CAR T-Cell Immunotherapy in Large B-Cell Lymphoma: Viral Etiology, Stem Cell Engineering, and AI/ML-Guided Gene Therapy for Cytokine Release Syndrome (CRS) and Immune Effector Cell-Associated Neurotoxicity Syndrome (ICANS)

Abstract

This hypothetical study investigated CAR T-cell immunotherapy in the context of viral status, CAR T-cell and stem-cell engineering characteristics, and AI/ML prediction of CRS and ICANS, focused on LBCL. Using statistical analyses and machine-learning designs, 120 adult patients undergoing CD19-targeted CAR T-cell therapy were simulated in a hypothetical retrospective study, based on clinical, viral, inflammatory, treatment, and cellular parameters. The results indicated that CRS (77.5%) and ICANS (44.2%) were more common, and that IL-6, ferritin, CRP, D-dimer, disease burden, and CAR T-cell expansion levels were higher in participants who experienced treatment-related side effects. The presence of viral markers was also correlated with severe CRS and ICANS. Gradient Boosting had the best hypothetical predictive performance out of the evaluated AI/ML models, with an AUC of 0.95 for CRS and 0.93 for ICANS, with IL-6, ferritin, CAR T-cell expansion, disease burden and CRP identified as important predictors. The study proposes that combining clinical, viral, laboratory, and cellular parameters with AI/ML could help stratify early toxicity and enable more personalized CAR T-cell therapy, but this remains a conjecture that needs to be assessed in larger prospective, multicenter, and independently validated clinical studies.