Skip to navigation Skip to main content Skip to footer

Approved research

Deep phenotyping of patients with a range of chronic pain conditions in the UK Biobank using a machine-learning approach.

Principal Investigator: Dr Andrew Segerdahl
Approved Research ID: 45465
Approval date: July 29th 2019

Lay summary

Chronic pain is a common and disabling problem which affects up to half of adults in the UK. The current treatments for chronic pain are often not very effective. This is partly because we do not fully understand how the pain is generated. We know that part of the problem is due to how pain signals are processed in the brain. By doing brain scans on patients with chronic pain, we can see how the different areas of the brain react to pain and how they are connected to one another. Studies involving brain scans in patients with chronic pain usually only involve a small number of patients. This means that we cannot be completely sure that what has been found so far is definitely correct. In addition we might have missed some important differences between patients that might exist even if they have been diagnosed with the same painful condition. The UK biobank study aims to image 100,000 volunteers by 2020. Of these participants, a significant percentage are likely to have one of the painful conditions that we would like to study. This means that we will be able to conduct a very powerful study to investigate the how the brain reacts to pain. We will also use a technique called 'machine-learning' to help analyse the data. This approach uses computer programming to find new patterns in the data which we may not otherwise be aware of. In doing so, we may uncover a better understanding of how the brain is working in people with chronic pain and we hope that this will help to shape future research looking at the best ways to help improve the quality of life for these people.

Scope extension:

Analysis 1 aims to identify brain-based features observed during the first scan session (when the participant is not suffering from pain) that predict the onset of chronic pain at the follow-up appointment (up to 3 years later). To do this, we will utilize the questionnaire data collected at baseline and during the online follow-up (category 153) to select participants who were pain-free at baseline but subsequently developed chronic pain. We hypothesize that altered connectivity between key pain regions will accurately discriminate patient groupings similar to what has been shown previously with chronic low back pain patients (13).

Analysis 2 aims to identify specific brain-based features that can reliably classify different subgroups of chronic pain patients compared to age and sex matched pain-free controls. To do this, we will utilize FMRI data collected from participants that identified themselves as suffering from any of the following conditions during the first scan session: Migraine, Irritable bowel syndrome, Inflammatory arthritis (Rheumatoid arthritis and Psoriatic arthritis), Osteroarthritis, Chronic widespread pain (general pain for 3 months or more). We hypothesize that brain-based features will allow precise classification of patients even if the clinical questionnaire data suggests there is a similar symptomatology.

In addition to the above, analysis 1 will be enhanced by utilising the more detailed pain questionnaire data which was collected via a further online follow-up assessment (Category 154). For analyses 1 and 2, we will also investigate the demographic, biopsychosocial and genotypic features which may contribute to the development and presentation of pain. Finally, an additional analysis 3 aims to establish the epidemiology of cognitive impairment in people with chronic pain, such as fibromyalgia syndrome, the underlying mechanisms and mediators in their development, including depression and sleep impairment, and the role of brain structural and functional changes. The project will be ongoing until the full amount of neuroimaging data has been released.