Using a multi-institutional pediatric learning health system to characterize induction therapy for incident lupus nephritis in children

Background: Childhood-onset systemic lupus erythematosus (SLE) poses unique challenges compared to adult-onset SLE, with higher rates of kidney involvement, tissue damage, and mortality. Despite this, data on the effectiveness of immunomodulatory therapies, particularly for lupus nephritis (LN), in pediatric populations are limited. We leveraged a U.S. national learning health system to describe induction treatment in children with LN. Methods: A previously validated LN case-finding algorithm was used to identify patients. Index date was the date of the first LN diagnosis code or the date of the first kidney biopsy, whichever came first. Induction medications included corticosteroids (IV and PO), cytotoxic agents (cyclophosphamide, IV), antimetabolites (mycophenolate mofetil, mycophenolic acid) and biologics (rituximab, belimumab). Results: 575 children with LN were identified. Median age at entry was 15.1 years, 78.6% had evidence of a kidney biopsy, and 56.3% of patients received at least one non-corticosteroid immunosuppressive medication within 6 months of the index date. Prescribed or administered agents during the induction period were: corticosteroids (85%), antimetabolite (61.6%), biologic agent (13.7%), cytotoxic agent (21%). Among the 15% not prescribed or administered corticosteroids, but prescribed or administered an additional immunosuppressive medication, 100% were prescribed or administered antimetabolites. The most common first non-corticosteroid agents were antimetabolites (61.6%), cytotoxic agents (21%) and biologics (13.7%), while biologics were the most common second-line agent. Conclusions: We utilized a validated, sensitive and specific computable phenotype to identify children and adolescents with LN and subsequently characterize their induction therapy regimens, a critical first step in designing future pragmatic clinical trials.