Serum metabolic signatures are associated with anti-drug antibody development in rheumatoid arthritis patients treated with adalimumab

Objectives Development of anti-drug antibodies (ADAs) is a barrier to long-term efficacy of biologic therapies in rheumatoid arthritis (RA), but no biomarkers exist to predict ADA formation. This study explored the potential of serum metabolomics to predict development of ADAs to adalimumab in patients with RA. Methods: Serum from patients with RA (n=47), treatment naive for tumour necrosis factor-alpha inhibitor therapy, were collected before, Month(M)1 and M12 following initiation of adalimumab therapy as standard of care. Sera were tested for ADAs and patients were stratified according to M12 ADA status (ADA-positive n=21; ADA-negative n=26). Serum metabolomics was performed using an NMR-based platform. Metabolomic and clinical data were analysed using machine learning (ML) to develop a signature associated with ADA development. Results: ML analysis of baseline serum metabolomics and clinical data identified a signature that distinguished patients according to their future M12 ADA status (ADA-positive/ADA-negative) prior to first adalimumab treatment (area under the receiver operator curve, AUC-ROC=0.78), which out-performed clinical parameters alone (AUC-ROC=0.78). Metabolites related to cholesterol transport including large high and very low-density lipoproteins (L-HDL/VLDL) and small low density-lipoprotein (S-LDL) and clinical markers body mass index (BMI) and erythrocyte sedimentation rate were top discriminating features. Patients stratified as ADA-positive/ADA-negative at baseline also had different serum metabolic responses to adalimumab at M1 and M12. Finally, a putative predictive score for future ADA status was generated comprising L-HDL, L-LDL, extra-large VLDL subsets and BMI. Conclusion: These results support the potential of serum metabolomics as a predictive tool for immunogenicity risk in RA.