Development and validation of an automated computational approach to grade immune effector cell-associated hematotoxicity.

Hematologic toxicity frequently complicates chimeric antigen receptor (CAR) T-cell therapy, resulting in significant morbidity and mortality. In an effort to standardize reporting, the European Hematology Association (EHA) and European Society of Blood and Marrow Transplantation (EBMT) devised the immune effector cell-associated hematotoxicity (ICAHT) grading system, distinguishing between early (day 0-30) and late (after day +30) events based on neutropenia depth and duration. However, manual implementation of ICAHT grading criteria is time-consuming and susceptible to subjectivity and error. To address these challenges, we introduce a novel computational approach, utilizing the R programming language, to automate early and late ICAHT grading. Given the complexities of early ICAHT grading, we benchmarked our approach both manually and computationally in two independent cohorts totaling 1251 patients. Our computational approach offers significant implications by streamlining grading processes, reducing manual time and effort, and promoting standardization across varied clinical settings. We provide this tool to the scientific community alongside a comprehensive implementation guide, fostering its widespread adoption and enhancing reporting consistency for ICAHT.