MICA: Towards using historical data for research prioritisation in children

Lead Research Organisation: Lancaster University
Department Name: Mathematics and Statistics

Abstract

Context of the research: New medicines for children must be rigorously assessed to ensure that they are both safe and effective. So that unnecessary clinical trials in children may be avoided, assessments of the risks and benefits of a new medicine should incorporate existing relevant data. For example, if adult data are considered relevant, we may extrapolate from evidence a medicine is effective in adults to conclude that it will be beneficial in children also if prescribed at doses yielding concentrations in the blood that are therapeutic in adults. However, for this conclusion to be valid, disease progression and the relationship between drug concentration and clinical response must be similar in adults and children. Erroneously assuming similar concentration-response relationships will lead to children receiving excessively toxic or sub-therapeutic doses. Regulators have proposed algorithmic approaches for determining which studies are needed in children to support medicine development but these do not accommodate uncertainty about extrapolation assumptions. Furthermore, assumptions may only hold in certain subgroups of children.

Aims and objectives: The proposed project aims to develop statistical methods for quantifying uncertainty about assumed similarities between adults and children. Specifically, the first workpackage of this project will develop an approach using data from historical clinical trials to measure the strength of evidence supporting a claim of similar concentration-response relationships in adults and children for a new medicine. Historical data will be down-weighted to account for differences between historical and future patients. Decision rules will be formulated which use this information to make decisions about whether additional clinical data in children are needed to verify an assumption of similar concentration-response relationships. Work will also formulate designs for clinical trials conducted with the aim of verifying an extrapolation assumption; designs will be efficient because decisions will be based on all available data. The second workpackage of the project will explore model-based approaches for classifying children into groups with different concentration-response relationships.

Potential applications and benefits: Research will be embedded in three conditions: while our primary focus will be on epilepsy research, for comparative purposes, applications to HIV and asthma medicine will also be considered. Anticipated beneficiaries of the proposed research span regulatory agencies, public sector clinical trials units, the pharmaceutical industry and academia. The research has the potential to benefit the health of children because methods will help investigators to make informed decisions about: a) the plausibility of similarities between adults and children; and b) which data are needed in children to evaluate the risks and benefits of a medicine. The project will work with GlaxoSmithKline and Novartis. Software implementing developed methodologies will be made publicly available to encourage their uptake by practitioners.

Technical Summary

The proposed research will be structured into two workpackages (WPs). Methodological developments will be embedded in three indications, namely, epilepsy, human immunodeficiency virus (HIV) and asthma. WP1 will develop Bayesian multivariate random effects meta-analytic models for synthesizing individual patient data from historical pharmacokinetic-pharmacodynamic (PK-PD) studies in order to quantify the strength of prior evidence supporting an assumption of similar PK-PD relationships in adults and children for a novel medicine. Historical data may be related, but not perfectly relevant, to the assumption we wish to verify because, for example, historical trials concerned a different but similar medicine to the one of current interest. Therefore, the meta-analytic model will incorporate expert opinion on the effects of biases on key parameters that arise because historical data are not perfectly relevant to the target question. Strategies and software for eliciting expert opinion on the effects of biases will be developed. A Bayesian decision theoretic approach will be adopted to measure the expected utility of collecting PK-PD data in children for verifying assumed similarities with adults, given the strength of historical evidence supporting assumptions. Utilities will incorporate sampling costs and the costs of making erroneous assumptions. Work will also formulate Bayesian adaptive designs for paediatric PK-PD studies conducted to verify an extrapolation assumption. Properties of developed methodologies will be illustrated using simulation and through application to datasets provided by GlaxoSmithKline. WP2 will explore using frequentist and Bayesian finite regression mixture modelling as a data driven approach for determining the number and definition of paediatric subgroups with different PK-PD relationships. Open source software implementing proposed methodologies will be developed.

Planned Impact

This project seeks to develop a timely and practical solution to the challenge of how to characterise and respond to uncertainty about an extrapolation assumption.

Who will benefit: We anticipate that beneficiaries of the proposed research will include children, regulatory agencies, public sector clinical trials units and the pharmaceutical industry. End-user interest in this work is evidenced by the collaborative links established with Novartis and GlaxoSmithKline. The Pathways to Impact statement accompanying this application outlines the steps that will be taken to ensure the proposed research has an international impact.

How will they benefit: The proposed research has the potential to benefit the health of children because many treatments are currently used off-label and hence may be used at inappropriate doses. The proposed work will help to make more rigorously evaluated treatments available to the paediatric population. Methods will contribute towards an improved approach to dose-finding in children, whilst also ensuring that opportunities for extrapolation are exploited when historical data support this.

Short-term beneficiaries of the proposed research include statisticians; clinical pharmacologists; and pharmacometricians involved with designing development programmes for new medicines for children. The proposed project will devise methods that these beneficiaries may use to derive objective and easily interpretable summaries of the current evidence supporting an assumption of similar pharmacokinetic-pharmacodynamic relationships in adults and children, and the value of collecting data to verify this. In the EU, we anticipate that such summaries could be cited to support a planned extrapolation strategy in an application for a paediatric investigation plan submitted to the European Medicines Agency (EMA). Regulatory bodies, such as the Paediatric Committee of the EMA, could then use this evidence to weigh up the risks and benefits of a proposed extrapolation strategy. To ensure these research benefits are realised, open source software implementing the novel methodologies will be developed. 'How-to' papers communicating research findings will also be prepared to encourage uptake of the proposed methods in practice.

There has been recent interest from regulatory agencies and public bodies about how to quantify and respond to uncertainty when evaluating new medicines. A recent EMA concept paper [1] highlighted the need for a coherent framework for extrapolation whereby uncertainty about assumptions is documented and used to develop an extrapolation concept. The Institute of Medicine recently held two workshops on "Characterising and communicating uncertainty in the assessment of benefits and risks of pharmaceutical products" (www.iom.edu/Activities/Research/DrugForum/2014-FEB-13.aspx). The proposed research will contribute to these wider discussions on drug development.

The proposed project will build capacity in trials methodology research. The project RA, Mr Wadsworth, will work under the supervision of two statisticians, Drs Hampson and Jaki, whilst also receiving support from an interdisciplinary team of experienced researchers. The PI, who currently holds an MRC Career Development Award in Biostatistics, will develop her leadership skills through co-ordinating the project, supported by co-applicants who have a strong track record in methodology research.

References:
[1] European Medicines Agency. (2012) Concept paper on extrapolation of efficacy and safety in medicine development.
 
Description GSK 
Organisation GlaxoSmithKline (GSK)
Country Global 
Sector Private 
PI Contribution Joint research
Collaborator Contribution Input to research and data provision
Impact Data have been made available by the partner
Start Year 2015
 
Description Liverpool 
Organisation University of Liverpool
Department Department of Molecular and Clinical Pharmacology
Country United Kingdom 
Sector Academic/University 
PI Contribution Joint research
Collaborator Contribution Input into research and practical perspective.
Impact Joint publications
Start Year 2015
 
Description Novartis 
Organisation Novartis
Department Novartis Statistical Methodology Group
Country United Kingdom 
Sector Private 
PI Contribution Joint research
Collaborator Contribution Input into joint research and practical perspective
Impact Joint publications
Start Year 2015
 
Title App for Prior Elicitation 
Description R package for eliciting priors for similarity assessment of PK-PD relationships in adults and children 
Type Of Technology Software 
Year Produced 2018 
Open Source License? Yes  
Impact NA 
URL https://github.com/iwadsworth/ElicitBiasPrior
 
Description BAYES Pharma 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Industry/Business
Results and Impact Bayes Pharma 2017 - Contributed talk "Using historical data to inform extrapolation decisions in children"
Year(s) Of Engagement Activity 2017
 
Description CEN-ISBS 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact CEN-ISBS 2017 - Contributed talk "Using historical data to inform extrapolation decisions in children"
Year(s) Of Engagement Activity 2017
 
Description Dissemination Workshop 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Industry/Business
Results and Impact We hosted a dissemination workshop at the Royal Statistical Society in London in May 2018 which around 40 people coming from industry, regulatory agency and academia attended. Strategies for pediatric drug development through extrapolation where presented and discussed.
Year(s) Of Engagement Activity 2018
URL http://www.lancaster.ac.uk/maths/paediatric-extrapolation/
 
Description ICTMC 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact 4th ICTMC & 38th Annual Meeting of SCT - Invited talk "Using historical data to inform extrapolation decisions in children"
Year(s) Of Engagement Activity 2017
 
Description ISCB 2016 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact ISCB 2016 - Contributed talk "Clinical drug development in epilepsy revisited: a proposal for a new paradigm streamlined using extrapolation"
Year(s) Of Engagement Activity 2016
 
Description PSI Extrapolation meeting 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Industry/Business
Results and Impact PSI One Day Meeting: Extrapolation - Invited talk "Using historical data to inform extrapolation decisions in children"
Year(s) Of Engagement Activity 2017