Advancing toxicity prediction in CAR-T cell therapy

Despite recent advances in the field, Chimeric Antigen Receptor T (CAR-T) cell toxicities remain a significant burden throughout both the early and late post-infusion periods, affecting patient outcomes and complicating clinical management. Several risk factors for these side effects have been identified and multiple predictive scores and models have been developed, focusing mainly on clinical and inflammatory parameters assessed before or after CAR-T cell infusion. However, the majority of these scores have been derived and validated in retrospective cohorts, with external validation available only for a few of them predicting only CRS, ICANS or ICAHT risk. Several newly described toxicities still lack any predictive tool, and even the existing scores for CRS and ICANS have not yet been validated across all CAR-T disease indications, limiting their generalizability. The CAR-HEMATOTOX (CAR-HT), initially developed for ICAHT risk prediction, remains the most extensively validated tool to date. Despite this, no predictive model has yet been prospectively validated within interventional trials randomizing patients according to predicted risk. This minireview aims to provide an up-to-date overview of currently validated toxicity prediction scores in CAR-T cell therapy, along with prediction tools under development, future research directions and their potential integration into routine clinical practice to guide risk-adapted patient management.