PSI One Day Meeting: Estimands – Examples for Statisticians!

  • Dates: 27 – 27 Sep, 2017
  • Location: 500 Brook Drive, RG2 6UU, Reading
  • Address: QuintilesIMS

Wednesday 27 September - QuintilesIMS, 500 Brook Drive, Reading, RG2 6UU, UK

With the recent release of the ICH E9 Addendum, Estimands is moving from a concept to reality with some Regulators already asking “What is your Estimand?”  But where do you start when defining your Estimand?  What is the Estimand of interest and to whom? What inter-current events could determine your Estimand and thus your analysis methodology?  The PSI Scientific Committee have put together this one day meeting to provide Statisticians with real world examples of how Estimands have been defined.  Presenters will give their insight into discussions with colleagues, working groups and regulators and there will be plenty of opportunity to ask your questions on defining an Estimand. 

Please click here to view the flyer.

The agenda is outlined below:

9.30 – 9.55

Registration

9.55 – 10.00

Welcome and Introduction

10.00 – 10.40

Estimands and Sensitivity Analyses: What’s in the new ICH E9 Addendum?

 

Chrissie Fletcher, Amgen

10.40 – 11.20

The Estimands Concept – Experiences when Introducing the Concept in a Global Development Organization

 

Christoph Gerlinger, Bayer

11.20 – 11.35

Break

11.35 – 12.15

Disentangling Estimands and the Intention-to-Treat Principle

 

Ann-Kristin Leuchs, BfArM

12.15 – 12:55

Use of Clear Estimands – Beyond Hypothesis Testing

 

Yolanda Barbachano, MHRA

12.55 – 13.45

Lunch

13.45 – 14.25

A Journey towards Estimand Specification in Pain: Motivation and Challenges

 

Francesca Callegari, Novartis

14.25 – 15.05

Estimands in Chronic symptomatic Diseases

 

Martin Jenkins, AstraZeneca

15.05 – 15.25

Break

15:25 – 16:05

Are Estimands Necessary for Time-To-Event Endpoints?

 

Chris Harbron, Roche

16:05 – 16:30

Panel Discussion

 

All presenters

16:30 – 16:35

Closing remarks – Finish



Estimands and Sensitivity Analyses: What’s in the new ICH E9 Addendum?

Chrissie Fletcher, Amgen

The new addendum to ICH E9 on estimands and sensitivity analyses introduces a new framework for clinical trial design, conduct, analysis and interpretation of results. In the new framework the first step is to ensure there is a clearly defined clinical trial objective. The trial objective will lead to defining the estimand, the treatment effect to be estimated, which will influence the choice of trial design. The estimand will lead to defining appropriate statistical analyses to derive estimates of treatment effects, including sensitivity analyses that are aligned to the estimand.

The new framework in the ICH E9 addendum will enable sponsors to discuss with regulators prior to the clinical trial commencing what estimand is of primary interest. Choices made in the study design and planned statistical analyses describing how intercurrent events, such as non-adherence, use of rescue medication, and deaths occurring in the study, will be handled can impact what treatment effect is actually being estimated in a clinical trial. Therefore alignment in the choice of estimand and planned statistical analyses, including sensitivity analyses, will improve the interpretation and understanding of trial results.

This presentation will provide an overview of the new addendum including examples illustrating how to use the new framework in designing clinical trials.

Click here to view the slides.

 

The Estimands Concept – Experiences when Introducing the Concept in a Global Development Organization

Christoph Gerlinger, Bayer

The forthcoming addendum to the ICH E9 guidance on statistical principles for clinical trials has major implications on almost all aspects of drug development trials. Despite being a multidisciplinary topic it is often perceived as an exclusively statistical topic in adjacent functions like medical, project management, or regulatory affairs.

A broad working group of statisticians interested in the estimands framework was founded in our company early 2016 to prepare not only the statisticians but also the whole company for the changes in the way we plan, run, and analyze clinical trials in the future.

This talk will review the actions taken before the release of the ICH draft addendum: Creation of a white paper, summaries of the key publications, and two pilot workshops all aimed mainly at clinicians in drug development. We will also discuss our plans to roll out the estimands concept both within the statistics department and also to the whole company once the draft addendum is published.

Click here to view the slides.

 

Disentangling Estimands and the Intention-to-Treat Principle

Ann-Kristin Leuchs, BfArM

The precise definition of the estimand of primary interest (treatment effect to be estimated) with regard to specification of handling intercurrent events such as recue medication or non-adherence is essential when planning and designing randomized controlled trials (RCT). The analysis should then be aligned to the agreed primary estimand. In this context the question arises how this fits in with the intention-to-treat (ITT) principle. Although the ITT principle has long-since been the gold standard of analyzing RCTs. Despite this there was and still is much ambiguity involved around what is considered to constitute ITT, especially in relation to the problem of missing data and, in recent years, also with regard to intercurrent events and estimands. While some argue it is simply analyzing all patients as randomized, others regard ITT as addressing the treatment policy estimand.

This talk focuses on the author’s thinking on the ITT principle and its definition, on how it is distinct from the missing data and estimand problem and on how to best move forward. Since ICH E9 is imprecise concerning the ITT principle, discussing an addendum to ICH E9 might be an ideal time point to solve the confusion and ambiguity around defining it. The ITT principle could well remain gold standard of analyzing RCTs even while allowing various different estimands to conform to it.

Click here to view the slides.

 

Use of Clear Estimands – Beyond Hypothesis Testing  

Yolanda Barbachano, MHRA

Though the Addendum to ICH E9 was clearly motivated by a need to more precisely define the measure of treatment effect in clinical trials, in the context of efficacy, the same framework is also applicable and helpful when thinking about how to collect and present the safety data. Furthermore, we should ideally think about the estimand of interest for each endpoint or trial objective separately, regardless of whether they are primary or secondary, hypothesis testing or descriptive.  In this talk I will move away from the usual discussion around the choice of estimand for the primary efficacy endpoint, and instead present some examples on safety, tolerability and quality of life, to illustrate the value of the estimand framework in a wider context.  

Click here to view the slides.

 

A Journey towards Estimand Specification in Pain: Motivation and Challenges 

Francesca Callegari, Novartis

An estimand clearly defines the treatment effect to be estimated in a clinical trial. An ICH E9 addendum is under preparation, which will introduce the concept of estimand and will provide a structured framework to link trial objectives of a clinical trial and statistical methods in a coherent way.

In the meantime, regulators are keen to know the definition of estimands for new clinical trials. In this presentation, we focus on a Phase 2 study in chronic pain. The definition of the primary estimand in this context takes into account relevant post-randomization events, which are often informative of the treatment effect of interest, such as intake of concomitant medications and premature discontinuations of study treatment. Other supplementary and secondary estimands are also defined to assess the treatment effect under different handling of the post-randomization events or under different specifications of the variable of interest. Some practical considerations coming from the development of the estimand concept for this trial from its inception till its detailed specification are summarized, outlining the challenges encountered and how these have been overcome.

Click here to view the slides.

 

Estimands in Chronic symptomatic Diseases

Martin Jenkins, AstraZeneca

In many chronic, systemic diseases the goal of treatment is to manage patient symptoms and to prevent disease flares. The effect of current therapies is generally reversible and as such it is not necessarily of primary interest to address a treatment policy estimand, but rather to consider the effects attributable to the initially randomised treatment. In addition, defined treatment pathways mean that it is common that estimands in this area must consider the handling of patients who use rescue treatments or escalate therapy. Drawing on examples in rheumatology, dermatology, autoimmune and respiratory disease areas I will compare different scenarios to describe how the precise choice of estimand should take into account the type of endpoint, current treatment paradigm and any retention of treatment effect upon discontinuation.

Click here to view the slides.

 

Are Estimands Necessary for Time-to-Event Endpoints?

Chris Harbron, Roche

Observed time-to-event endpoints typically contain many censored observations. As a consequence, many of the standard analysis approaches e.g. Kaplan-Meier and Proportional Hazards are specifically designed to address these partially missing data. This ability to cope with data being missing due to censoring has frequently led to the benefits of estimands for addressing other types of intercurrent events being overlooked.

In this presentation I will discuss how the estimand framework provides a vehicle for explicitly describing and addressing several of the challenges within time-to-event analyses such as treatment cross-over, informative censoring, lack of blinding and inconsistent definition of endpoints.

Click here to view the slides.

 

Registration is now closed.
Add event to: ICal Outlook Google Calendar