Predicting Remission in Antiphospholipase A2 Receptor Antibody-Associated Membranous Nephropathy

Key Points In phospholipase A2 receptor-membranous nephropathy, clinical variables and antibody levels over 3 or 6 months of treatment were predictors of remission at 1 year. Prediction models at 3 and 6 months had similar performance predicting remission, with the 3-month model allowing for earlier assessment of response. The 3-month and 6-month prediction models could be applied in those treated with supportive therapy, rituximab, calcineurin inhibitors, or cyclophosphamide. Background In patients with antiphospholipase A2 receptor antibody–associated membranous nephropathy, there is currently no accepted method to predict an individual's probability of remission after treatment with immunosuppression or supportive therapy using changes in antibody levels and clinical variables during the first 3–6 months of therapy. Methods Using a cohort of 187 patients from the Glomérulopathie extramembraneuse rituximab, Membranous Nephropathy Trial of Rituximab, and Sequential Treatment with Tacrolimus and Rituximab Versus Alternating Corticosteroids and Cyclophosphamide in primary Membranous Nephropathy clinical trials with antibody levels at baseline ≥14 RU/ml, we derived logistic regression models to predict proteinuria remission at 12 months that can be used at baseline or after 3 or 6 months of treatment. Treatment exposures in the trials included supportive therapy, rituximab, calcineurin inhibitors, and cyclophosphamide. Predictors in the models included male sex and baseline and changes in serum albumin, proteinuria, and antibody levels, with or without changes in eGFR. Results Proteinuria remission at 12 months was achieved in 107 patients. Compared with the model at baseline, the 3-month and 6-month models had better model fit with lower Akaike information criterion (186/158 versus 225) and higher R 2 (52.7%/62.4% versus 25.8%), better discrimination with higher C-statistic (0.87 and 0.91 versus 0.75, P < 0.001), and better calibration with lower integrated calibration index (0.89%/2.22% versus 2.51%). The 3-month and 6-month models had no consistent difference in prediction performance, and decision curve analysis demonstrated similar net benefit for treatment decisions based on either model up to a threshold probability of 31%. Prediction performance was similar after internal validation using optimism correction. Prediction performance was maintained within subgroups of different treatment regimens, including supportive therapy, rituximab, calcineurin inhibitors, and cyclophosphamide. Conclusions Either the 3-month or 6-month models can be used in patients with antiphospholipase A2 receptor antibody associated membranous nephropathy after 3 or 6 months of treatment with a variety of immunosuppression or supportive therapy to predict remission status at 12 months.