Research Results: Machine-learning and JIA 

Dr Stephanie Shoop-Worrall discusses her latest research from the CLUSTER consortium (which includes data from the UK JIA Biologics Register). This research looks at how Artificial Intelligence (AI) can help determine which treatment will work best for children and young people with JIA.  

Background: 

Methotrexate (MTX) is a common first-choice treatment for young people with juvenile idiopathic arthritis (JIA). However, it only works well for about half of them.  

To improve treatment decisions, researchers in CLUSTER used artificial intelligence (AI) to find groups of young people who experienced different patterns in their disease and its impact after taking MTX. They also wanted to see how these AI-based patterns compared to traditional ways of measuring treatment ‘success’. 

What they did: 

The researchers studied children and young people who started MTX before 2018 using data from multiple hospitals in the UK. They tracked key disease impacts over a year: joint swelling, how well the doctor thought the young person was, and wellbeing of the young person and a blood marker of inflammation . They used AI to group patients into six response types, based on how their condition and its impact changed over time. 

What they found: 

The six response types were: Fast Improvers (key impacts all got better six months after starting MTX), Slow Improvers (key impacts took a year to get better), Improve-Relapse (some improvement then worsening), Persistent Disease (little to no improvement), Persistent Doctor Concern (young person felt better, but doctors still thought the disease could be better controlled), and Persistent Parent Concern (doctor thought disease looked well controlled, young person still had issues like pain and issues doing everyday tasks).  

• Factors like age, ethnicity, and initial disease severity influenced which group patients fell into.  

• Traditional scoring methods couldn’t fully capture how different patients improved or relapsed over time. They also couldn’t pick out young people whose disease and its impacts didn’t all improve at the same rate 

Why it matters: 

This AI-based approach shows that the standard way of judging treatment success (a simple yes/no measure) is too limited. Instead, recognizing distinct response patterns could help doctors personalize treatment plans for children with JIA. This would make sure each young person gets the best possible care, tailored to which part of their disease and its impact are causing issues. 

Should you wish to read this scientific paper in full, the text can be found online here: https://pubmed.ncbi.nlm.nih.gov/38194741/ 

Related