Abstract Background/Aims The BEAT-lupus double-blind randomised trial demonstrated that belimumab (anti-BAFF antibody) reduced the number of severe-flares (compared to placebo) after B-cell depletion (rituximab) in patients with systemic lupus erythematosus (SLE)(1). We longitudinally immune profiled patients recruited to the trial to identify which patients would benefit most from belimumab after rituximab combination therapy. Methods We constructed two models (because the number of patients who provided serum/blood RNA samples and peripheral blood mononuclear cells differed) using baseline data to predict a major clinical response (MCR)(2) to belimumab after rituximab at 52-weeks utilising conventional and machine learning approaches: A) Model 1: serum autoantibodies and cytokines, peripheral blood RNA expression, clinical data, and routinely collected laboratory-markers (10 responders, 11 non-responders). B) Model 2: B-cell flow-cytometry (8 responders, 8 non-responders). The emerging key variables were then used as effect modifiers comparing response to belimumab and placebo after rituximab. Results Baseline serum IgA2 anti-dsDNA antibody levels emerged from model 1 as the key predictor of MCR with an AUROC (area under the curve of the receiver operator characteristic) of 0.81 (95% CI: 0.70-0.96) in predicting an MCR by 52-weeks in the belimumab arm. An optimal cut point of serum IgA2 anti-dsDNA antibody (10.7 AU/arbitrary unit), derived from AUROC, was applied as an effect modifier which improved the delta between response to belimumab vs placebo from 13% (without the effect modifier) to 48% (95% CI: 10-70; p = 0.021) (Belimumab - 64% [9 responders, 5 non-responders], placebo - 16% [2 responders, 10 non-responders]). Belimumab significantly reduced serum IgA2 anti-dsDNA antibody levels compared to placebo at 52-weeks (difference: 20 AU; 95% CI: 13-27; p < 0.001) only in those patients who achieved an MCR, falling by 60% from baseline (difference: 23 AU; 95% CI: 11-35; p < 0.001, compared to non-response). Prediction model 2 identified the frequency of T-bet+CD11c+DNM-2 (double negative memory B-cell, type-2: CD11c+T-bet+CD38-CD27-IgD-CD21-CD24-) as the strongest predictor of MCR by 52-weeks in the belimumab arm with an AUROC of 1.00 (95% CI: 1.00-1.00). Using T-bet+CD11c+DNM-2 as an effect modifier (optimal cut point: 14.2% of total B-cells) the delta improved from 13% to 71% (95% CI: 23-92; p = 0.007) (Belimumab - 100% [8 responders, 0 non-responders], placebo - 29% [2 responders, 5 non-responders]). The frequency of T-bet+CD11c+DNM-2 was significantly lower in the belimumab arm at 52-weeks compared to placebo (difference: 6%; 95% CI: 3-11; p = 0.001), and dropped by 48% from baseline in belimumab responders (difference with non-responders: 11%; 95% CI: 6-15; p < 0.001). Conclusion High baseline serum IgA2 anti-dsDNA antibody and peripheral blood T-bet+CD11c+DNM-2 B-cell frequency could guide patient selection for belimumab after rituximab combination therapy thereby improving outcomes and access to targeted therapies for patients with SLE. Exploration of the underlying mechanisms and confirmation of these results including combining the two prediction models in a larger clinical trial are required. Reference 1. Shipa M, et al. Annals of Internal Medicine. 2021;174:1647-57. 2. McDonald S, et al. Lupus Sci Med. 2022;9:e000584. Disclosure M. Shipa: Grants/research support; Versus arthritis, LUPUS-UK. M. Parvaz: None. D. Nguyen: None. L. Santos: None. D.A. Isenberg: Consultancies; Astra Zeneca, Eli Lilly, Merck Serono, Servier and UCB. C. Gordon: Consultancies; Center for Disease Control and Prevention, AbbVie, Amgen, Astra-Zeneca, EMD Serono, MGP, Sanofi and UCB. M.R. Ehrenstein: Consultancies; GSK. Grants/research support; Versus arthritis.