A TRL3-Stage Feasibility Study for a Prognostic and Predictive Biomarker System in Rheumatoid Arthritis

RA is one of several high-burden chronic inflammatory diseases (cIDs) affecting Northern Ireland (NI), the mainland UK, and Ireland. Current care pathways suffer from two major issues:

1. Late Action – RA diagnosis and treatment typically occur after the disease has fully developed.

2. Imprecision – Clinicians often rely on guesswork when selecting treatments for non-responding patients.

Both problems stem from a lack of effective chronic inflammation (cI) biomarkers. Current medical practice relies on markers of acute inflammation (aI), which are not reliable for guiding long-term care in RA or other cIDs.

ProVision-RA addresses these systemic issues by combining three sources of data to deliver high clinical utility:

1. Omics biomarkers – including glycomics patterns of immune system components

2. Conventional clinical markers

3. Digitally-acquired patient data

These are used to:

• Render disease-specific cI processes visible at early stages of disease progression and flare development

• Model the patient’s likely future health trajectory, enabling long-range prediction and personalisation

Background and Rationale

ProVision-RA is based on the validated design of ProVision-IBD, a multi-biomarker system for inflammatory bowel disease now at TRL6. Clinical studies of ProVision-IBD (at TRL5) demonstrated that the system enables early action to mitigate inflammation and provides reliable long-range prognoses. Preliminary data also showed glycomic markers can indicate likely response to specific biologic drugs.

ProVision-RA aims to transfer this capability to the RA setting, using RA-specific omics and clinical data.

This TRL3 study was conducted in Northern Ireland (NI) to answer key questions about ProVision-RA’s translation and deployment across NI, mainland UK, and Ireland. The project investigated whether ProVision-RA can:

a. Combine aetiologically relevant longitudinal patient data (glycomic, proteomic, genetic/epigenetic, and clinical) to build a long-range biomarker system for RA
b. Provide faster, more precise information to rheumatologists on patients’ disease states and likely treatment response
c. Enable earlier diagnosis and intervention, improving quality of life and reducing burden on families
d. Improve health outcomes and reduce healthcare costs
e. Cut down on drug waste and avoidable hospital visits

To support the development and validation of ProVision-RA, we carried out both primary and secondary research to assess the urgent unmet need in rheumatoid arthritis (RA) care across Northern Ireland, the UK, and Ireland. This included:

• Primary research:
  Interviews with RA patients, general practitioners, rheumatologists, and healthcare stakeholders across NI to understand diagnostic delays, treatment challenges, and system-level barriers.

• Secondary research:
  A structured literature review on RA burden, treatment failure rates, healthcare costs, and existing biomarker gaps in chronic inflammatory diseases (cIDs).

These assessments confirmed two systemic issues in RA care:

1. Late action, where patients are diagnosed and treated after full disease manifestation

2. Imprecision, where clinicians rely on trial-and-error to find an effective treatment for non-responders

To quantify the potential value of ProVision-RA, we developed a health economic model called Chadwich-RA, which evaluated the cost-effectiveness and impact of deploying the biomarker system in routine care. Key outputs included:

• Projected savings of £82,094 per 100 RA patients over 6 months

• Earlier identification of non-responders to anti-TNF therapies with up to 89% accuracy at baseline

• Reduction in drug waste and avoidable hospital visits

• Increased opportunity for clinicians to intervene at the early prodromal stage of flare development

Biomarker Algorithms and Performance

The ProVision-RA study tested six algorithmic models for predicting treatment response using different combinations of data sources. Results are summarised below:

Algorithm Performance

Combined 1 & 3 (glycomics + clinical data) offers the best overall performance, combining high accuracy with balanced sensitivity and specificity—allowing for early identification of responders and non-responders alike.

Although full clinical utility can only be established through a larger, independent cohort study, our TRL3 data show that glycomic-only and glycomic-integrated approaches outperform both standard care and clinical data alone.

These findings support the strong predictive potential of ProVision-RA, particularly when leveraging glycomics data in conjunction with clinical variables such as pain scoring, age, and sex. In our models, the AUC reached 0.83 for predicting response to Infliximab, and glycomics showed excellent p-values when confirming response at the 6-month timepoint.