Objectives . To analyse the effect of a dose increase in patients with severe rheumatoid arthritis (RA) with insufficient clinical response to 3 mg/kg infliximab every 8 weeks. Methods . Patients suffering from active refractory RA despite methotrexate, were treated with i.v. infusions of infliximab (3 mg/kg) on week 0, 2, 6 and every 8 weeks thereafter. Based on the clinical judgement at week 22, patients received a dose increase of 100 mg from week 30 on. The American College of Rheumatology (ACR) core set for disease activity measures was regularly assessed. Results . Five hundred and eleven RA patients were included. At week 22, 61.4, 34 and 14.1% of all patients met ACR 20, ACR 50 and ACR 70 criteria, respectively, and 6.1% of patients were in remission. A low swollen joint count at baseline was correlated with improvement at week 22 for ACR 20 ( P <0.06), ACR 50 ( P <0.06) and ACR 70 ( P <0.005). The change in HAQ score between weeks 0 and 22 was predictive for response at week 54 ( P <0.01). The dose of infliximab was increased by 100 mg in 22% of the patients. Most baseline values of patients requiring dose increase were higher ( P ≤ 0.001) than the baseline values of the remaining patients. Increasing the dose of infliximab by one vial from week 30 on could circumvent the partial loss of response in these patients. Conclusion . Infliximab use in this large out-patient cohort resulted in a significant clinical improvement. A subgroup that partially lost response during the first 22 weeks could regain response by adding 100 mg of infliximab to the subsequent doses. Due to the current study design, however, a regression to the mean like effect could not be ruled out.