(C-198) Decoding IgA Nephropathy: Large Language Model-Powered Case Report Analysis of Etiologies and Sequential Treatment Strategies in Primary and Secondary Forms

Background: IgA nephropathy (IgAN) is a heterogeneous renal disorder with primary and secondary forms. However, real-world datasets remain limited, particularly for rare or complex secondary IgAN. Published case reports provide a valuable proxy for capturing the natural history and treatment patterns of these rare presentations.

Objectives: This study assessed the feasibility of using a large language model (LLM) to analyze published IgAN case reports for demographic patterns, secondary etiologies, and sequential treatment pathways across geographic regions, primary versus secondary IgAN, and age groups.

Methods: We systematically extracted data from 83 IgAN case reports using the GPT-4-o1 LLM. The LLM system classified cases into primary (absence of underlying cause) and secondary IgAN (presence of systemic disease), analyzed sequential treatment regimens (up to five lines), and visualized using Sankey diagrams. Variations were explored by geographic origin and age.

Results: The cohort included 51 males (61.4%) and 32 females (38.6%), aged 5 to 81 years, with 63.9% identified as East Asian. Sixty-four cases (77.11%) were classified as secondary IgAN. Among secondary cases, autoimmune etiologies (e.g., anti-GBM disease, systemic lupus erythematosus, vasculitis) were observed in 19 cases (29.69%), genetic causes (e.g., Alport syndrome, UMOD mutation) in 10 cases (15.63%), and iatrogenic triggers (e.g., post–COVID-19 vaccination, ICI) in another 10 cases (15.63%). Infectious causes were reported in 5 cases (7.81%), paraneoplastic in 4 cases (6.25%), and both metabolic (e.g., Wilson’s disease) and other causes in 4 (6.25%) and 6 cases (9.38%), respectively. East Asian secondary IgAN patients were typically managed with aggressive immunosuppression—methylprednisolone pulses combined with oral prednisone, cyclophosphamide, mycophenolate mofetil, or mTOR inhibitors, alongside renin–angiotensin system inhibitors—whereas non-Asian cases more often received conservative management. Secondary IgAN generally required more complex, multi-step treatment pathways than primary IgAN. Age-related differences were notable: pediatric patients often received disease-specific therapies (e.g., D-penicillamine for Wilson’s disease), while older patients underwent extensive combination immunosuppression. A minority received targeted agents (e.g., eculizumab) or alternative treatments.

Conclusions: Our study demonstrates that AI/LLM-enabled analysis of case reports can serve as a viable real-world data source for understanding the natural history and treatment patterns of rare diseases like IgAN. The observed variations, influenced by etiology, geographic origin, and age, underscore the need for individualized therapeutic strategies.