During her summer ‘micro-internship,’ Ramlah Riza examined rheumatoid arthritis prescribing and how the current trial-and-error model can leave patients without appropriate treatment for long periods.
——
Despite an expanding range of treatments, finding the right therapy for the right RA patient can still involve sequential switching.
Part of the challenge is that RA isn’t biologically uniform. Patients can differ in their inflammatory pathways, molecular profiles, and disease characteristics, meaning the same treatment may produce very different responses.
This is where AI-powered immune profiling becomes interesting.
By analysing complex molecular and clinical data, machine-learning models may identify patterns associated with treatment response that conventional biomarkers alone cannot.
Early findings are promising. One model using synovial gene-expression signatures reported an AUC of 0.92 in predicting infliximab response.
But promising prediction isn’t the same as clinical utility.
These approaches still need robust prospective validation, integration with different patient data, and evidence that they can improve real-world treatment decisions.
The opportunity may therefore be less about replacing clinical judgment with AI, and more about giving clinicians better evidence to choose the right treatment earlier.
Could immune profiling ultimately help shift RA from trial-and-error towards more personalised treatment selection?