Variability of the metabolomic profiles in terms of potential confounding variables in the four principal components identified by PCA and expressed as F-values together with the corresponding p-values (in brackets).

ANOVA test, significant values (pSmoking, co-morbidity, ACPA and RF variables are defined as presence or absence; dietary habits variable is defined as mediterranean diet or non-mediterranean diet; age and disease duration variables are defined as numerical values (number of months); concomitant treatment variable is defined as four groups: glucocorticoids, DMARDs, DMARDs and glucocorticoids and no immunosuppressants.Among subject-related and disease-related factors, co-morbidity and RF, respectively, affected the clustering of metabolic profiles in terms of a minor likelihood of response to treatment in patients positive for either of them. However, by using Pearson’s simple and partial correlation analysis to simultaneously investigate PCA score, responder or non-responder status, and co-morbidities, the metabolic variations on PC1 were mainly associated with the prediction of response to etanercept.ACPA: anti-citrullinated protein/peptide antibodies; RF: rheumatoid factor.Variability of the metabolomic profiles in terms of potential confounding variables in the four principal components identified by PCA and expressed as F-values together with the corresponding p-values (in brackets).