EPSRC Centre for Predictive Modelling in Healthcare

Our Centre brings together a world leading team of mathematicians, statisticians and clinicians with a range of industrial partners, patients and other stakeholders to focus on the development of new methods for managing and treating chronic health conditions using predictive mathematical models. This unique approach is underpinned by the expertise and breadth of experience of the Centre's team and innovative approaches to both the research and translational aspects.

At present, many chronic disorders are diagnosed and managed based upon easily identifiable phenomena in clinically collected data. For example, features of the electrical activity of the heart of brain are used to diagnose arrhythmias and epilepsy. Sampling hormone levels in the blood is used for a range of endocrine conditions, and psychological testing is used in dementia and schizophrenia. However, it is becoming increasingly understood that these clinical observables are not static, but rather a reflection of a highly dynamic and evolving system at a single snapshot in time. The qualitative nature of these criteria, combined with observational data which is incomplete and changes over time, results in the potential for non-optimal decision-making.

As our population ages, the number of people living with a chronic disorder is forecast to rise dramatically, increasing an already unsustainable financial burden of healthcare costs on society and potentially a substantial reduction in quality of life for the many affected individuals. Critical to averting this are early and accurate diagnoses, optimal use of available medications, as well as new methods of surgery. Our Centre will facilitate these through developing mathematical and statistical tools necessary to inform clinical decision making on a patient-by-patient basis. The basis of this approach is patient-specific mathematical models, the parameters of which are determined directly from clinical data obtained from the patient. As an example of this, our recent research in the field of epilepsy has revealed that seizures may emerge from the interplay between the activity in specific regions of the brain, and the network structures formed between those regions. This hypothesis has been tested in a cohort of people with epilepsy and we identified differences in their brain networks, compared to healthy volunteers. Mathematical analysis of these networks demonstrated that they had a significantly increased propensity to generate seizures, in silico, which we proposed as a novel biomarker of epilepsy. To validate this, an early phase clinical trial at King's Health Partners in London has recently commenced, the success of which could ultimately lead to a revolution in diagnosis of epilepsy by enabling diagnosis from markers that are present even in the absence of seizures; reducing time spent in clinic and increasing accuracy of diagnosis. Indeed it may even make diagnosis in the GP clinic a reality.

However, epilepsy is just the tip of the iceberg! Patient-specific mathematical models have the potential to revolutionise a wide range of clinical conditions. For example, early diagnosis of dementia could enable much more effective use of existing medication and result in enhanced quality and quantity of life for millions of people. For other conditions, such as cortisolism and diabetes where a range of treatment options exist, identifying the optimal medication, and the pattern of its delivery, based upon the profile of the individual will enable us to maximise efficacy, whilst minimising unwanted side effects.

Building on a substantial track-record of engagement with end-users (including patients and their carers, clinicians, device manufacturers and diagnostics companies, the pharmaceutical industry and funding agencies) we have developed a set of activities for User engagement that will build upon the research core of the Centre and form the basis of future impact. These activities include:

- bespoke training and a visitor programme (for both clinical researchers and practitioners) to introduce key concepts and to develop understanding of predictive models; a key first step to facilitate the ultimate uptake of these methods in practice
- researcher exchanges where members of the Centre and researchers from our industry partners will spend time working on projects of mutual interest
- "hot topics" workshops and sandpits where groups of scientists and clinicians spend a focused period discussing the very latest findings and challenges and suggest a roadmap for future travel
- "research incubator" retreats. Here we will bring mathematicians, clinicians and clinical scientists together with relevant industry to develop proposals for pump-priming new areas of research. This will be facilitated by our industry partner Life Sciences SouthWest and members of the Universities' Research & Knowledge Transfer teams, focused on research development, impact and commercialisation will help teams to develop bids and budgets to maximise the likelihood of substantive impact arising from these projects

Having utilised similar activities to establish a strong track-record of partnering with clinicians and industry already, we foresee substantial impact in the areas of epilepsy diagnostics and dynamic treatments for cortisolism arising during the initial period of EPSRC funding. In epilepsy, we have already taken steps to protect intellectual property through patenting the concept that time-varying changes in a network model of electroencephalography can predict clinical outcomes (joint patent between the University of Exeter and King's College London). We are currently undertaking an early phase clinical trial to validate this hypothesis. From here there is the opportunity for commercialisation, either linking with our project partner MentisCura or through a University spin-out company. To explore this latter option, we have recently secured SetSquared funding through their innovation to commercialisation scheme. Similarly, we have also patented a device for the dynamic delivery of hormone replacement with our project partner DesignWorks. Here we are exploring whether dynamic delivery of steroids - that mimic the natural secretory patterns in healthy humans - will optimise therapeutic benefit, whilst reducing significant side effects associated with long-term steroid use. Building closed loop systems based upon sensing precursor hormones offers significant potential for a dramatic impact for people with adrenal dysfunction. Building on this experience we also foresee impact in the areas of cardiology (notably arrhythmias), dementia and diabetes where we have developing collaborations with internationally leading clinical partners.

Research arising in our Centre will present the opportunity for impact for other commercial sector R&D. For example, parameter evolution of a patient-specific models extracted from clinical data brings the ability to characterise the effect of a drug or neuromodulation device. Similarly, it may be possible to develop biomarkers for response to newly developed compounds in early-stage trials, based upon parameter evolution of a patient-specific model. Based on our track-record to date, there will also be opportunities for industries focussed on device development and technologies for aiding clinical decision making. Our approach to co-develop the research focus and design of methodologies will ensure the development of mathematical and statistical methodologies that are challenge led and readily translatable.