Predicting responses to anti-TNFα therapy in patients with rheumatoid arthritis using metabolomic analysis of urine

Background: Anti-TNFα therapies are highly effective in the treatment of rheumatoid arthritis (RA) but a significant proportion of patients have an inadequate response. Given the important role of TNFα in regulating systemic and localized metabolism, we sought to determine if the metabolic profile of patients prior to therapy could be used to predict responses to anti-TNFα agents. Methods: Urine was collected from 16 patients with RA before and during therapy with infliximab or etanercept as part of a multicentre study. All patients were female and the mean age was 51.5. 14 patients were positive for rheumatoid factor and 14 for the anti-CCP antibody. All patients had a DAS28>4 at baseline. Urine metabolic profiles were assessed using NMR spectroscopy. The relationship between metabolic profiles and clinical outcomes was assessed (using partial least square discriminant analysis (PLSDA), Galgo and PLS-R analysis) and relevant metabolites were identified (using metabolite databases and Chenomix). Results: Baseline urine metabolic profiles were able to discriminate between RA patients who did (7 patients) or did not (9 patients) have a good response to anti-TNFα therapy according to EULAR criteria with a sensitivity of 85.9% and specificity of 85.7% with several metabolites (in particular citrate, creatinine and cresol) contributing. There was a significant correlation between baseline metabolic profiles in the urine samples and the extent of change in DAS 28 (PLS-R analysis p=0.04). In patients with RA who responded to TNFα antagonists, a good response to therapy was associated with changes in the following urinary metabolites: erythritol, phenylacetic acid, cresol, propionic acid, methylamine, citrate, hippuric acid and creatinine. Urine samples were also available for 20 patients with psoriatic arthritis (PsA). Similar metabolites were identified in the urine samples of the patients with PsA that responded to TNFα antagonists. We were unable to study the ability of baseline urinary metabolite profiles to predict response in PsA as all but one of the PsA patients responded according to predefined criteria. Conclusions: There are clear differences in the metabolic profiles of baseline urine samples of RA patients who go on to respond well to anti-TNFα therapy. This may be relevant in the development of clinically useful predictive strategies.