Inflammatory Bowel Disease (IBD) is a complex, chronic condition characterised by unpredictable periods of remission and relapse. Identifying which patients are at high risk for disease flares or severe complications is a major clinical challenge.
Our group is developing enhanced patient stratification methods to support precision medicine in IBD. By leveraging longitudinal data and advanced statistical modelling, we aim to provide clinicians with dynamic risk prediction tools that can adapt as a patient’s clinical profile changes over time.
This work represents a long-standing collaboration with Prof Charlie Lees
A core foundation of our research relies on integrating routinely collected real-world data to build robust, scalable stratification models. By utilising the Lothian IBD Registry, describing nearly all IBD patients receiving secondary care from NHS Lothian, we incorporate deeply phenotyped Electronic Health Records (EHR) into our predictive frameworks.
By analysing repeated biomarker measurements, our statistical models can capture the dynamic nature of the disease. This moves our predictive capabilities beyond baseline-only assessments, offering a continuously updated view of a patient’s risk profile as their disease progresses.
While local EHRs provide a macro-level view of patient pathways, we also investigate the granular trajectory of IBD through highly phenotyped datasets. We work extensively with data from the PREdiCCt study, a major prospective cohort designed to investigate how environmental factors, diet, the gut microbiome, and biomarkers influence IBD outcomes.
To translate longitudinal data into actionable clinical insights, we employ survival analysis techniques, specifically focusing on landmarking. This approach allows us to update survival and risk predictions as new clinical information becomes available, providing a real-time assessment of the risk of disease flare or the need for therapy escalation.
To facilitate these analyses, our team actively develops open-source software, including landmaRk
Group members that have contributed to this project: