The present invention relates to a method for predicting the response of a subject with an autoimmune disease to a B cell-targeting therapy. In particular, the invention relates to predicting the response of a patient by analysing the level of soluble CD23 (sCD23) and B cell activation factor (BAFF) in a sample from said subject. The B cell-targeting therapy can be an anti-CD20 antibody. The autoimmune disease can be systemic lupus erythematosus (SLE). Methods of treatment, compositions and kits are disclosed.