Integrative AI-Based Multiomic and Neurophysiological Profiling of Chronic Pain in Rheumatoid Arthritis

Overview

Rheumatoid arthritis (RA) is a chronic, systemic autoimmune disease in which pain remains the most prominent and burdensome symptom from the patient's perspective. Although the introduction of biological and targeted synthetic disease-modifying antirheumatic drugs (bDMARDs and tsDMARDs) has significantly improved the control of inflammatory activity, an estimated 20-30% of patients continues to experience persistent moderate-to-severe pain despite achieving clinical remission. This discordance between objective inflammatory markers and subjective pain perception reflects an important unmet clinical need and suggests that chronic pain in RA may become partially or fully independent of peripheral inflammation, driven instead by central sensitization and nociplastic mechanisms. The RA-PAIN-AI study is a prospective, observational, case-control study integrating clinical assessment, patient-reported outcome measures, neurophysiological evaluation, and multiomic profiling, analyzed using artificial intelligence (AI)-based methods. Two groups of adult patients with RA are enrolled: a study group of patients with active disease qualifying for biological therapy under the Polish national drug program B.33, and a control group of patients in sustained clinical remission (at least 1.5 years) under biological therapy, without clinically significant chronic pain. All participants undergo clinical assessment, standardized questionnaires, peripheral blood collection at the baseline (Visit 1) and follow-up (Visit 2) visits. Electroencephalography (EEG) is offered as an optional procedure, at the discretion of the investigator. Biological samples are analyzed using advanced multiomic technologies, including whole genome sequencing (genomics), mRNA expression profiling (transcriptomics), and measurement of circulating proteins (secretomics). All clinical, biological, and neurophysiological data are then integrated using AI methods, including dimensionality reduction, clustering, and supervised machine learning, to identify distinct pain-related patient subtypes (endotypes). The primary objective of the study is to identify and characterize clinical, neurophysiological, and multiomic signatures associated with chronic pain in RA. Expected outcomes include identification of biomarkers predictive of chronic pain persistence despite effective anti-inflammatory treatment, improved differentiation between inflammation-driven pain and pain mediated by central nervous system changes, and support for the development of more personalized, mechanism-based treatment strategies in rheumatology. The study is non-interventional: no experimental therapeutic interventions are administered, and all participants continue to receive standard clinical care. All biological analyses are performed ex vivo.

Background and rationale. Chronic pain in rheumatoid arthritis (RA) represents a complex, multidimensional clinical phenomenon extending beyond the traditional paradigm of inflammation-driven nociception. Despite significant therapeutic advances enabling effective control of peripheral inflammation with biological and targeted synthetic disease-modifying antirheumatic drugs (bDMARDs and tsDMARDs), an estimated 20-30% of patients continues to experience persistent moderate-to-severe pain, even after achieving clinical remission. This therapeutic gap highlights the evolution of pain mechanisms from peripheral nociception toward central sensitization and nociplastic pain. Evidence from quantitative sensory testing (QST) studies demonstrates widespread pressure pain threshold reduction at sites remote from inflamed joints, indicating generalized central sensitization. Conditioned pain modulation (CPM) protocols reveal impaired descending inhibitory pathways. Functional magnetic resonance imaging (fMRI) studies show pathological hyperconnectivity between the default mode network and the insula, correlating with pain centralization independently of inflammatory markers. At the molecular level, transcriptomic analyses of human dorsal root ganglia have revealed pathological remodeling of primary sensory neurons, including ectopic immunoglobulin signaling and neurogenesis. Metabolomic and lipidomic profiling demonstrates hypoxia-driven glycolytic reprogramming, lactate accumulation acting as an independent algogen through acid-sensing ion channels, and glycerophospholipid metabolism disruption impairing production of specialized pro-resolving mediators. Integrative multi-omic studies have further identified distinct immunometabolic signatures between ACPA-positive and ACPA-negative RA, suggesting serotype-dependent pain mechanisms. Study design. The RA-PAIN-AI study is designed as a prospective, observational, case-control study integrating clinical, laboratory, neurophysiological, and multiomic data. Two main groups are enrolled: a study group of patients with confirmed RA in an active phase of the disease, qualified for biological therapy according to the Polish national drug program B.33, and a control group of patients with RA in long-term remission (≥1.5 years) under biological therapy, without clinically significant chronic pain. Non-interventional nature. The study is non-interventional. No experimental therapeutic interventions are administered as part of this study, and all biological analyses are performed ex vivo. All participants continue to receive standard clinical care according to routine rheumatology practice; study procedures are aligned with routine clinical visits whenever possible. Participation is associated with minimal risk. The primary procedures involving participants include venous blood collection, questionnaire assessments, and, optionally (at the discretion of the investigator), electroencephalography (EEG). The risks associated with these procedures are comparable to those encountered in routine clinical practice. Study procedures - Visit 1 (Baseline, Day 0). All participants undergo: verification of eligibility criteria; collection of informed consent; clinical assessment and medical history; completion of standardized questionnaires (EQ-5D-5L, PHQ-9, VAS, DN4, painDETECT, HAQ, SF-36, FAS, ISI, HADS, GAD-7, CSQ, RS-25); peripheral blood collection (approximately 30 mL total). EEG assessment is performed in selected participants, at the discretion of the investigator. Study procedures - Visit 2 (Follow-up, 3-6 months after baseline). Participants undergo repeat peripheral blood collection (approximately 10 mL), completion of standardized questionnaires (EQ-5D-5L, PHQ-9, VAS, DN4, painDETECT, HAQ, SF-36, FAS, ISI, HADS, GAD-7, CSQ, RS-25) and, in participants who underwent EEG at Visit 1, repeat EEG assessment, at the discretion of the investigator. Key assessment tools. The study integrates three complementary assessment domains: (i) patient-reported outcome measures (PROMs) covering pain intensity and character (VAS, DN4, painDETECT), functional status (HAQ), health-related quality of life (EQ-5D-5L, SF-36), psychological status (PHQ-9, HADS, GAD-7, CSQ, RS-25), and sleep quality (ISI), and fatigue (FAS); (ii)optional neurophysiological assessment by EEG (in selected participants, at the discretion of the investigator); and (iii) biological samples (peripheral blood) analyzed for multiomic and laboratory profiles, including whole genome sequencing (genomics), mRNA expression profiling (transcriptomics), and circulating protein measurement (secretomics). Study setting. The study is conducted at the 4th Military Clinical Hospital with Polyclinic (4WSzKzP SPZOZ) in Wrocław, Poland, a tertiary referral center. Laboratory and multiomic analyses are conducted in collaboration with the Institute of Immunology and Experimental Therapy, Polish Academy of Sciences (IITD PAN; responsible for genomic and transcriptomic analyses) and the Biobank and Research Group at Łukasiewicz - PORT Polish Center for Technology Development (blood processing for genomic analysis and long-term biobanking). All procedures are conducted in accordance with Good Clinical Practice (GCP) guidelines and applicable national regulations. Analysis framework. All clinical, biological, and neurophysiological data will be analyzed using both conventional statistical methods and artificial intelligence (AI)-based approaches. Statistical analyses will include descriptive statistics, between-group comparisons, correlation analyses, and multivariable analyses, as appropriate. In addition, AI methods, including dimensionality reduction, clustering, and supervised machine learning, will be used to identify distinct pain-related endotypes, develop predictive models for chronic pain persistence and treatment response, and support the transition from inflammation-centered management toward mechanism-based, personalized pain treatment in RA. The study also aims to create a structured, high-quality dataset and biobank resource to support future research in precision medicine and chronic pain in RA.