Abstract 2439: Predictive modeling of venetoclax/azacitidine response in acute myeloid leukemia using inflammatory cytokine profiling and clinical data from the NCT04267081 trial cohort

Abstract Cytokines in acute myeloid leukemia (AML) remodel the bone marrow microenvironment, drive disease progression, and contribute to chemoresistance. Cytokines such as IL-8 and CXCL12 promote AML cell survival and drug resistance, while some others influence therapeutic response, making cytokine signaling a key factor for predicting outcomes. Venetoclax (VEN) plus azacitidine (AZA) is standard therapy for older patients or those unable to receive intensive chemotherapy, yet responses vary widely. To address this, we developed predictive models for VEN/AZA response and survival using cytokine and clinical data. Cytokine profiling was performed on samples from 85 AML patients in the VenEx trial (NCT04267081). Using the Olink Target 48 platform, the protein expression of 44 inflammatory cytokines was measured and normalized for downstream analysis. Clinical measurements and VEN/AZA treatment responses were collected, including 68 responders and 17 non-responders. T-test was used to compare the difference in cytokine expression between responders and non-responders. Elastic net regression models incorporating cytokine profiles and clinical variables were used to predict patient response and identified key cytokines associated with treatment response. To assess survival risk, we performed univariate Cox regression for each cytokine and generated Kaplan-Meier curves by stratifying patients into high-expression (above the third quartile) and low-expression groups. Comparing the responders vs non-responders, we found potential biomarkers, including IL13 had higher expression in responders, while IL18, IL10, VEGFA, OSM, and CSF1 were expressed higher in non-responders. Using cytokine profiling to predict patient response, the model achieved an accuracy of 0.74 using 5-fold cross validation. Adding cell population measurements in the model, the accuracy increased to 0.83. All potential biomarkers are the top predictors for ven/aza response. Beyond that, we also identified TSLP, IL15, IL1B and TNFSF12 are among the top 10 predictors. Cox’s proportional hazard model identified cytokines linked to poor survival, including CSF1, CCL3, CCL19, IL18, HGF, IL1B and OSM (ranked by hazard ratio). Stratification by FAB type revealed distinct cytokine associations with survival and drug response, TNFSF10 was siginificantly higher in non-responders in FAB M0-M2 (n=51; 41 responders, 10 non-responders). Integrating cytokine and clinical data enables robust prediction of VEN/AZA response and survival in AML. Key cytokines such as CSF1, IL18 and IL13 may guide risk stratification and personalized therapy. Future work will map pathways linked to these cytokines and explore their roles in survival and treatment response. Citation Format: Ankita Srivastava, Bahar Tercan, David L. Gibbs, Heikki Kuusanmäki, Mika Kontro, Caroline A. Heckman, Guangrong Qin. Predictive modeling of venetoclax/azacitidine response in acute myeloid leukemia using inflammatory cytokine profiling and clinical data from the NCT04267081 trial cohort [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2439.