Pain is the most prominent and burdensome symptom reported by people with Rheumatoid arthritis (RA). Although modern biologic and targeted synthetic disease-modifying antirheumatic drugs can achieve effective control of peripheral inflammation, an estimated 20–30% of patients continue to experience moderate-to-severe chronic pain even after reaching clinical remission. This persistent pain suggests a transition of mechanisms from peripheral nociception toward central sensitisation and nociplastic pain, processes that are less likely to respond to peripheral anti-inflammatory therapies.
These observations motivate an integrative approach to patient stratification that combines clinical phenotyping, neurophysiological measures and high-dimensional molecular profiling. The RA-PAIN-AI study is designed to discover and characterise signatures associated with chronic pain in RA by applying integrative artificial intelligence (AI) and machine-learning analytical methods to combined clinical, neurophysiological and multiomic datasets.
Primary aims include identifying clinical, neurophysiological and multiomic features that associate with persistent pain in RA and developing integrative signatures that may support patient stratification. Secondary aims include exploring longitudinal changes over a 3–6 month interval and conducting feasibility evaluation of exploratory neurophysiological measures in a subset of participants.
RA-PAIN-AI is a prospective, observational, single-centre case–control study. Two adult RA groups will be enrolled:
Eligibility criteria for the exploratory EEG subset are predefined and include feasibility and safety considerations; EEG will not be performed in all participants but only in those meeting those subset criteria.
All participants will undergo comprehensive clinical assessment at baseline. Validated patient-reported outcome measures will be collected to characterise symptom burden, including pain severity and related domains. Questionnaires will be repeated at the 3–6 month follow-up visit to capture change over time. The exact interval between baseline and follow-up visits will be recorded for inclusion in longitudinal analyses.
Peripheral blood samples will be collected at baseline for all participants and again at the 3–6 month visit. Biological analyses planned for the study include:
Laboratory and bioinformatic workflows are intended to integrate these data modalities for downstream AI-driven analyses.
Electroencephalography will be undertaken only in an exploratory subset of participants selected according to predefined feasibility, safety and eligibility criteria. The inclusion of EEG aims to probe central nervous system correlates of pain and potential signatures of central sensitisation. Because EEG is exploratory, its deployment is limited and findings will inform feasibility and future hypothesis-driven studies.
Recognising the high-dimensional and exploratory nature of the dataset, planned analytical methods include prespecified feature filtering, dimensionality reduction and regularisation techniques. Machine-learning and AI approaches will use internal validation procedures to assess model performance. The protocol emphasises that any predictive models developed from this study will require external validation in independent cohorts before clinical application.
A minimum of 50 participants per group will be recruited. Baseline data and biological samples will be collected from all participants, with repeat questionnaires and peripheral blood collection at a 3–6 month interval. The specific timing of follow-up visits will be recorded and incorporated into longitudinal analyses when appropriate.
The study protocol received a favourable opinion from the Bioethics Committee at the Hirszfeld Institute of Immunology and Experimental Therapy, Polish Academy of Sciences, Wroclaw, Poland (Opinion No. KB-17/2025, dated 25 September 2025). Written informed consent will be obtained from all participants prior to study procedures. Planned dissemination routes include peer-reviewed publications, conference presentations and communication with scientific, clinical and patient communities.
This study is registered with the identifier NCT07769619.