Objective To evaluate the interleukin-6 receptor inhibitor tocilizumab for the treatment of patients with polyarticular-course juvenile idiopathic arthritis (pcJIA). Methods This three-part, randomised, placebo-controlled, double-blind withdrawal study (NCT00988221) included patients who had active pcJIA for ≥6 months and inadequate responses to methotrexate. During part 1, patients received open-label tocilizumab every 4 weeks (8 or 10 mg/kg for body weight (BW) <30 kg; 8 mg/kg for BW ≥30 kg). At week 16, patients with ≥JIA-American College of Rheumatology (ACR) 30 improvement entered the 24-week, double-blind part 2 after randomisation 1:1 to placebo or tocilizumab (stratified by methotrexate and steroid background therapy) for evaluation of the primary end point: JIA flare, compared with week 16. Patients flaring or completing part 2 received open-label tocilizumab. Results In part 1, 188 patients received tocilizumab (<30 kg: 10 mg/kg (n=35) or 8 mg/kg (n=34); ≥30 kg: n=119). In part 2, 163 patients received tocilizumab (n=82) or placebo (n=81). JIA flare occurred in 48.1% of patients on placebo versus 25.6% continuing tocilizumab (difference in means adjusted for stratification: −0.21; 95% CI −0.35 to −0.08; p=0.0024). At the end of part 2, 64.6% and 45.1% of patients receiving tocilizumab had JIA-ACR70 and JIA-ACR90 responses, respectively. Rates/100 patient-years (PY) of adverse events (AEs) and serious AEs (SAEs) were 480 and 12.5, respectively; infections were the most common SAE (4.9/100 PY). Conclusions Tocilizumab treatment results in significant improvement, maintained over time, of pcJIA signs and symptoms and has a safety profile consistent with that for adults with rheumatoid arthritis. Trial registration number: NCT00988221.