Filling the mechanistic and methodological gaps for predicting cytokine release syndrome in humans

Cytokine release syndrome (CRS) is a severe immune-mediated toxicity frequently associated with immunotherapies such as CAR-T cells and bispecific T-cell engagers. Predicting CRS in humans remains difficult because current animal models and in vitro systems do not adequately capture the complex, systemic, and patient-specific nature of pathological inflammation. This perspective highlights key mechanistic and methodological limitations in current immunotoxicology approaches and proposes an immune-related Adverse Outcome Pathway (irAOP) framework adapted for biotherapeutics. The irAOP approach structures CRS as a sequence of interconnected key events (KEs), including immune cell activation, recruitment of pro-inflammatory cells, and excessive cytokine release. These KEs are harmonized from existing AOPs to distinguish pathological inflammation from normal immune responses. Emphasis is placed on downstream effects such as endothelial activation and vascular leakage, which are central to CRS severity. Clinical manifestations - including fever, hypotension, hypoxia, edema, and multi-organ dysfunction - are linked to late-stage KEs, enabling a mechanistic bridge between molecular events and patient outcomes. Insights from COVID-19-related irAOP development and international initiatives like the imSAVAR consortium inform a roadmap for improving CRS prediction. This includes defining context of use, strengthening mechanistic evidence through quantitative key event relationships, and advancing human-relevant New Approach Methodologies (NAMs). By integrating advanced in vitro and ex vivo systems within an irAOP framework, fragmented data can be transformed into coherent, decision-relevant tools. Overall, this strategy supports a shift toward predictive, human-centered approaches for assessing and managing CRS risk in immunotherapy.