Stratifying Chronic Pain Patients By Pathological Mechanism- A Multimodal Investigation Using Functional MRI, Psychometric And Clinical Assessment

Approximately one person in five suffers with pain every day. Despite our best efforts we often struggle even to partly alleviate their pain. One reason is that current diagnoses are based upon symptoms, but the same symptoms can occur for different reasons. If we could diagnose pain based upon 'mechanisms', the biological processes underlying symptoms, then patients might get more precise treatment earlier.

Several mechanisms can be faulty in chronic pain. They can be 'peripheral'- when sensors in the body that detect danger persistently send messages via the spinal cord to the brain. Sometimes 'crossed-wires' occur in the spine, where messages representing touch become scrambled, resulting in an incorrect or amplified pain signal- 'central sensitisation'. These processes might be due to disease (e.g. osteoarthritis (OA)) or because of nerve damage-'neuropathy'. Brain imaging has told us that changes in the way parts of the brain communicate with one another -'connectivity'- predicts transition to chronic pain, but we don't know whether connectivity differences cause pain or are a 'knock-on' effect of dysfunction elsewhere; perhaps peripherally or in the cord. Finally, pain control systems in the base of the brain (brainstem)- the 'descending modulatory system'- can fail, producing symptoms similar to central sensitisation. One or all these mechanisms might be involved in patients with chronic pain. As the symptoms can look the same we don't currently know which.

This project aims to predict the mechanisms underlying individual patients with chronic pain of the face or upper limb. We will use computerised pattern recognition (PR) techniques to determine which combination of clinical assessments (examinations, interviews, questionnaires); specialised nervous system tests and brain/spine imaging techniques best detect the underlying pathophysiological mechanisms. Historically it has been difficult to get clear 'functional' images of the brainstem and spine during rest and stimulation but we now have new methods to help solve these problems.

First, we will use electrical stimulation in the arms of healthy, pain-free people to see how the periphery transmits a normal ongoing pain signal. By changing the characteristics of the stimulation we can also temporarily create central sensitisation in the spine. In the face, we can examine pain due to peripheral and central sensitisation after wisdom tooth surgery. We can use an anaesthetic injection to 'block' the peripheral signal to look at central sensitisation only in these patients. Rarely, but sometimes wisdom tooth surgery produces nerve damage, leading to chronic facial pain. We will also study these patients too, again using anaesthetic injections to look at the peripheral and central signals separately. We will also study patients with chronic arm pain due to OA. Historically OA was considered a 'peripheral' disease, but some patients may also have 'central' changes, which we will determine in the brain and spine. Finally, we will use a technique called 'Conditioned Pain Modulation' (CPM) to assess, in all patients and healthy people, how well their 'descending modulatory' pain control systems are working.

We will capitalise on all of these clinical data (imaging, examination, questionnaires) and use PR to develop distinct 'fingerprints' that classify peripheral and central pain mechanisms. We will apply the classifier in each chronic face and arm patient to make predictions about their individual underlying pathophysiology. Similarly, a second classifier will be trained to recognise 'normal' versus 'abnormal' descending modulatory pain control in each chronic pain patient. Success in this project will help us get the best treatment, more quickly to suit the needs of each patient in persistent pain. The new knowledge that we generate about how the brain and spine represent these mechanisms will also stimulate the development of much-needed new treatments.

Chronic pain remains an area of considerable unmet need. Up to 20% of the population suffer psychological distress and poor health status. Classification remains disease-based, despite evidence that multiple pathophysiological processes underpin diseases, and that patients with similar symptoms can have different aetiologies. As most current pharmacotherapies target only one mechanism, it is not perhaps surprising that many offer only moderate pain relief. Mechanism-based patient stratification should facilitate timely, appropriate and cost-effective treatments.

We will use multimodal techniques to stratify patients according to their underlying pain mechanisms, examining patients with intractable painful osteoarthritis, orofacial pain and pain-free controls. Evoked-response and resting-state functional and perfusion magnetic resonance imaging will be employed to investigate brain, brainstem and spinal cord structures underpinning mechanisms of ongoing peripheral drive, central sensitisation, descending pain modulation (DPM) and functional connectivity. We will decouple peripheral from central phenomena using anaesthetic blocks. Gold-standard subjective report, psychometric and sensory testing data will also be acquired. We will use multimodal machine learning across all data to predict the mechanisms underlying patients' persistent pain, namely: (i) does persistent pain originate peripherally, centrally or as a combined state? (ii) is the DPM system functioning? We will determine which data types contribute most to patient stratification, deriving simple, identifiable fingerprints of each pathophysiological state for clinical use.

Success in this project should impact on clinical practice. Diagnostic uncertainty is associated with poor outcomes, distress and suffering. Delivery of new mechanistic insights should catalyse treatment development; better identification of aetiologies should expedite implementation of treatments tailored to individual patients.

Who will benefit from this research and how?

Successful completion of this project will have identified a new way of stratifying patients according to the mechanisms underlying their persistent pain. We envisage that a wide-range of patients awaiting investigation of their pain or struggling to achieve effective pain relief will be the most likely beneficiaries. The impact of pain on individuals overall is immense. Miserable and debilitating in itself, it damages quality of life and long-term mental and physical well-being; over half of these patients will develop symptoms of depression. Pain costs jobs. 19% of people with chronic pain will eventually lose their job due an inability to work. The estimated cost to the UK economy for back pain alone is £12.3 billion per year. The last pan-European survey, completed almost ten years ago, suggested that one in every five people was suffering with chronic pain. Every indication suggests that this number is gradually increasing. We know that uncertain diagnoses of pain strongly predict a poor treatment outcome and that the longer pain persists, the more intractable to treatment it becomes. A mechanism-based approach should help sufferers get on the right treatment pathway as quickly as possible. This approach is very much aligned to research in psychiatry where 'RDoC (Research Domain Criteria) has been proposed as a new way of classifying mental health disorders based on behavioural dimensions and neurobiological measures, thus using the power of modern research to better categorise and treat patients.

We also envisage academic, commercial and healthcare communities may benefit from this work. At its simplest, healthcare cost efficiencies can be made by starting with the correct treatment. For example, in the treatment of low back pain, practically all patients are still considered to have pain of a peripheral origin, despite the fact that increasingly it is recognised as many as a third of those patients are likely to have a significant central nervous system component involved in their pain. Academics, clinicians and industries dedicated to the development of new treatments may also stand to gain. We will generate vital new insights into how the human central nervous system represents pain that will be informative to these groups of individuals. The optimised strategies for imaging brain and brainstem in particular will further our neurobiological understanding. Pattern classification methodologies are well-suited to drug development. Classifiers learn by example; here we used them to stratify patients but a real opportunity exists in late phase I/early phase II studies to use these technologies to refine inclusion criteria. Further, the pharmaceutical industry has already demonstrated their interest, not only to understand putative new analgesics and how they might work but also to provide an early indication of when drug development should be discontinued, or 'no-go' decision-making.