AUTHORS
Kyeongmin Kim, PharmD, PhD, Ana Maria Avila Rodriguez, MD2, Matias Eugenio Sanchez, MD3, Damiano Rondelli, MD4, Michael Buege, PharmD5, John Gerard Quigley, MD6, Nadim Mahmud, MD, PhD, MBA7, Karen Sweiss, PharmD8
AFFILIATIONS
1 University of Illinois Chicago, Chicago, Illinois, United States
2 University of Illinois Chicago, Chicago, Illinois, United States
3 University of Illinois Chicago, Chicago, Illinois, United States
4 University of Illinois Chicago, Chicago, Illinois, United States
5 University of Illinois Chicago, Chicago, Illinois, United States
6 University of Illinois Chicago, Chicago, Illinois, United States
7 University of Illinois Chicago, Chicago, Illinois, United States
8 University of Illinois Chicago, Chicago, Illinois, United States
ABSTRACT
Introduction: Bispecific antibodies (BsAbs) have transformed the treatment landscape for relapsed/refractory multiple myeloma (RRMM), yet baseline prognostic markers that identify patients at high risk of severe toxicities or treatment failure remain limited. Neutrophil-to-lymphocyte ratio (NLR), monocyte-to-lymphocyte ratio (MLR), neutrophil-monocyte-to-lymphocyte ratio (NMLR), and prognostic nutritional index (PNI), which utilize readily available peripheral blood laboratory values, reflect host immune, inflammatory and nutritional status and are associated with outcomes in MM. However, their prognostic significance in patients receiving BsAbs remains unclear. Since BsAbs depend on functional T cells to mediate antimyeloma activity, we hypothesized that biomarkers broadly reflecting baseline immune function, inflammation, and nutritional status may have prognostic value in patients receiving BsAbs.
Large federated electronic health record (EHR) platforms such as Epic Cosmos have enormous potential for real-world oncology research but remain underutilized for longitudinal therapeutic outcome studies due to challenges in identifying treatment cohorts and disease-specific outcomes. We therefore developed a scalable analytic framework for longitudinal outcomes research in Epic Cosmos and applied it to identify these specific biomarkers in patients receiving BsAbs.
Methods: A retrospective cohort study was conducted using Epic Cosmos, a network comprising records from >310 million patients across 2,198 hospitals and 49,300 clinics. Adults with MM who received ≥3 administrations of teclistamab, elranatamab, talquetamab, or linvoseltamab before April 1, 2026, and had baseline and post-treatment M-protein measurements were included. Event-free survival (EFS) was defined based on IMWG progression criteria, using serial serum or urine M-protein measurements, a new treatment initiation post-BsAb, new extramedullary disease, or death. Baseline clinical variables, MLR, NLR, NMLR, and PNI were evaluated using Cox proportional hazards regression. A core multivariable model was developed using available clinically relevant baseline covariates, and candidate biomarkers were evaluated for incremental prognostic value beyond the core model.
Results: Among 2,754 eligible patients, 903 had complete clinical and laboratory variables of interest and were included in the analysis. The median age at BsAb initiation was 69 years (range: 31–96), 55.0% were male, and 19.5% were Black. With a median follow-up of 18.8 months (range: 0.03–40.7), 476 EFS events occurred and median EFS was 14.4 months (95% CI: 11.6–16.8). Median baseline MLR, NLR, NMLR, and PNI were 0.5 (range: 0–5.0), 3.0 (range: 0–146.0), 3.7 (range: 0.1–148.1), and 41.5 (range: 18.7–69.5), respectively.
In univariable analyses, higher baseline serum M-protein and LDH, along with lower hemoglobin and platelet count, were associated with shorter EFS, supporting the validity of the cohort and the analytic framework. Higher MLR, NLR, and NMLR, lower eosinophil count, lower PNI, and male sex were also associated with shorter EFS (all P<0.05). A core clinical model was constructed using age, sex, serum M-protein, LDH, hemoglobin, and platelet count based on clinical relevance and the univariate analysis results. Among the evaluated host-related biomarkers, PNI provided the greatest incremental improvement in model performance, with additional improvement following incorporation of MLR. In the final multivariable model, lower PNI (adjusted HR per 1-unit increase: 0.97; 95% CI: 0.96–0.99) and higher MLR (adjusted HR per 1-unit increase: 1.16; 95% CI: 1.03–1.32) remained independently associated with shorter EFS.
Conclusion: To our knowledge, this is the first multicenter real-world analysis demonstrating the feasibility of leveraging de-identified Epic Cosmos data for scalable longitudinal outcomes research in MM through structured ascertainment of treatment exposure and outcomes. PNI and MLR were identified as novel predictors of EFS in patients receiving BsAbs, independent of known clinical prognostic factors. As readily available clinical biomarkers, they may allow improved patient risk stratification for BsAb treatment, particularly identifying those who may have inferior outcomes due to baseline compromised nutritional and/or immune reserve. Additional outcomes, including treatment-related toxicities, will be analyzed under our existing Epic workflow and presented at the conference.
DISCLOSURES
Kyeongmin Kim, PharmD, PhD: No relevant conflicts of interest to declare.