Novel model-based designs for early phase trials of medicines for children

In 2006 the European Union introduced legislation (Regulation on Medicines for Paediatric Use) to ensure that new medicines to be prescribed to children are subject to rigorous testing to establish their safety and effectiveness in this age group. Essential to this process is finding the optimal dose of the new medicine in children which balances potential benefits for health against the risks of an adverse side-effect. This dose will often differ from that used in adults because the ways in which the body breaks down and reacts to a drug may change subtly as it matures. Therefore using adult data to infer the optimal dose in children can lead to medicines being prescribed at ineffective or toxic doses, although this is currently often done because clinical trials are predominantly conducted in adults. It is imperative that prescribing decisions in children are supported by evidence from well-designed clinical trials in this patient group. The problem of designing paediatric trials is complex and the Fellow will engage throughout the project with parties with expertise in issues pertinent to developing medicines for children.

The main aim of this research is to devise statistical methods for trials intended to estimate the optimal dose of a medicine in children. Trials allocate each child participant a dose from a range of alternatives; data are then collected on dose received, the concentration of the drug in the blood over time and the patient's response to the drug (which may measure changes in their health). Statistical models are used to summarise how dose affects concentration and how concentration relates to response, and these models are updated as participants in the trial complete their treatment and more data become available. Based upon the updated models, each new patient enrolled in the trial is allocated the dose which is judged to maximise our learning about what the true optimal dose is.

Before the trial begins, data will often be available on how the drug performs in adults and this can influence investigators' beliefs about the optimal dose in children. This project will use methods to incorporate these existing data into the statistical models for the drug's effect in children with the intention of reducing the number of children needed to precisely estimate the optimal dose we seek. Clearly this dose may differ between young children and adolescents, for example. Methods will be extended to incorporate information on patient characteristics such as age, as well as considering adaptations necessary for successful application of the designs to trials of medicines for different diseases. Trial designs will also be proposed which shift to testing the drug's effectiveness at improving health relative to placebo once the optimal dose has been found, using all available data to make a final decision.

The evidence basis for medicines prescribed to children is often limited with efficacy and dosage extrapolated from data generated by clinical trials in adults. Extrapolation is not straightforward as differences in physiology and function, not just size, separate children and adults; this may lead to dose recommendations being either inefficacious or so high as to be unacceptably toxic. The primary aim of this project is to develop novel dose-finding procedures to estimate the optimal dose of a medicine in children which can formally borrow strength from existing data in adults or data on a different indication in children. Further project aims are to engage with key stakeholders in the clinical trials community (regulators; industry; publicly funded clinical trials units; clinical pharmacologists; clinicians) to increase the relevance of this work.

Dose-finding procedures will use Bayesian non-linear hierarchical models to synthesise available data on dose, pharmacokinetics and pharmacodynamics with the aim of identifying a quantile of the implied dose-response relationship. As each new participant enters the trial dose-assignments are made according to a Bayesian decision theoretic criterion. The performance of procedures will be evaluated. Within the Bayesian paradigm, formal use of existing information on drug behaviour in adults is possible. For the hierarchical model, prior distributions will be constructed for model forms and parameters which correctly reflect the information contributed by adult data. Extensions to dose-finding procedures include incorporating covariates such as age, and relaxing constraints initially made by the hierarchical model. Trials which progress seamlessly to tests of efficacy on termination of dose-finding will be considered, combining information accrued in both stages to make a final decision. Opportunities for applying the proposed methodology to different therapeutic areas will be actively sought throughout the Fellowship.

This research will develop statistical methods to address a highly relevant problem in health research, that of how to efficiently and accurately find the optimal dose of a medicine in children and confirm its efficacy. As such, research findings have the potential for high impact in terms of changing research practice as novel dose-finding procedures are implemented, and ultimately improving the health of children. Obtaining accurate estimates of the optimal paediatric dose which do not make extrapolations from adult data will improve child health by reducing the number treated at sub-therapeutic or toxic doses. Inaccurate paediatric dosing recommendations may be a significant problem in practice; 53 paediatric studies conducted between 1998-2002 in response to FDA requests for 33 drugs, resulted in new dosing information for seven of them [1]. The benefits of this research for child health would be felt in the short-term as new medicines begin development, since the early-phase trial designs this project proposes are intended to make efficient use of all available data in order to limit the burden of experimentation in children. Long-term benefits would become apparent as new drugs are approved for general use in children at efficacious doses.

The findings of this research activity will be of interest to all parties involved with developing and prescribing medicines to children, specifically the pharmaceutical industry, publicly funded clinical trials units, medicine regulatory bodies, clinicians and clinical pharmacologists. Particularly in the context of industry sponsored studies where paediatric drug development must follow PIPs, our designs may play a part in contributing high-quality evidence to support drug approval. The medical community would benefit as there would be new tools to improve the evidence basis for prescribing decisions. Furthermore, efficient early phase designs may make possible trials in circumstances hitherto unfeasible due to ethical or sample size constraints. Improving the quality of UK clinical research in children has been identified as a government priority [2], as illustrated by the establishment of the NIHR MCRN in 2004. The novel dose-finding designs in this proposal may also be applied more generally in adult studies with the methodology modified so that uninformative priors for models and parameters are used. Realistically, the rate of uptake of our new dose-finding methodology will be increased if user-friendly open-source packages exist for statisticians to implement them.

This project would have a positive impact on the Fellow's skills and future research career. Whilst undertaking this research, the Fellow would develop several transferable skills which could be applied in all employment sectors: leadership and project-management skills as the Fellow leads the project from inception to completion; communication skills as she communicates statistical results with non-statistical collaborators; and skills of managing relationships with inter-disciplinary collaborators.

[1]Roberts R, Rodriguez W, Murphy D, Crescenzi T. (2003) Pediatric drug labelling. Improving the Safety and Efficacy of Pediatric Therapies. JAMA 290: 905-911.
[2] MHRA/Department of Health Strategy on Medicines for Children. Accessed on http://www.mhra.gov.uk/Howweregulate/Medicines/Medicinesforchildren/index.htm#l6