To assess the efficacy and safety of tocilizumab (TCZ) plus methotrexate/placebo (MTX/PBO) over 2 years and the course of disease activity in patients who discontinued TCZ due to sustained remission. ACT-RAY was a double-blind 3-year trial. Patients with active rheumatoid arthritis despite MTX were randomised to add TCZ to ongoing MTX (add-on strategy) or switch to TCZ plus PBO (switch strategy). Using a treat-to-target approach, open-label conventional synthetic disease-modifying antirheumatic drugs (csDMARDs), other than MTX, were added from week 24 if Disease Activity Score in 28 joints based on erythrocyte sedimentation rate (DAS28-ESR) >3.2. Between weeks 52 and 104, patients in sustained clinical remission (DAS28-ESR <2.6 at two consecutive visits 12 weeks apart) discontinued TCZ and were assessed every 4 weeks for 1 year. If sustained remission was maintained, added csDMARDs, then MTX/PBO, were discontinued. Of the 556 randomised patients, 76% completed year 2. Of patients entering year 2, 50.4% discontinued TCZ after achieving sustained remission and 5.9% achieved drug-free remission. Most patients who discontinued TCZ (84.0%) had a subsequent flare, but responded well to TCZ reintroduction. Despite many patients temporarily stopping TCZ, radiographic progression was minimal, with differences favouring add-on treatment. Rates of serious adverse events and serious infections per 100 patient-years were 12.2 and 4.4 in add-on and 15.0 and 3.7 in switch patients. In patients with normal baseline values, alanine aminotransferase elevations >3×upper limit of normal were more frequent in add-on (14.3%) versus switch patients (5.4%). Treat-to-target strategies could be successfully implemented with TCZ to achieve sustained remission, after which TCZ was stopped. Biologic-free remission was maintained for about 3 months, but most patients eventually flared. TCZ restart led to rapid improvement. NCT00810199.