pavel Mozganov, University of Lancaster

12.20 - 13.00

Bayesian hierarchical model for dose escalation
Alessandro Matano, Novartis

13.00 - 13.50

Lunch

13.50 - 14.30

The use of formally elicited priors to aid the design of a phase 2B dose ranging study
Doug Thompson, GlaxoSmithKline

14.30 -15.10

Bayesian methods for design and analysis of clinical trials assessing multiple treatments and biomarkers
Prof James Wason, MRC Biostatistics Unit

15.10 - 15.30

Break

15.30 - 16.10

Error rates control via shrinkage priors in multivariate error-in-variable dose-response models
Fabio Rigat, GlaxoSmithKline

16.10 - 16.50

Prior elicitation and translation of biomarkers to aid clinical development plans and compound selection
Trevor Smart, Lilly

16.50 - 17.00

Close


Abstracts


rev bayes 
Rev Thomas Bayes
Bayesian methods are being increasingly used in the design and analysis of trials through all stages of clinical drug development. This meeting will include talks on two areas of application commonly used in early phase trials – dose selection and biomarkers. The meeting will include a range of case studies presented from different therapeutic areas. 
Andy Grieve, UCB First-in-Human Studies: Practice or Principle

Abstract: 
The last 12 years has seen two FIH studies hit the newspaper headlines: the study of TGN1412 conducted at Northwick Park Hospital in the UK and the study of BIA 10‐2474 conducted in Rennes in France. Both studies led to catastrophic outcomes for the volunteers involved. As a result of the first study two investigations were established. The first was set-up by the MHRA and their final report, Duff et al (2006), appeared in December 2006. The second was established by the Royal Statistical Society(RSS) to look at statistical issues involved in FIH studies. The resulting report, Senn et al (2007), made 21 recommendations, both practical and methodological. After the second study a sub-group of the RSS working party provide a commentary on statistical issues involved, Bird et al (2017). As a result of the Duff report the Commission for Human Medicines established an Expert Advisory Group (EAG) on Clinical Trials in 2007 to review FIH protocols for three types of new compounds. I was a co-author on the two statistical reports and served on the EAG. In this talk I review practical and methodological issues involved in FIH studies based on my experience serving on these three groups.

Bird SM, Bailey RA, Grieve AP, Senn S. Statistical issues in first‐in‐human studies on BIA 10‐2474: Neglected comparison of protocol against practice. Pharmaceutical Statistics 2017;16:100-6.

Expert Scientific Group on Phase One Clinical Trials (Final report of a group convened by UK's Medicines and Healthcare Regulatory Authority and chaired by Professor Sir Gordon Duff), Final report online 6 December 2006, see http://webarchive.nationalarchives.gov.uk/+ /dh.gov.uk/ en/ publicationsandstatistics/ publications/publicationspolicyandguidance/ dh_063117 (acc. Jan 2018).

Senn S, Amin D, Bailey RA, Bird SM, Bogacka B, Colman P, Garrett A, Grieve A, Lachmann P. Statistical issues in first‐in‐man studies. Journal of the Royal Statistical Society: Series A (Statistics in Society) 2007; 170:517-79.

Phil Overend, AstraZeneca

Model based dose escalation: in practice

Abstract: At AstraZeneca, the implementation of a model based Bayesian dose escalation design in Oncology phase 1 studies has been facilitated by putting practical considerations in place that address the operational challenges that arise due to increased statistical complexity that accompanies the methodology. We will discuss a practical case study, describing the first time that this approach was implemented at AstraZeneca, including the options explored for specifying the prior, the advantages of the approach together with practical and methodological learnings for future studies.  The extension to 2-drug combination dose escalation studies will be introduced.

Pavel Mozgunov A Bayesian information theoretic design for phase I dose finding trials without monotonicity assumption

Abstract: Methods for finding the highest dose that has an acceptable risk of toxicity in Phase I dose-escalation clinical trials assuming a monotonic dose-response relationship have been studied extensively in recent decades. The assumption of monotonicity is fundamental in these methods. As a result, such designs fail to identify the correct dose when the dose-response relationship is non-monotonic, for example, in dose-schedules and dose-combinations trials. We propose a dose-escalation method that does not require monotonicity or any pre-specified relationship between dose levels. The design is based on the Bayesian model incorporating an information-theoretic criterion for dose selections. We will discuss the practical aspects of the design such as prior and safety stopping. For small sample size typical for dose-escalation studies we will show in simulations that the proposed method is comparable to well-studied and used methods under the assumption of monotonicity and outperforms them when this assumption is violated. We conclude with an extension of the design which allows to include an efficacy endpoint.

Biography: Pavel did his undergraduate studies in the National Research University Higher School of Economics, Moscow, Russia. He worked on information-theoretic aspects in statistics in the International Laboratory of Stochastic Analysis. Currently, Pavel is a PhD student in Lancaster University where he works on the project “Dose finding for combination trials with many treatments” under supervision of Professor Thomas Jaki. Pavel is a Marie-Curie Fellow in the IDEAS network. The aim of his project is to develop novel dose finding designs which do not require parametric assumptions and can be applied to various clinical trials. 

Alessandro Matano Bayesian hierarchical model for dose escalation

Abstract: Clinical trials with multiple strata (e.g. indication, dose regimen, mutation status) are increasingly used in drug development. In dose escalation trials, where data are often sparse and a multiple strata analysis may be the only option to study a new treatment, good statistical inference and decision-making can be challenging. Inferences from simple pooling or stratification are known to be inferior to hierarchical modeling methods, which build on exchangeable strata parameters to allow borrowing information across strata. However, the standard exchangeability (EX) assumption bears the risk of too much shrinkage and excessive borrowing for extreme strata. This talk will present the exchangeability-nonexchangeability (EXNEX) approach in dose finding studies, as a robust extension of the standard EX approach. It allows each stratum-specific parameter to be exchangeable with other similar strata parameters or nonexchangeable with any of them. A case study from a Phase I trial will also be shown.

Biography: After the degree in Statistics for biomedicine, environment and technology from the university of Rome “La Sapienza” (2008), Alessandro joined the pharmaceutical industry (Novartis). He is currently working as Associate Director of Biostatistics in the area of Phase I/II Oncology clinical trials in Early Development Biostatistics (EDB) group. Prior to join the EDB Oncology group Alessandro has worked as trainee in the CIS-Methodology Group, supporting the methodological development and implementation of oncology Bayesian single agent/combination dose-finding study design.

Doug Thompson

Doug Thompson, GSK
The use of formally elicited priors to aid the design of a phase 2B dose ranging study

Abstract: The use of assurance to quantify the probability of success of a proposed parallel group clinical trial is well understood. This framework extends naturally to studies with other designs; including when evaluating dose response is the primary focus. Prior distributions formally elicited about the mean response at certain dose levels can be used to set certain constraints such that a spread of plausible three parameter E-max curves may be generated. In this presentation, we will discuss our experiences in the development of a dose response trial simulator. We illustrate how such an approach may be used to help characterise interim futility rules, dose adaptation and assurance.

Biography: Doug Thompson has been at GlaxoSmithKline for over three years working on the design, analysis and reporting of clinical trials in rare diseases. He studied statistics at the University of Glasgow as an undergraduate before going on to undertake a PhD at the University of Edinburgh on the development of a risk stratified approach for the treatment of acute stroke patients, funded through the MRC Hubs for Trials Methodology Research.

James Wason Bayesian methods for design and analysis of clinical trials assessing multiple treatments and biomarkers

Abstract: Response to treatments is often highly heterogeneous, especially ones that are targeted at specific biological pathways. The increasing availability of biomarkers and targeted treatments has led to the need for trial designs that efficiently test new treatments in biomarker-stratified patient subgroups. Often new treatments are targeted at a specific biomarker subgroup, but may in fact work in a narrower or broader set of patients.

I will discuss Bayesian adaptive methodology for trials that have multiple treatments and biomarkers. The proposed design incorporates biological hypotheses about the links between treatments and biomarker subgroups, but allows alternative links to be formed during the trial. The statistical properties of the method compare well to alternative designs available.  This design has been developed for trials in ovarian cancer and breast cancer and some methodology issues specific to each application will be discussed. These include the use of continuous biomarker information to allocate patients and adding in new treatments and biomarkers during the trial.

Biography: James Wason is a Professor of Biostatistics at Newcastle University and a Programme Leader at the MRC Biostatistics Unit. He leads a research group focused on development of statistical methods for more efficient trials. His research interests are adaptive designs, using biomarkers prospectively during trials and analysis of composite endpoints. As well as working on methodology, James is very interested in applying novel methods in real clinical trials and is currently involved in several ongoing trials using novel designs.

Fabio Rigat, GSK Error rates control via shrinkage priors in multivariate error-in-variable dose-response models

Abstract: When planning & analysing limited numbers of dose-response data, the risk of false positive results may be considerable. In this case, priors can be used to penalise inferences for dose-response parameters so as to reduce this risk. We demonstrate a simple example of such priors for analysing the multivariate relation between the dose of an investigational asset administered to a cohort of cancer patients and the associated average change in the expression of several target genes. This analysis is further robustified by identifying error terms capturing the technical variability of these biomarker data from those representing sampling variation. 

trevor smart Prior elicitation and translation of biomarkers to aid clinical development plans and compound selection

Abstract: The talk will show how prior elicitation of a pharmacodynamic (PD) biomarker can be used to estimate a design prior to be used in designing clinical trials.  In addition, if the translation of the biomarker to efficacy or safety is elicited, then this can help evaluate the benefits of the biomarker and how decisions may be made based on the biomarker.  An example will be given where efficacy and safety biomarkers are used to compare several compounds being considered for the same indication.  The elicitation of the translation allows us to weigh up which biomarker is most relevant in making the decision over which compound to progress first.  When this decision can be made, pre-clinical, Phase 1 or Phase 2, can be compared.

Biography: Trevor has been the Group Lead for Early Phase Neuroscience Statistics at Eli Lilly since September 2011 covering neurodegeneration, psychiatric illnesses and pain.  This covers the clinical planning after Candidate Identification through to the decision to proceed to Phase 3.  Prior to this he was at Pfizer in Kent from 1998, where he worked on early clinical development as well.  Much of the work was in pain and many biomarker and methodology studies including: fMRI, other imaging biomarkers, healthy volunteer challenge models and patient biomarkers. Trevor has also worked for BioSS in Scotland and at Rothamsted.



Registration is now closed.

 

Registration Fees

PSI Member

£40 + VAT

Non-Member

£135 + VAT
(This includes free PSI membership for the remainder of 2018)



Please register before Wednesday 21st February 2018. 

Please contact the PSI secretariat on psi@mci-group.com if you have any dietary requirements or queries about the event.

Andy Grieve, UCB Pharma

Andy Grieve, UCB Pharma

Andy Grieve, UCB Pharma

Bayesian methods in a seamless Phase 2/3 diabetes trial
Tom Parke, Berry Consultants

Error rates control via shrinkage priors in multivariate error-in-variable dose-response models
Fabio Rigat, GSK

Error rates control via shrinkage priors in multivariate error-in-variable dose-response models
Fabio Rigat, GSK

Astrazeneca

Head, Statistical Modelling & Simulation Services

read more...

image001

GlaxoSmithKline is looking for Talented Clinical Statisticians

read more...

Sanofi

Biostatistician Project Leader- Early Clinical Development

read more...