Plasma metabolomics reveals candidate biomarkers associated with individual treatment response in psoriatic arthritis

Psoriatic arthritis (PsA) is a clinically heterogeneous inflammatory disease for which optimal treatment selection remains unclear. Robust biomarkers that predict treatment response and identify patient-specific factors to inform treatment choice could advance precision medicine in PsA. We therefore investigated whether baseline plasma metabolomic profiles could predict response to tofacitinib, methotrexate, and etanercept in patients with PsA. Baseline plasma concentrations of 355 metabolites were measured in 80 patients with PsA from the TOFA-PREDICT trial using liquid chromatography–mass spectrometry (LC–MS) assays. Patients were randomized to tofacitinib or methotrexate (DMARD-naïve group), or to tofacitinib or etanercept (DMARD-failure group). Baseline metabolites predictive of response (MDA, minimal disease activity) at 16 weeks were selected using machine learning approaches. Multiple prediction models incorporating the selected metabolites and clinical variables were developed and evaluated using nested cross-validation. Longitudinal changes in the selected metabolites were additionally assessed. The health assessment questionnaire (HAQ) score and the metabolites 1-methylhistidine, 3-methylglutaric acid, deoxycarnitine, allantoin, cLPA(16:1), LPI(20:4), 5-HEPE, 8,12-iso-iPF2α-VI isoprostane, and sphinganine were selected as predictors. The best performing model was a heterogeneous treatment-effect model using principal component-based predictors, which achieved an AUC of 0.83. Predicted response probabilities varied substantially between treatments within individual patients. The selected metabolites represented distinct classes, including amino acids, (bioactive/signaling) lipids, and markers related to oxidative stress. In this exploratory analysis, a panel of baseline metabolites combined with HAQ predicted differential treatment response between treatment options in individual patients with PsA. This indicates that metabolomic markers may carry information relevant to differential treatment response. The identified candidate biomarkers support further investigation of metabolomics-informed precision medicine in PsA but need validation in independent cohorts. EU Clinical Trials, 2017–003900-28.