525P Robustness of LLM-assisted network meta-analysis in second-line or later HER2-positive metastatic breast cancer: A cross-model validation study

In HER2-positive metastatic breast cancer (HER2+ mBC), many regimens are used in second-line or later (2L+), yet direct comparisons are scarce. Network meta-analysis (NMA) enables indirect comparisons, but results are very sensitive to study selection. We developed a large language model (LLM)-assisted pipeline for trial selection and assessed the concordance and robustness of NMA results.