Background Although biological therapies have transformed the outlook for those with rheumatoid arthritis, there is a lack of any meaningful response in approximately 40% of patients. The role of B cells in rheumatoid arthritis pathogenesis is well recognised and is supported by the clinical efficacy of the B-cell-depleting agent rituximab (MabThera, F. Hoffman La-Roche Ltd, Basel, Switzerland). Rituximab is licensed for use in rheumatoid arthritis following failure of conventional synthetic disease-modifying antirheumatic drugs and tumour necrosis factor inhibitor therapy. However, over 50% of patients show low/absent synovial B-cell infiltration, suggesting that, in these patients, inflammation is driven by alternative cell types. This prompted us to test the hypothesis that, in synovial biopsy B-cell-poor patients, tocilizumab (RoActemra, F. Hoffman La-Roche Ltd, Basel, Switzerland) (targeting interleukin 6) is superior to rituximab (targeting CD20⁺/B cells). Design The R4–RA (A Randomised, open-labelled study in anti-TNFalpha inadequate responders to investigate the mechanisms for Response, Resistance to Rituximab versus Tocilizumab in Rheumatoid Arthritis patients) trial is a 48-week Phase IV, open-label, randomised controlled trial conducted in 19 European centres that recruited patients failing on or intolerant to conventional synthetic disease-modifying antirheumatic drug therapy and at least one tumour necrosis factor inhibitor. Participants Synovial tissue was obtained at trial entry and classified histologically as B-cell rich or B-cell poor to inform balanced stratification. Patients were randomised on a 1 : 1 basis to receive standard therapy with rituximab or tocilizumab. B-cell-poor/-rich molecular classification was also carried out. The study was powered to test the superiority of tocilizumab over rituximab at 16 weeks in the B-cell-poor population. Main outcome measures The primary end point was defined as an improvement in the Clinical Disease Activity Index (CDAI) score of ≥ 50% from baseline. In addition, patients were considered to be non-responders if they did not reach an improvement in CDAI score of ≥ 50% and a CDAI score of < 10.1, defined for simplicity as CDAI major treatment response (CDAI-MTR). Secondary outcomes included the assessment of CDAI response in the B-cell-rich cohort, in which the non-inferiority of rituximab compared with tocilizumab was evaluated. Safety data up to week 48 are reported. Results In total, 164 patients were randomised: 83 patients received rituximab and 81 received tocilizumab. Eighty-one out of 83 rituximab patients and 73 out of 81 tocilizumab patients completed treatment up to week 16 (primary end point). Baseline characteristics were comparable between the treatment groups. In the histologically classified B-cell-poor population (n = 79), no significant difference was observed in the primary outcome, an improvement in CDAI score of ≥ 50% from baseline (risk ratio 1.25, 95% confidence interval 0.80 to 1.96). A supplementary analysis of the CDAI-MTR, however, did reach statistical significance (risk ratio 1.96, 95% confidence interval 1.01 to 3.78). In addition, when B-cell-poor classification was determined molecularly, both the primary end point and the CDAI-MTR were statistically significant (risk ratio 1.72, 95% confidence interval 1.02 to 2.91, and risk ratio 4.12, 95% confidence interval 1.55 to 11.01, respectively). Moreover, a larger number of secondary end points achieved significance when classified molecularly than when classified histologically. In the B-cell-rich population, there was no significant difference between treatments in the majority of both primary and secondary end points. There were more adverse events and serious adverse events, such as infections, in the tocilizumab group than in the rituximab group. Conclusion To our knowledge, this is the first biopsy-based, multicentre, randomised controlled trial of rheumatoid arthritis. We were unable to demonstrate that tocilizumab was more effective than rituximab in patients with a B-cell-poor pathotype in our primary analysis. However, superiority was shown in most of the supplementary and secondary analyses using a molecular classification. These analyses overcame possible unavoidable weaknesses in our original study plan, in which the histological method of determining B-cell status may have misclassified some participants and our chosen primary outcome was insufficiently sensitive. Given the significant results observed using the molecular classification, future research will focus on refining this stratification method and evaluating its clinical utility. Trial registration Current Controlled Trials ISRCTN97443826. Funding This project was funded by the Efficacy and Mechanism Evaluation (EME) programme, a Medical Research Council and National Institute for Health and Care Research (NIHR) partnership. This will be published in full in Efficacy and Mechanism Evaluation; Vol. 9, No. 7. See the NIHR Journals Library website for further project information.