A precision health approach to osteoarthritis: early diagnosis, stratification and risk prediction using state-of-the-art machine learning models.

The research targets osteoarthritis (OA), a disabling MSK condition affecting over 500 million people globally, with more than 10 million cases reported in the UK alone. With an ageing population, these numbers are likely to rise, posing great challenges to healthcare systems worldwide. As there is presently no approved treatment to prevent or reverse disease progression, the primary focus of current medical practices is on symptom management and, eventually, joint replacement.

The main aims and objectives of the research study.

The overall aim of the research is to develop novel precision health tools for the early diagnosis, progression prediction, stratification and, eventually, management of OA. The project plans to harness the power of cutting-edge machine learning methodologies to achieve this. These tools will pave the way for the design of streamlined clinical trials, the identification of novel treatment targets, and ultimately, the creation of efficacious therapies that can halt or even reverse OA progression.

How this research is going to help address MSK health.

The research focuses on enhancing the management of OA, aiming for early diagnosis and tailored treatment strategies to significantly improve patient quality of life. The precision health tools will be designed to assist clinicians in optimising patient care, leading to better outcomes and potential healthcare cost reductions through more efficient OA management. The long-term advantages extend to improved patient care through personalised treatments, increased clinical trial efficiency, discovery of new therapeutic targets, and overall cost savings from early interventions.

The main research methods, or datasets being used.

The main research methods in the project involve collecting data from publicly available datasets such as the Osteoarthritis Initiative (OAI) or the UK Biobank, which offer a wealth of clinical, radiological, biological, and accelerometry data. Subsequently, the project aims to harness the advanced machine learning tools developed by the ‘van der Schaar lab’ to construct predictive models based on these comprehensive datasets. A key focus of the research is ensuring the fairness and transparency of these models. Ultimately, the project intends to conduct local prospective studies with real patients to validate the models in a clinical setting. These studies are crucial in evaluating the potential clinical utility of the tools, ensuring their effectiveness and applicability across diverse patient groups.

 Research results generated.

The preliminary models demonstrated high reliability in predicting OA progression using clinical, X-ray, MRI and biochemical marker data. Patient-reported outcomes and MRI features emerged as primary predictors of progression. To enhance clinical utility, we developed web applications to provide intuitive visualisation of personalised predictions.

Next steps of this research project.

The researchers plan to externally validate and refine their models by integrating additional data, such as movement data from wearable accelerometers and -omics data.

They also plan to adapt their methodology to diagnose and predict progression of other MSK conditions as well as other complex chronic non-MSK diseases.

Publications and presentations related to this fellowship.

Conferences:

• EORS (European Orthopaedic Research Society) 2023: (Castagno et al., 2024).
• OARSI (Osteoarthritis Research Society International) 2024.

Castagno, S., Birch, M., Schaar, M. van der, McCaskie, A., 2024. A precision health approach for osteoarthritis: prediction of rapid knee osteoarthritis progression using automated machine learning. Orthopaedic Proceedings 106-B, 19–19. https://doi.org/10.1302/1358-992X.2024.2.019

Awards or recognition received related to this project.

EORS2023 New Investigator Best Clinical Presentation Award

Engagement with research users, special interest groups and the general public to inform them about the research.

The researchers have forged a relationship with the PPI team at Addenbrooke’s Hospital, Cambridge. Recently, they convened a focus group comprising both OA and non-OA patients to discuss their research. This dialogue underscored the urgent need for innovative tools to enhance OA management through early diagnosis and patient stratification. It also emphasised the critical role of public involvement in the early stages of the research, particularly when developing tools, such as our web-based app, that could be directly utilised by the public. The conversation brought to light potential concerns about Machine Learning, including the risk of bias in the data used to train our models and the opacity of “black-box’ models that are not easily understood or interpreted. We are committed to maintaining an ongoing dialogue with this group throughout our research process. We plan to hold regular focus group meetings to discuss our research, solicit feedback, and ensure our work remains patient-centric.

Researcher: Dr Simone Castagno.

Supervisors: Prof Andrew McCaskie (Professor of Trauma and Orthopaedic Surgery and Head of the Department of Surgery), Prof Mihaela van der Schaar (founder of the Cambridge Centre for AI in Medicine) and Dr Mark Birch (Cambridge Stem Cell Institute).

University or Trust: University of Cambridge.

Award stream: ORUK/Versus Arthritis AI in MSK Research Fellowship.

Award duration: 2 years.

Amount rewarded: £100,000.

Other funders: Versus Arthritis.

Collaborations/ partners: Dr Kirsten Rennie (MRC Epidemiology Unit) and Prof Stephen McDonnell (Associate Professor at the University of Cambridge and Consultant Orthopaedic Surgeon at Addenbrooke’s Hospital in Cambridge).