Exploring the Potential of Patient-Reported Daily Symptoms to Predict Probability of Treatment Response to Disease-Modifying Anti-Rheumatic Drugs in Rheumatoid Arthritis

Rheumatoid arthritis (RA) is a long-term disease in which the body’s immune system mistakenly attacks its own joints. This leads to pain, stiffness, swelling, and fatigue. Without early treatment, RA can cause permanent joint damage. More than 18 million people worldwide live with RA.

When doctors diagnose RA, they usually prescribe medicines called Disease-Modifying Anti-Rheumatic Drugs (DMARDs). These medicines can slow or stop joint damage by helping the immune system to calm down. Doctors check how well treatment is working during clinic visits using tests such as the Disease Activity Score for 28 joints (DAS28). However, these visits often happen weeks or months apart. Because of this, it may take a long time to know if a treatment is not working. This delay can mean that some patients remain on ineffective treatments for months before doctors make changes.

Patients often notice changes in their pain, fatigue, or stiffness earlier than doctors can detect changes during clinic visits. In large clinical studies called randomized controlled trials (RCTs), patients are given medication, and their symptoms are observed and recorded. When results from previous trials were averaged across all patients, these symptom reports showed clear improvement within a few days of starting DMARD treatment. What is not yet known is whether these daily reports can help identify different groups of patients, such as those who improve quickly, those who improve slowly, or those who do not respond at all.

In our project, we will analyze the daily symptom data collected from the RCTs. We will use computer-based methods to group patients according to how their symptoms change over time. We will then compare these groups to standard clinical measures such as DAS28 to see how closely they match. We will also test whether we can predict early in treatment which group each patient is likely to belong to, by using a model that updates its predictions as new daily symptom information is added.

This study is possible because participants in RCTs are more likely to take their medicines as directed and consistently record their symptoms, which makes it easier to link changes in symptoms directly to the treatment being studied. By using a model that updates predictions over time, we can better capture how accuracy increases with each additional day of information.

If successful, our research will help doctors identify patients who are not responding to treatment much earlier than is currently possible. This could allow doctors to switch to a different treatment sooner, reducing delays and improving outcomes. Ultimately, this approach may support more personalized care and help match each patient with the right treatment at the right time.