ObjectiveTo assess factors associated with primary and secondary non-response to rituximab in systemic lupus erythematosus (SLE) and evaluate management of secondary non-depletion non-response (2NDNR).Methods125 patients with SLE treated with rituximab over 12 years were studied prospectively. A major clinical response was defined as improvement of all active British Isles Lupus Assessment Group (BILAG)-2004 domains to grade C/better and no A/B flare. Partial responders were defined by one persistent BILAG B. B-cell subsets were measured using highly sensitive flow cytometry. Patients with 2NDNR, defined by infusion reaction and defective depletion, were treated with ocrelizumab or ofatumumab.Results117 patients had evaluable data. In cycle 1 (C1), 96/117 (82%) achieved BILAG response (major=50%, partial=32%). In multivariable analysis, younger age (OR 0.97, 95% CI 0.94 to 1.00) and B-cell depletion at 6 weeks (OR 3.22, 95% CI 1.24 to 8.33) increased the odds of major response. Complete depletion was predicted by normal complement and lower pre-rituximab plasmablasts and was not associated with increased serious infection post-rituximab. Seventy-seven (with data on 72) C1 responders were retreated on clinical relapse. Of these, 61/72 (85%) responded in cycle 2 (C2). Of the 11 C2 non-responders, nine met 2NDNR criteria (incidence=12%) and tested positive for anti-rituximab antibodies. Lack of concomitant immunosuppressant and higher pre-rituximab plasmablasts predicted 2NDNR. Five were switched to ocrelizumab/ofatumumab, and all depleted and responded.ConclusionTreatment with anti-CD20 agents can be guided by B-cell monitoring and should aim to achieve complete depletion. 2NDNR is associated with anti-rituximab antibodies, and switching to humanised agents restores depletion and response. In SLE, alternative anti-CD20 antibodies may be more consistently effective.