Research into AI model for reducing rheumatism medication

The drug use of rheumatism patients is usually phased out once the disease process has been brought under control. There is always the danger of a flare-up of the complaints, after which the use of medication has to be increased again. If it becomes possible to predict such a flare-up with the help of an AI model, and to tailor the phasing out of drug use accordingly, that would of course be good news for rheumatism patients. At the moment it is still very difficult to predict a flare-up. As a result, both doctors and patients are often reluctant to taper medication.

AI model of the UMC Utrecht

From development of an AI model to reduce the risk of a flare-up was carried out by a multidisciplinary team at UMC Utrecht. The developed computer model uses digitally available data from the patient. This includes previous blood results, medication use and disease activity. “Our computer model predicts whether it is possible to reduce the biological in a responsible manner without the disease flare-up,” explains Marianne Messelink, medical researcher at the Department of Rheumatology and Clinical Immunology.

Messeling is also doing PhD research into the reduction of biologicals. “With the support of a subsidy from ZonMw, we want to test whether the use of the developed computer model in clinical practice can actually reduce the number of flare-ups during the phase-out of biologicals. That is why we have set up the PATIO study together with the Sint Maartenskliniek in Nijmegen,” says the doctor-researcher.

PATIO investigation

In this study, where PATIO stands for Prediction Aided Tapering In RA patients treated with biOlogicals, 60 RA patients are followed for 18 months. Blood is drawn every three months, followed by a visit to the rheumatology outpatient clinic. Then it is checked whether the biological can be further reduced. The participants in the study are divided into two groups.

In addition to the conventional method, the developed AI model is also used in the second group. When the model predicts that there is a high risk of a flare-up, no further reductions are made. The results of the study are expected in 2024. It should then become clear whether the predictive model has added value. In other words, there were fewer flare-ups in the group where the model was used.