OBJECTIVE : To investigate ex-vivo drug-inhibited cytokine production before the start of a biological DMARD (bDMARD) as predictor of treatment response in rheumatoid arthritis (RA). METHODS : In a prospective RA cohort study [BIO-TOP], blood samples were obtained from patients before the start of a bDMARD (abatacept, adalimumab, etanercept, rituximab or tocilizumab). Peripheral blood mononuclear cells were pre-incubated for 1 hour with the therapeutic in-vivo concentration of the bDMARD and stimulated for 24 hours with heat-killed Candida albicans or Pam3Cys. Concentrations of IL-1β, IL-6, TNFα, IL-17 and IFNγ were determined by ELISA. EULAR response (good vs. moderate/no) was assessed at month 6. Area under the receiver operating characteristic curves (AUCs) were generated to evaluate the predictive value of baseline characteristics and ex-vivo cytokine production (including stimulated cytokine concentrations and absolute changes after inhibition by a bDMARD). Logistic prediction models were created to assess the added value of potential cytokine predictors. RESULTS : 277 RA patients were included with 330 blood samples. Good response was reached in 39% of the cases. DAS28-CRP was predictive for response to adalimumab (AUC 0.70, 95%CI 0.57-0.83), etanercept (AUC 0.68, 95%CI 0.58-0.78) and rituximab (AUC 0.76, 95%CI 0.65-0.86). ACPA was modestly predictive for response to abatacept (AUC 0.63, 95%CI 0.52-0.75). In the ex-vivo analysis, 4 of 64 (6%) tests showed some predictive value but these had no added value to clinical factors routinely measured in RA, such as DAS28-CRP. CONCLUSIONS : Ex-vivo inhibition of cytokine production by bDMARDs is unable to help prediction of treatment response to bDMARDs in RA.